setup.c 28 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159
  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. #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. struct boot_params boot_params;
  130. /*
  131. * Machine setup..
  132. */
  133. static struct resource data_resource = {
  134. .name = "Kernel data",
  135. .start = 0,
  136. .end = 0,
  137. .flags = IORESOURCE_BUSY | IORESOURCE_MEM
  138. };
  139. static struct resource code_resource = {
  140. .name = "Kernel code",
  141. .start = 0,
  142. .end = 0,
  143. .flags = IORESOURCE_BUSY | IORESOURCE_MEM
  144. };
  145. static struct resource bss_resource = {
  146. .name = "Kernel bss",
  147. .start = 0,
  148. .end = 0,
  149. .flags = IORESOURCE_BUSY | IORESOURCE_MEM
  150. };
  151. #ifdef CONFIG_X86_32
  152. /* cpu data as detected by the assembly code in head.S */
  153. struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1};
  154. /* common cpu data for all cpus */
  155. struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1};
  156. EXPORT_SYMBOL(boot_cpu_data);
  157. unsigned int def_to_bigsmp;
  158. /* for MCA, but anyone else can use it if they want */
  159. unsigned int machine_id;
  160. unsigned int machine_submodel_id;
  161. unsigned int BIOS_revision;
  162. struct apm_info apm_info;
  163. EXPORT_SYMBOL(apm_info);
  164. #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
  165. defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
  166. struct ist_info ist_info;
  167. EXPORT_SYMBOL(ist_info);
  168. #else
  169. struct ist_info ist_info;
  170. #endif
  171. #else
  172. struct cpuinfo_x86 boot_cpu_data __read_mostly = {
  173. .x86_phys_bits = MAX_PHYSMEM_BITS,
  174. };
  175. EXPORT_SYMBOL(boot_cpu_data);
  176. #endif
  177. #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
  178. unsigned long mmu_cr4_features;
  179. #else
  180. unsigned long mmu_cr4_features = X86_CR4_PAE;
  181. #endif
  182. /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
  183. int bootloader_type, bootloader_version;
  184. /*
  185. * Setup options
  186. */
  187. struct screen_info screen_info;
  188. EXPORT_SYMBOL(screen_info);
  189. struct edid_info edid_info;
  190. EXPORT_SYMBOL_GPL(edid_info);
  191. extern int root_mountflags;
  192. unsigned long saved_video_mode;
  193. #define RAMDISK_IMAGE_START_MASK 0x07FF
  194. #define RAMDISK_PROMPT_FLAG 0x8000
  195. #define RAMDISK_LOAD_FLAG 0x4000
  196. static char __initdata command_line[COMMAND_LINE_SIZE];
  197. #ifdef CONFIG_CMDLINE_BOOL
  198. static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
  199. #endif
  200. #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
  201. struct edd edd;
  202. #ifdef CONFIG_EDD_MODULE
  203. EXPORT_SYMBOL(edd);
  204. #endif
  205. /**
  206. * copy_edd() - Copy the BIOS EDD information
  207. * from boot_params into a safe place.
  208. *
  209. */
  210. static inline void __init copy_edd(void)
  211. {
  212. memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
  213. sizeof(edd.mbr_signature));
  214. memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
  215. edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
  216. edd.edd_info_nr = boot_params.eddbuf_entries;
  217. }
  218. #else
  219. static inline void __init copy_edd(void)
  220. {
  221. }
  222. #endif
  223. void * __init extend_brk(size_t size, size_t align)
  224. {
  225. size_t mask = align - 1;
  226. void *ret;
  227. BUG_ON(_brk_start == 0);
  228. BUG_ON(align & mask);
  229. _brk_end = (_brk_end + mask) & ~mask;
  230. BUG_ON((char *)(_brk_end + size) > __brk_limit);
  231. ret = (void *)_brk_end;
  232. _brk_end += size;
  233. memset(ret, 0, size);
  234. return ret;
  235. }
  236. #ifdef CONFIG_X86_64
  237. static void __init init_gbpages(void)
  238. {
  239. if (direct_gbpages && cpu_has_gbpages)
  240. printk(KERN_INFO "Using GB pages for direct mapping\n");
  241. else
  242. direct_gbpages = 0;
  243. }
  244. #else
  245. static inline void init_gbpages(void)
  246. {
  247. }
  248. static void __init cleanup_highmap(void)
  249. {
  250. }
  251. #endif
  252. static void __init reserve_brk(void)
  253. {
  254. if (_brk_end > _brk_start)
  255. memblock_reserve(__pa(_brk_start),
  256. __pa(_brk_end) - __pa(_brk_start));
  257. /* Mark brk area as locked down and no longer taking any
  258. new allocations */
  259. _brk_start = 0;
  260. }
  261. #ifdef CONFIG_BLK_DEV_INITRD
  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 = boot_params.hdr.ramdisk_image;
  267. u64 ramdisk_size = boot_params.hdr.ramdisk_size;
  268. u64 area_size = PAGE_ALIGN(ramdisk_size);
  269. u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
  270. u64 ramdisk_here;
  271. unsigned long slop, clen, mapaddr;
  272. char *p, *q;
  273. /* We need to move the initrd down into lowmem */
  274. ramdisk_here = memblock_find_in_range(0, end_of_lowmem, area_size,
  275. PAGE_SIZE);
  276. if (!ramdisk_here)
  277. panic("Cannot find place for new RAMDISK of size %lld\n",
  278. ramdisk_size);
  279. /* Note: this includes all the lowmem currently occupied by
  280. the initrd, we rely on that fact to keep the data intact. */
  281. memblock_reserve(ramdisk_here, area_size);
  282. initrd_start = ramdisk_here + PAGE_OFFSET;
  283. initrd_end = initrd_start + ramdisk_size;
  284. printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
  285. ramdisk_here, ramdisk_here + ramdisk_size - 1);
  286. q = (char *)initrd_start;
  287. /* Copy any lowmem portion of the initrd */
  288. if (ramdisk_image < end_of_lowmem) {
  289. clen = end_of_lowmem - ramdisk_image;
  290. p = (char *)__va(ramdisk_image);
  291. memcpy(q, p, clen);
  292. q += clen;
  293. ramdisk_image += clen;
  294. ramdisk_size -= clen;
  295. }
  296. /* Copy the highmem portion of the initrd */
  297. while (ramdisk_size) {
  298. slop = ramdisk_image & ~PAGE_MASK;
  299. clen = ramdisk_size;
  300. if (clen > MAX_MAP_CHUNK-slop)
  301. clen = MAX_MAP_CHUNK-slop;
  302. mapaddr = ramdisk_image & PAGE_MASK;
  303. p = early_memremap(mapaddr, clen+slop);
  304. memcpy(q, p+slop, clen);
  305. early_iounmap(p, clen+slop);
  306. q += clen;
  307. ramdisk_image += clen;
  308. ramdisk_size -= clen;
  309. }
  310. /* high pages is not converted by early_res_to_bootmem */
  311. ramdisk_image = boot_params.hdr.ramdisk_image;
  312. ramdisk_size = boot_params.hdr.ramdisk_size;
  313. printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
  314. " [mem %#010llx-%#010llx]\n",
  315. ramdisk_image, ramdisk_image + ramdisk_size - 1,
  316. ramdisk_here, ramdisk_here + ramdisk_size - 1);
  317. }
  318. static void __init reserve_initrd(void)
  319. {
  320. /* Assume only end is not page aligned */
  321. u64 ramdisk_image = boot_params.hdr.ramdisk_image;
  322. u64 ramdisk_size = boot_params.hdr.ramdisk_size;
  323. u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
  324. u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
  325. if (!boot_params.hdr.type_of_loader ||
  326. !ramdisk_image || !ramdisk_size)
  327. return; /* No initrd provided by bootloader */
  328. initrd_start = 0;
  329. if (ramdisk_size >= (end_of_lowmem>>1)) {
  330. panic("initrd too large to handle, "
  331. "disabling initrd (%lld needed, %lld available)\n",
  332. ramdisk_size, end_of_lowmem>>1);
  333. }
  334. printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
  335. ramdisk_end - 1);
  336. if (ramdisk_end <= end_of_lowmem) {
  337. /* All in lowmem, easy case */
  338. /*
  339. * don't need to reserve again, already reserved early
  340. * in i386_start_kernel
  341. */
  342. initrd_start = ramdisk_image + PAGE_OFFSET;
  343. initrd_end = initrd_start + ramdisk_size;
  344. return;
  345. }
  346. relocate_initrd();
  347. memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
  348. }
  349. #else
  350. static void __init reserve_initrd(void)
  351. {
  352. }
  353. #endif /* CONFIG_BLK_DEV_INITRD */
  354. static void __init parse_setup_data(void)
  355. {
  356. struct setup_data *data;
  357. u64 pa_data;
  358. if (boot_params.hdr.version < 0x0209)
  359. return;
  360. pa_data = boot_params.hdr.setup_data;
  361. while (pa_data) {
  362. u32 data_len, map_len;
  363. map_len = max(PAGE_SIZE - (pa_data & ~PAGE_MASK),
  364. (u64)sizeof(struct setup_data));
  365. data = early_memremap(pa_data, map_len);
  366. data_len = data->len + sizeof(struct setup_data);
  367. if (data_len > map_len) {
  368. early_iounmap(data, map_len);
  369. data = early_memremap(pa_data, data_len);
  370. map_len = data_len;
  371. }
  372. switch (data->type) {
  373. case SETUP_E820_EXT:
  374. parse_e820_ext(data);
  375. break;
  376. case SETUP_DTB:
  377. add_dtb(pa_data);
  378. break;
  379. default:
  380. break;
  381. }
  382. pa_data = data->next;
  383. early_iounmap(data, map_len);
  384. }
  385. }
  386. static void __init e820_reserve_setup_data(void)
  387. {
  388. struct setup_data *data;
  389. u64 pa_data;
  390. int found = 0;
  391. if (boot_params.hdr.version < 0x0209)
  392. return;
  393. pa_data = boot_params.hdr.setup_data;
  394. while (pa_data) {
  395. data = early_memremap(pa_data, sizeof(*data));
  396. e820_update_range(pa_data, sizeof(*data)+data->len,
  397. E820_RAM, E820_RESERVED_KERN);
  398. found = 1;
  399. pa_data = data->next;
  400. early_iounmap(data, sizeof(*data));
  401. }
  402. if (!found)
  403. return;
  404. sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
  405. memcpy(&e820_saved, &e820, sizeof(struct e820map));
  406. printk(KERN_INFO "extended physical RAM map:\n");
  407. e820_print_map("reserve setup_data");
  408. }
  409. static void __init memblock_x86_reserve_range_setup_data(void)
  410. {
  411. struct setup_data *data;
  412. u64 pa_data;
  413. if (boot_params.hdr.version < 0x0209)
  414. return;
  415. pa_data = boot_params.hdr.setup_data;
  416. while (pa_data) {
  417. data = early_memremap(pa_data, sizeof(*data));
  418. memblock_reserve(pa_data, sizeof(*data) + data->len);
  419. pa_data = data->next;
  420. early_iounmap(data, sizeof(*data));
  421. }
  422. }
  423. /*
  424. * --------- Crashkernel reservation ------------------------------
  425. */
  426. #ifdef CONFIG_KEXEC
  427. /*
  428. * Keep the crash kernel below this limit. On 32 bits earlier kernels
  429. * would limit the kernel to the low 512 MiB due to mapping restrictions.
  430. * On 64 bits, kexec-tools currently limits us to 896 MiB; increase this
  431. * limit once kexec-tools are fixed.
  432. */
  433. #ifdef CONFIG_X86_32
  434. # define CRASH_KERNEL_ADDR_MAX (512 << 20)
  435. #else
  436. # define CRASH_KERNEL_ADDR_MAX (896 << 20)
  437. #endif
  438. static void __init reserve_crashkernel(void)
  439. {
  440. unsigned long long total_mem;
  441. unsigned long long crash_size, crash_base;
  442. int ret;
  443. total_mem = memblock_phys_mem_size();
  444. ret = parse_crashkernel(boot_command_line, total_mem,
  445. &crash_size, &crash_base);
  446. if (ret != 0 || crash_size <= 0)
  447. return;
  448. /* 0 means: find the address automatically */
  449. if (crash_base <= 0) {
  450. const unsigned long long alignment = 16<<20; /* 16M */
  451. /*
  452. * kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
  453. */
  454. crash_base = memblock_find_in_range(alignment,
  455. CRASH_KERNEL_ADDR_MAX, crash_size, alignment);
  456. if (!crash_base) {
  457. pr_info("crashkernel reservation failed - No suitable area found.\n");
  458. return;
  459. }
  460. } else {
  461. unsigned long long start;
  462. start = memblock_find_in_range(crash_base,
  463. crash_base + crash_size, crash_size, 1<<20);
  464. if (start != crash_base) {
  465. pr_info("crashkernel reservation failed - memory is in use.\n");
  466. return;
  467. }
  468. }
  469. memblock_reserve(crash_base, crash_size);
  470. printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
  471. "for crashkernel (System RAM: %ldMB)\n",
  472. (unsigned long)(crash_size >> 20),
  473. (unsigned long)(crash_base >> 20),
  474. (unsigned long)(total_mem >> 20));
  475. crashk_res.start = crash_base;
  476. crashk_res.end = crash_base + crash_size - 1;
  477. insert_resource(&iomem_resource, &crashk_res);
  478. }
  479. #else
  480. static void __init reserve_crashkernel(void)
  481. {
  482. }
  483. #endif
  484. static struct resource standard_io_resources[] = {
  485. { .name = "dma1", .start = 0x00, .end = 0x1f,
  486. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  487. { .name = "pic1", .start = 0x20, .end = 0x21,
  488. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  489. { .name = "timer0", .start = 0x40, .end = 0x43,
  490. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  491. { .name = "timer1", .start = 0x50, .end = 0x53,
  492. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  493. { .name = "keyboard", .start = 0x60, .end = 0x60,
  494. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  495. { .name = "keyboard", .start = 0x64, .end = 0x64,
  496. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  497. { .name = "dma page reg", .start = 0x80, .end = 0x8f,
  498. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  499. { .name = "pic2", .start = 0xa0, .end = 0xa1,
  500. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  501. { .name = "dma2", .start = 0xc0, .end = 0xdf,
  502. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  503. { .name = "fpu", .start = 0xf0, .end = 0xff,
  504. .flags = IORESOURCE_BUSY | IORESOURCE_IO }
  505. };
  506. void __init reserve_standard_io_resources(void)
  507. {
  508. int i;
  509. /* request I/O space for devices used on all i[345]86 PCs */
  510. for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
  511. request_resource(&ioport_resource, &standard_io_resources[i]);
  512. }
  513. static __init void reserve_ibft_region(void)
  514. {
  515. unsigned long addr, size = 0;
  516. addr = find_ibft_region(&size);
  517. if (size)
  518. memblock_reserve(addr, size);
  519. }
  520. static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
  521. static bool __init snb_gfx_workaround_needed(void)
  522. {
  523. int i;
  524. u16 vendor, devid;
  525. static const u16 snb_ids[] = {
  526. 0x0102,
  527. 0x0112,
  528. 0x0122,
  529. 0x0106,
  530. 0x0116,
  531. 0x0126,
  532. 0x010a,
  533. };
  534. /* Assume no if something weird is going on with PCI */
  535. if (!early_pci_allowed())
  536. return false;
  537. vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
  538. if (vendor != 0x8086)
  539. return false;
  540. devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
  541. for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
  542. if (devid == snb_ids[i])
  543. return true;
  544. return false;
  545. }
  546. /*
  547. * Sandy Bridge graphics has trouble with certain ranges, exclude
  548. * them from allocation.
  549. */
  550. static void __init trim_snb_memory(void)
  551. {
  552. static const unsigned long bad_pages[] = {
  553. 0x20050000,
  554. 0x20110000,
  555. 0x20130000,
  556. 0x20138000,
  557. 0x40004000,
  558. };
  559. int i;
  560. if (!snb_gfx_workaround_needed())
  561. return;
  562. printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
  563. /*
  564. * Reserve all memory below the 1 MB mark that has not
  565. * already been reserved.
  566. */
  567. memblock_reserve(0, 1<<20);
  568. for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
  569. if (memblock_reserve(bad_pages[i], PAGE_SIZE))
  570. printk(KERN_WARNING "failed to reserve 0x%08lx\n",
  571. bad_pages[i]);
  572. }
  573. }
  574. /*
  575. * Here we put platform-specific memory range workarounds, i.e.
  576. * memory known to be corrupt or otherwise in need to be reserved on
  577. * specific platforms.
  578. *
  579. * If this gets used more widely it could use a real dispatch mechanism.
  580. */
  581. static void __init trim_platform_memory_ranges(void)
  582. {
  583. trim_snb_memory();
  584. }
  585. static void __init trim_bios_range(void)
  586. {
  587. /*
  588. * A special case is the first 4Kb of memory;
  589. * This is a BIOS owned area, not kernel ram, but generally
  590. * not listed as such in the E820 table.
  591. *
  592. * This typically reserves additional memory (64KiB by default)
  593. * since some BIOSes are known to corrupt low memory. See the
  594. * Kconfig help text for X86_RESERVE_LOW.
  595. */
  596. e820_update_range(0, ALIGN(reserve_low, PAGE_SIZE),
  597. E820_RAM, E820_RESERVED);
  598. /*
  599. * special case: Some BIOSen report the PC BIOS
  600. * area (640->1Mb) as ram even though it is not.
  601. * take them out.
  602. */
  603. e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
  604. sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
  605. }
  606. static int __init parse_reservelow(char *p)
  607. {
  608. unsigned long long size;
  609. if (!p)
  610. return -EINVAL;
  611. size = memparse(p, &p);
  612. if (size < 4096)
  613. size = 4096;
  614. if (size > 640*1024)
  615. size = 640*1024;
  616. reserve_low = size;
  617. return 0;
  618. }
  619. early_param("reservelow", parse_reservelow);
  620. /*
  621. * Determine if we were loaded by an EFI loader. If so, then we have also been
  622. * passed the efi memmap, systab, etc., so we should use these data structures
  623. * for initialization. Note, the efi init code path is determined by the
  624. * global efi_enabled. This allows the same kernel image to be used on existing
  625. * systems (with a traditional BIOS) as well as on EFI systems.
  626. */
  627. /*
  628. * setup_arch - architecture-specific boot-time initializations
  629. *
  630. * Note: On x86_64, fixmaps are ready for use even before this is called.
  631. */
  632. void __init setup_arch(char **cmdline_p)
  633. {
  634. #ifdef CONFIG_X86_32
  635. memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
  636. visws_early_detect();
  637. /*
  638. * copy kernel address range established so far and switch
  639. * to the proper swapper page table
  640. */
  641. clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY,
  642. initial_page_table + KERNEL_PGD_BOUNDARY,
  643. KERNEL_PGD_PTRS);
  644. load_cr3(swapper_pg_dir);
  645. __flush_tlb_all();
  646. #else
  647. printk(KERN_INFO "Command line: %s\n", boot_command_line);
  648. #endif
  649. /*
  650. * If we have OLPC OFW, we might end up relocating the fixmap due to
  651. * reserve_top(), so do this before touching the ioremap area.
  652. */
  653. olpc_ofw_detect();
  654. early_trap_init();
  655. early_cpu_init();
  656. early_ioremap_init();
  657. setup_olpc_ofw_pgd();
  658. ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
  659. screen_info = boot_params.screen_info;
  660. edid_info = boot_params.edid_info;
  661. #ifdef CONFIG_X86_32
  662. apm_info.bios = boot_params.apm_bios_info;
  663. ist_info = boot_params.ist_info;
  664. if (boot_params.sys_desc_table.length != 0) {
  665. machine_id = boot_params.sys_desc_table.table[0];
  666. machine_submodel_id = boot_params.sys_desc_table.table[1];
  667. BIOS_revision = boot_params.sys_desc_table.table[2];
  668. }
  669. #endif
  670. saved_video_mode = boot_params.hdr.vid_mode;
  671. bootloader_type = boot_params.hdr.type_of_loader;
  672. if ((bootloader_type >> 4) == 0xe) {
  673. bootloader_type &= 0xf;
  674. bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
  675. }
  676. bootloader_version = bootloader_type & 0xf;
  677. bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
  678. #ifdef CONFIG_BLK_DEV_RAM
  679. rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
  680. rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
  681. rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
  682. #endif
  683. #ifdef CONFIG_EFI
  684. if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
  685. "EL32", 4)) {
  686. efi_enabled = 1;
  687. efi_64bit = false;
  688. } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
  689. "EL64", 4)) {
  690. efi_enabled = 1;
  691. efi_64bit = true;
  692. }
  693. if (efi_enabled && efi_memblock_x86_reserve_range())
  694. efi_enabled = 0;
  695. #endif
  696. x86_init.oem.arch_setup();
  697. iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
  698. setup_memory_map();
  699. parse_setup_data();
  700. /* update the e820_saved too */
  701. e820_reserve_setup_data();
  702. copy_edd();
  703. if (!boot_params.hdr.root_flags)
  704. root_mountflags &= ~MS_RDONLY;
  705. init_mm.start_code = (unsigned long) _text;
  706. init_mm.end_code = (unsigned long) _etext;
  707. init_mm.end_data = (unsigned long) _edata;
  708. init_mm.brk = _brk_end;
  709. code_resource.start = virt_to_phys(_text);
  710. code_resource.end = virt_to_phys(_etext)-1;
  711. data_resource.start = virt_to_phys(_etext);
  712. data_resource.end = virt_to_phys(_edata)-1;
  713. bss_resource.start = virt_to_phys(&__bss_start);
  714. bss_resource.end = virt_to_phys(&__bss_stop)-1;
  715. #ifdef CONFIG_CMDLINE_BOOL
  716. #ifdef CONFIG_CMDLINE_OVERRIDE
  717. strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
  718. #else
  719. if (builtin_cmdline[0]) {
  720. /* append boot loader cmdline to builtin */
  721. strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
  722. strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
  723. strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
  724. }
  725. #endif
  726. #endif
  727. strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
  728. *cmdline_p = command_line;
  729. /*
  730. * x86_configure_nx() is called before parse_early_param() to detect
  731. * whether hardware doesn't support NX (so that the early EHCI debug
  732. * console setup can safely call set_fixmap()). It may then be called
  733. * again from within noexec_setup() during parsing early parameters
  734. * to honor the respective command line option.
  735. */
  736. x86_configure_nx();
  737. parse_early_param();
  738. x86_report_nx();
  739. /* after early param, so could get panic from serial */
  740. memblock_x86_reserve_range_setup_data();
  741. if (acpi_mps_check()) {
  742. #ifdef CONFIG_X86_LOCAL_APIC
  743. disable_apic = 1;
  744. #endif
  745. setup_clear_cpu_cap(X86_FEATURE_APIC);
  746. }
  747. #ifdef CONFIG_PCI
  748. if (pci_early_dump_regs)
  749. early_dump_pci_devices();
  750. #endif
  751. finish_e820_parsing();
  752. if (efi_enabled)
  753. efi_init();
  754. dmi_scan_machine();
  755. /*
  756. * VMware detection requires dmi to be available, so this
  757. * needs to be done after dmi_scan_machine, for the BP.
  758. */
  759. init_hypervisor_platform();
  760. x86_init.resources.probe_roms();
  761. /* after parse_early_param, so could debug it */
  762. insert_resource(&iomem_resource, &code_resource);
  763. insert_resource(&iomem_resource, &data_resource);
  764. insert_resource(&iomem_resource, &bss_resource);
  765. trim_bios_range();
  766. #ifdef CONFIG_X86_32
  767. if (ppro_with_ram_bug()) {
  768. e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
  769. E820_RESERVED);
  770. sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
  771. printk(KERN_INFO "fixed physical RAM map:\n");
  772. e820_print_map("bad_ppro");
  773. }
  774. #else
  775. early_gart_iommu_check();
  776. #endif
  777. /*
  778. * partially used pages are not usable - thus
  779. * we are rounding upwards:
  780. */
  781. max_pfn = e820_end_of_ram_pfn();
  782. /* update e820 for memory not covered by WB MTRRs */
  783. mtrr_bp_init();
  784. if (mtrr_trim_uncached_memory(max_pfn))
  785. max_pfn = e820_end_of_ram_pfn();
  786. #ifdef CONFIG_X86_32
  787. /* max_low_pfn get updated here */
  788. find_low_pfn_range();
  789. #else
  790. num_physpages = max_pfn;
  791. check_x2apic();
  792. /* How many end-of-memory variables you have, grandma! */
  793. /* need this before calling reserve_initrd */
  794. if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
  795. max_low_pfn = e820_end_of_low_ram_pfn();
  796. else
  797. max_low_pfn = max_pfn;
  798. high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
  799. #endif
  800. /*
  801. * Find and reserve possible boot-time SMP configuration:
  802. */
  803. find_smp_config();
  804. reserve_ibft_region();
  805. /*
  806. * Need to conclude brk, before memblock_x86_fill()
  807. * it could use memblock_find_in_range, could overlap with
  808. * brk area.
  809. */
  810. reserve_brk();
  811. cleanup_highmap();
  812. memblock.current_limit = get_max_mapped();
  813. memblock_x86_fill();
  814. /*
  815. * The EFI specification says that boot service code won't be called
  816. * after ExitBootServices(). This is, in fact, a lie.
  817. */
  818. if (efi_enabled)
  819. efi_reserve_boot_services();
  820. /* preallocate 4k for mptable mpc */
  821. early_reserve_e820_mpc_new();
  822. #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
  823. setup_bios_corruption_check();
  824. #endif
  825. printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
  826. (max_pfn_mapped<<PAGE_SHIFT) - 1);
  827. setup_real_mode();
  828. trim_platform_memory_ranges();
  829. init_gbpages();
  830. /* max_pfn_mapped is updated here */
  831. max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
  832. max_pfn_mapped = max_low_pfn_mapped;
  833. #ifdef CONFIG_X86_64
  834. if (max_pfn > max_low_pfn) {
  835. int i;
  836. unsigned long start, end;
  837. unsigned long start_pfn, end_pfn;
  838. for_each_mem_pfn_range(i, MAX_NUMNODES, &start_pfn, &end_pfn,
  839. NULL) {
  840. end = PFN_PHYS(end_pfn);
  841. if (end <= (1UL<<32))
  842. continue;
  843. start = PFN_PHYS(start_pfn);
  844. max_pfn_mapped = init_memory_mapping(
  845. max((1UL<<32), start), end);
  846. }
  847. /* can we preseve max_low_pfn ?*/
  848. max_low_pfn = max_pfn;
  849. }
  850. #endif
  851. memblock.current_limit = get_max_mapped();
  852. dma_contiguous_reserve(0);
  853. /*
  854. * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
  855. */
  856. #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
  857. if (init_ohci1394_dma_early)
  858. init_ohci1394_dma_on_all_controllers();
  859. #endif
  860. /* Allocate bigger log buffer */
  861. setup_log_buf(1);
  862. reserve_initrd();
  863. #if defined(CONFIG_ACPI) && defined(CONFIG_BLK_DEV_INITRD)
  864. acpi_initrd_override((void *)initrd_start, initrd_end - initrd_start);
  865. #endif
  866. reserve_crashkernel();
  867. vsmp_init();
  868. io_delay_init();
  869. /*
  870. * Parse the ACPI tables for possible boot-time SMP configuration.
  871. */
  872. acpi_boot_table_init();
  873. early_acpi_boot_init();
  874. initmem_init();
  875. memblock_find_dma_reserve();
  876. #ifdef CONFIG_KVM_GUEST
  877. kvmclock_init();
  878. #endif
  879. x86_init.paging.pagetable_init();
  880. if (boot_cpu_data.cpuid_level >= 0) {
  881. /* A CPU has %cr4 if and only if it has CPUID */
  882. mmu_cr4_features = read_cr4();
  883. if (trampoline_cr4_features)
  884. *trampoline_cr4_features = mmu_cr4_features;
  885. }
  886. #ifdef CONFIG_X86_32
  887. /* sync back kernel address range */
  888. clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
  889. swapper_pg_dir + KERNEL_PGD_BOUNDARY,
  890. KERNEL_PGD_PTRS);
  891. #endif
  892. tboot_probe();
  893. #ifdef CONFIG_X86_64
  894. map_vsyscall();
  895. #endif
  896. generic_apic_probe();
  897. early_quirks();
  898. /*
  899. * Read APIC and some other early information from ACPI tables.
  900. */
  901. acpi_boot_init();
  902. sfi_init();
  903. x86_dtb_init();
  904. /*
  905. * get boot-time SMP configuration:
  906. */
  907. if (smp_found_config)
  908. get_smp_config();
  909. prefill_possible_map();
  910. init_cpu_to_node();
  911. init_apic_mappings();
  912. if (x86_io_apic_ops.init)
  913. x86_io_apic_ops.init();
  914. kvm_guest_init();
  915. e820_reserve_resources();
  916. e820_mark_nosave_regions(max_low_pfn);
  917. x86_init.resources.reserve_resources();
  918. e820_setup_gap();
  919. #ifdef CONFIG_VT
  920. #if defined(CONFIG_VGA_CONSOLE)
  921. if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
  922. conswitchp = &vga_con;
  923. #elif defined(CONFIG_DUMMY_CONSOLE)
  924. conswitchp = &dummy_con;
  925. #endif
  926. #endif
  927. x86_init.oem.banner();
  928. x86_init.timers.wallclock_init();
  929. mcheck_init();
  930. arch_init_ideal_nops();
  931. register_refined_jiffies(CLOCK_TICK_RATE);
  932. #ifdef CONFIG_EFI
  933. /* Once setup is done above, disable efi_enabled on mismatched
  934. * firmware/kernel archtectures since there is no support for
  935. * runtime services.
  936. */
  937. if (efi_enabled && IS_ENABLED(CONFIG_X86_64) != efi_64bit) {
  938. pr_info("efi: Setup done, disabling due to 32/64-bit mismatch\n");
  939. efi_unmap_memmap();
  940. efi_enabled = 0;
  941. }
  942. #endif
  943. }
  944. #ifdef CONFIG_X86_32
  945. static struct resource video_ram_resource = {
  946. .name = "Video RAM area",
  947. .start = 0xa0000,
  948. .end = 0xbffff,
  949. .flags = IORESOURCE_BUSY | IORESOURCE_MEM
  950. };
  951. void __init i386_reserve_resources(void)
  952. {
  953. request_resource(&iomem_resource, &video_ram_resource);
  954. reserve_standard_io_resources();
  955. }
  956. #endif /* CONFIG_X86_32 */