setup.c 28 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161
  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. #ifdef CONFIG_PCI
  524. int i;
  525. u16 vendor, devid;
  526. static const __initconst u16 snb_ids[] = {
  527. 0x0102,
  528. 0x0112,
  529. 0x0122,
  530. 0x0106,
  531. 0x0116,
  532. 0x0126,
  533. 0x010a,
  534. };
  535. /* Assume no if something weird is going on with PCI */
  536. if (!early_pci_allowed())
  537. return false;
  538. vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
  539. if (vendor != 0x8086)
  540. return false;
  541. devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
  542. for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
  543. if (devid == snb_ids[i])
  544. return true;
  545. #endif
  546. return false;
  547. }
  548. /*
  549. * Sandy Bridge graphics has trouble with certain ranges, exclude
  550. * them from allocation.
  551. */
  552. static void __init trim_snb_memory(void)
  553. {
  554. static const __initconst unsigned long bad_pages[] = {
  555. 0x20050000,
  556. 0x20110000,
  557. 0x20130000,
  558. 0x20138000,
  559. 0x40004000,
  560. };
  561. int i;
  562. if (!snb_gfx_workaround_needed())
  563. return;
  564. printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
  565. /*
  566. * Reserve all memory below the 1 MB mark that has not
  567. * already been reserved.
  568. */
  569. memblock_reserve(0, 1<<20);
  570. for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
  571. if (memblock_reserve(bad_pages[i], PAGE_SIZE))
  572. printk(KERN_WARNING "failed to reserve 0x%08lx\n",
  573. bad_pages[i]);
  574. }
  575. }
  576. /*
  577. * Here we put platform-specific memory range workarounds, i.e.
  578. * memory known to be corrupt or otherwise in need to be reserved on
  579. * specific platforms.
  580. *
  581. * If this gets used more widely it could use a real dispatch mechanism.
  582. */
  583. static void __init trim_platform_memory_ranges(void)
  584. {
  585. trim_snb_memory();
  586. }
  587. static void __init trim_bios_range(void)
  588. {
  589. /*
  590. * A special case is the first 4Kb of memory;
  591. * This is a BIOS owned area, not kernel ram, but generally
  592. * not listed as such in the E820 table.
  593. *
  594. * This typically reserves additional memory (64KiB by default)
  595. * since some BIOSes are known to corrupt low memory. See the
  596. * Kconfig help text for X86_RESERVE_LOW.
  597. */
  598. e820_update_range(0, ALIGN(reserve_low, PAGE_SIZE),
  599. E820_RAM, E820_RESERVED);
  600. /*
  601. * special case: Some BIOSen report the PC BIOS
  602. * area (640->1Mb) as ram even though it is not.
  603. * take them out.
  604. */
  605. e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
  606. sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
  607. }
  608. static int __init parse_reservelow(char *p)
  609. {
  610. unsigned long long size;
  611. if (!p)
  612. return -EINVAL;
  613. size = memparse(p, &p);
  614. if (size < 4096)
  615. size = 4096;
  616. if (size > 640*1024)
  617. size = 640*1024;
  618. reserve_low = size;
  619. return 0;
  620. }
  621. early_param("reservelow", parse_reservelow);
  622. /*
  623. * Determine if we were loaded by an EFI loader. If so, then we have also been
  624. * passed the efi memmap, systab, etc., so we should use these data structures
  625. * for initialization. Note, the efi init code path is determined by the
  626. * global efi_enabled. This allows the same kernel image to be used on existing
  627. * systems (with a traditional BIOS) as well as on EFI systems.
  628. */
  629. /*
  630. * setup_arch - architecture-specific boot-time initializations
  631. *
  632. * Note: On x86_64, fixmaps are ready for use even before this is called.
  633. */
  634. void __init setup_arch(char **cmdline_p)
  635. {
  636. #ifdef CONFIG_X86_32
  637. memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
  638. visws_early_detect();
  639. /*
  640. * copy kernel address range established so far and switch
  641. * to the proper swapper page table
  642. */
  643. clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY,
  644. initial_page_table + KERNEL_PGD_BOUNDARY,
  645. KERNEL_PGD_PTRS);
  646. load_cr3(swapper_pg_dir);
  647. __flush_tlb_all();
  648. #else
  649. printk(KERN_INFO "Command line: %s\n", boot_command_line);
  650. #endif
  651. /*
  652. * If we have OLPC OFW, we might end up relocating the fixmap due to
  653. * reserve_top(), so do this before touching the ioremap area.
  654. */
  655. olpc_ofw_detect();
  656. early_trap_init();
  657. early_cpu_init();
  658. early_ioremap_init();
  659. setup_olpc_ofw_pgd();
  660. ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
  661. screen_info = boot_params.screen_info;
  662. edid_info = boot_params.edid_info;
  663. #ifdef CONFIG_X86_32
  664. apm_info.bios = boot_params.apm_bios_info;
  665. ist_info = boot_params.ist_info;
  666. if (boot_params.sys_desc_table.length != 0) {
  667. machine_id = boot_params.sys_desc_table.table[0];
  668. machine_submodel_id = boot_params.sys_desc_table.table[1];
  669. BIOS_revision = boot_params.sys_desc_table.table[2];
  670. }
  671. #endif
  672. saved_video_mode = boot_params.hdr.vid_mode;
  673. bootloader_type = boot_params.hdr.type_of_loader;
  674. if ((bootloader_type >> 4) == 0xe) {
  675. bootloader_type &= 0xf;
  676. bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
  677. }
  678. bootloader_version = bootloader_type & 0xf;
  679. bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
  680. #ifdef CONFIG_BLK_DEV_RAM
  681. rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
  682. rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
  683. rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
  684. #endif
  685. #ifdef CONFIG_EFI
  686. if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
  687. "EL32", 4)) {
  688. efi_enabled = 1;
  689. efi_64bit = false;
  690. } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
  691. "EL64", 4)) {
  692. efi_enabled = 1;
  693. efi_64bit = true;
  694. }
  695. if (efi_enabled && efi_memblock_x86_reserve_range())
  696. efi_enabled = 0;
  697. #endif
  698. x86_init.oem.arch_setup();
  699. iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
  700. setup_memory_map();
  701. parse_setup_data();
  702. /* update the e820_saved too */
  703. e820_reserve_setup_data();
  704. copy_edd();
  705. if (!boot_params.hdr.root_flags)
  706. root_mountflags &= ~MS_RDONLY;
  707. init_mm.start_code = (unsigned long) _text;
  708. init_mm.end_code = (unsigned long) _etext;
  709. init_mm.end_data = (unsigned long) _edata;
  710. init_mm.brk = _brk_end;
  711. code_resource.start = virt_to_phys(_text);
  712. code_resource.end = virt_to_phys(_etext)-1;
  713. data_resource.start = virt_to_phys(_etext);
  714. data_resource.end = virt_to_phys(_edata)-1;
  715. bss_resource.start = virt_to_phys(&__bss_start);
  716. bss_resource.end = virt_to_phys(&__bss_stop)-1;
  717. #ifdef CONFIG_CMDLINE_BOOL
  718. #ifdef CONFIG_CMDLINE_OVERRIDE
  719. strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
  720. #else
  721. if (builtin_cmdline[0]) {
  722. /* append boot loader cmdline to builtin */
  723. strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
  724. strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
  725. strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
  726. }
  727. #endif
  728. #endif
  729. strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
  730. *cmdline_p = command_line;
  731. /*
  732. * x86_configure_nx() is called before parse_early_param() to detect
  733. * whether hardware doesn't support NX (so that the early EHCI debug
  734. * console setup can safely call set_fixmap()). It may then be called
  735. * again from within noexec_setup() during parsing early parameters
  736. * to honor the respective command line option.
  737. */
  738. x86_configure_nx();
  739. parse_early_param();
  740. x86_report_nx();
  741. /* after early param, so could get panic from serial */
  742. memblock_x86_reserve_range_setup_data();
  743. if (acpi_mps_check()) {
  744. #ifdef CONFIG_X86_LOCAL_APIC
  745. disable_apic = 1;
  746. #endif
  747. setup_clear_cpu_cap(X86_FEATURE_APIC);
  748. }
  749. #ifdef CONFIG_PCI
  750. if (pci_early_dump_regs)
  751. early_dump_pci_devices();
  752. #endif
  753. finish_e820_parsing();
  754. if (efi_enabled)
  755. efi_init();
  756. dmi_scan_machine();
  757. /*
  758. * VMware detection requires dmi to be available, so this
  759. * needs to be done after dmi_scan_machine, for the BP.
  760. */
  761. init_hypervisor_platform();
  762. x86_init.resources.probe_roms();
  763. /* after parse_early_param, so could debug it */
  764. insert_resource(&iomem_resource, &code_resource);
  765. insert_resource(&iomem_resource, &data_resource);
  766. insert_resource(&iomem_resource, &bss_resource);
  767. trim_bios_range();
  768. #ifdef CONFIG_X86_32
  769. if (ppro_with_ram_bug()) {
  770. e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
  771. E820_RESERVED);
  772. sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
  773. printk(KERN_INFO "fixed physical RAM map:\n");
  774. e820_print_map("bad_ppro");
  775. }
  776. #else
  777. early_gart_iommu_check();
  778. #endif
  779. /*
  780. * partially used pages are not usable - thus
  781. * we are rounding upwards:
  782. */
  783. max_pfn = e820_end_of_ram_pfn();
  784. /* update e820 for memory not covered by WB MTRRs */
  785. mtrr_bp_init();
  786. if (mtrr_trim_uncached_memory(max_pfn))
  787. max_pfn = e820_end_of_ram_pfn();
  788. #ifdef CONFIG_X86_32
  789. /* max_low_pfn get updated here */
  790. find_low_pfn_range();
  791. #else
  792. num_physpages = max_pfn;
  793. check_x2apic();
  794. /* How many end-of-memory variables you have, grandma! */
  795. /* need this before calling reserve_initrd */
  796. if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
  797. max_low_pfn = e820_end_of_low_ram_pfn();
  798. else
  799. max_low_pfn = max_pfn;
  800. high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
  801. #endif
  802. /*
  803. * Find and reserve possible boot-time SMP configuration:
  804. */
  805. find_smp_config();
  806. reserve_ibft_region();
  807. /*
  808. * Need to conclude brk, before memblock_x86_fill()
  809. * it could use memblock_find_in_range, could overlap with
  810. * brk area.
  811. */
  812. reserve_brk();
  813. cleanup_highmap();
  814. memblock.current_limit = get_max_mapped();
  815. memblock_x86_fill();
  816. /*
  817. * The EFI specification says that boot service code won't be called
  818. * after ExitBootServices(). This is, in fact, a lie.
  819. */
  820. if (efi_enabled)
  821. efi_reserve_boot_services();
  822. /* preallocate 4k for mptable mpc */
  823. early_reserve_e820_mpc_new();
  824. #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
  825. setup_bios_corruption_check();
  826. #endif
  827. printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
  828. (max_pfn_mapped<<PAGE_SHIFT) - 1);
  829. setup_real_mode();
  830. trim_platform_memory_ranges();
  831. init_gbpages();
  832. /* max_pfn_mapped is updated here */
  833. max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
  834. max_pfn_mapped = max_low_pfn_mapped;
  835. #ifdef CONFIG_X86_64
  836. if (max_pfn > max_low_pfn) {
  837. int i;
  838. unsigned long start, end;
  839. unsigned long start_pfn, end_pfn;
  840. for_each_mem_pfn_range(i, MAX_NUMNODES, &start_pfn, &end_pfn,
  841. NULL) {
  842. end = PFN_PHYS(end_pfn);
  843. if (end <= (1UL<<32))
  844. continue;
  845. start = PFN_PHYS(start_pfn);
  846. max_pfn_mapped = init_memory_mapping(
  847. max((1UL<<32), start), end);
  848. }
  849. /* can we preseve max_low_pfn ?*/
  850. max_low_pfn = max_pfn;
  851. }
  852. #endif
  853. memblock.current_limit = get_max_mapped();
  854. dma_contiguous_reserve(0);
  855. /*
  856. * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
  857. */
  858. #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
  859. if (init_ohci1394_dma_early)
  860. init_ohci1394_dma_on_all_controllers();
  861. #endif
  862. /* Allocate bigger log buffer */
  863. setup_log_buf(1);
  864. reserve_initrd();
  865. #if defined(CONFIG_ACPI) && defined(CONFIG_BLK_DEV_INITRD)
  866. acpi_initrd_override((void *)initrd_start, initrd_end - initrd_start);
  867. #endif
  868. reserve_crashkernel();
  869. vsmp_init();
  870. io_delay_init();
  871. /*
  872. * Parse the ACPI tables for possible boot-time SMP configuration.
  873. */
  874. acpi_boot_table_init();
  875. early_acpi_boot_init();
  876. initmem_init();
  877. memblock_find_dma_reserve();
  878. #ifdef CONFIG_KVM_GUEST
  879. kvmclock_init();
  880. #endif
  881. x86_init.paging.pagetable_init();
  882. if (boot_cpu_data.cpuid_level >= 0) {
  883. /* A CPU has %cr4 if and only if it has CPUID */
  884. mmu_cr4_features = read_cr4();
  885. if (trampoline_cr4_features)
  886. *trampoline_cr4_features = mmu_cr4_features;
  887. }
  888. #ifdef CONFIG_X86_32
  889. /* sync back kernel address range */
  890. clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
  891. swapper_pg_dir + KERNEL_PGD_BOUNDARY,
  892. KERNEL_PGD_PTRS);
  893. #endif
  894. tboot_probe();
  895. #ifdef CONFIG_X86_64
  896. map_vsyscall();
  897. #endif
  898. generic_apic_probe();
  899. early_quirks();
  900. /*
  901. * Read APIC and some other early information from ACPI tables.
  902. */
  903. acpi_boot_init();
  904. sfi_init();
  905. x86_dtb_init();
  906. /*
  907. * get boot-time SMP configuration:
  908. */
  909. if (smp_found_config)
  910. get_smp_config();
  911. prefill_possible_map();
  912. init_cpu_to_node();
  913. init_apic_mappings();
  914. if (x86_io_apic_ops.init)
  915. x86_io_apic_ops.init();
  916. kvm_guest_init();
  917. e820_reserve_resources();
  918. e820_mark_nosave_regions(max_low_pfn);
  919. x86_init.resources.reserve_resources();
  920. e820_setup_gap();
  921. #ifdef CONFIG_VT
  922. #if defined(CONFIG_VGA_CONSOLE)
  923. if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
  924. conswitchp = &vga_con;
  925. #elif defined(CONFIG_DUMMY_CONSOLE)
  926. conswitchp = &dummy_con;
  927. #endif
  928. #endif
  929. x86_init.oem.banner();
  930. x86_init.timers.wallclock_init();
  931. mcheck_init();
  932. arch_init_ideal_nops();
  933. register_refined_jiffies(CLOCK_TICK_RATE);
  934. #ifdef CONFIG_EFI
  935. /* Once setup is done above, disable efi_enabled on mismatched
  936. * firmware/kernel archtectures since there is no support for
  937. * runtime services.
  938. */
  939. if (efi_enabled && IS_ENABLED(CONFIG_X86_64) != efi_64bit) {
  940. pr_info("efi: Setup done, disabling due to 32/64-bit mismatch\n");
  941. efi_unmap_memmap();
  942. efi_enabled = 0;
  943. }
  944. #endif
  945. }
  946. #ifdef CONFIG_X86_32
  947. static struct resource video_ram_resource = {
  948. .name = "Video RAM area",
  949. .start = 0xa0000,
  950. .end = 0xbffff,
  951. .flags = IORESOURCE_BUSY | IORESOURCE_MEM
  952. };
  953. void __init i386_reserve_resources(void)
  954. {
  955. request_resource(&iomem_resource, &video_ram_resource);
  956. reserve_standard_io_resources();
  957. }
  958. #endif /* CONFIG_X86_32 */