setup.c 26 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955
  1. /*
  2. * S390 version
  3. * Copyright IBM Corp. 1999, 2012
  4. * Author(s): Hartmut Penner (hp@de.ibm.com),
  5. * Martin Schwidefsky (schwidefsky@de.ibm.com)
  6. *
  7. * Derived from "arch/i386/kernel/setup.c"
  8. * Copyright (C) 1995, Linus Torvalds
  9. */
  10. /*
  11. * This file handles the architecture-dependent parts of initialization
  12. */
  13. #define KMSG_COMPONENT "setup"
  14. #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
  15. #include <linux/errno.h>
  16. #include <linux/export.h>
  17. #include <linux/sched.h>
  18. #include <linux/kernel.h>
  19. #include <linux/memblock.h>
  20. #include <linux/mm.h>
  21. #include <linux/stddef.h>
  22. #include <linux/unistd.h>
  23. #include <linux/ptrace.h>
  24. #include <linux/random.h>
  25. #include <linux/user.h>
  26. #include <linux/tty.h>
  27. #include <linux/ioport.h>
  28. #include <linux/delay.h>
  29. #include <linux/init.h>
  30. #include <linux/initrd.h>
  31. #include <linux/bootmem.h>
  32. #include <linux/root_dev.h>
  33. #include <linux/console.h>
  34. #include <linux/kernel_stat.h>
  35. #include <linux/device.h>
  36. #include <linux/notifier.h>
  37. #include <linux/pfn.h>
  38. #include <linux/ctype.h>
  39. #include <linux/reboot.h>
  40. #include <linux/topology.h>
  41. #include <linux/ftrace.h>
  42. #include <linux/kexec.h>
  43. #include <linux/crash_dump.h>
  44. #include <linux/memory.h>
  45. #include <linux/compat.h>
  46. #include <asm/ipl.h>
  47. #include <asm/facility.h>
  48. #include <asm/smp.h>
  49. #include <asm/mmu_context.h>
  50. #include <asm/cpcmd.h>
  51. #include <asm/lowcore.h>
  52. #include <asm/irq.h>
  53. #include <asm/page.h>
  54. #include <asm/ptrace.h>
  55. #include <asm/sections.h>
  56. #include <asm/ebcdic.h>
  57. #include <asm/kvm_virtio.h>
  58. #include <asm/diag.h>
  59. #include <asm/os_info.h>
  60. #include <asm/sclp.h>
  61. #include <asm/sysinfo.h>
  62. #include "entry.h"
  63. /*
  64. * Machine setup..
  65. */
  66. unsigned int console_mode = 0;
  67. EXPORT_SYMBOL(console_mode);
  68. unsigned int console_devno = -1;
  69. EXPORT_SYMBOL(console_devno);
  70. unsigned int console_irq = -1;
  71. EXPORT_SYMBOL(console_irq);
  72. unsigned long elf_hwcap = 0;
  73. char elf_platform[ELF_PLATFORM_SIZE];
  74. int __initdata memory_end_set;
  75. unsigned long __initdata memory_end;
  76. unsigned long __initdata max_physmem_end;
  77. unsigned long VMALLOC_START;
  78. EXPORT_SYMBOL(VMALLOC_START);
  79. unsigned long VMALLOC_END;
  80. EXPORT_SYMBOL(VMALLOC_END);
  81. struct page *vmemmap;
  82. EXPORT_SYMBOL(vmemmap);
  83. #ifdef CONFIG_64BIT
  84. unsigned long MODULES_VADDR;
  85. unsigned long MODULES_END;
  86. #endif
  87. /* An array with a pointer to the lowcore of every CPU. */
  88. struct _lowcore *lowcore_ptr[NR_CPUS];
  89. EXPORT_SYMBOL(lowcore_ptr);
  90. /*
  91. * This is set up by the setup-routine at boot-time
  92. * for S390 need to find out, what we have to setup
  93. * using address 0x10400 ...
  94. */
  95. #include <asm/setup.h>
  96. /*
  97. * condev= and conmode= setup parameter.
  98. */
  99. static int __init condev_setup(char *str)
  100. {
  101. int vdev;
  102. vdev = simple_strtoul(str, &str, 0);
  103. if (vdev >= 0 && vdev < 65536) {
  104. console_devno = vdev;
  105. console_irq = -1;
  106. }
  107. return 1;
  108. }
  109. __setup("condev=", condev_setup);
  110. static void __init set_preferred_console(void)
  111. {
  112. if (MACHINE_IS_KVM) {
  113. if (sclp_has_vt220())
  114. add_preferred_console("ttyS", 1, NULL);
  115. else if (sclp_has_linemode())
  116. add_preferred_console("ttyS", 0, NULL);
  117. else
  118. add_preferred_console("hvc", 0, NULL);
  119. } else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
  120. add_preferred_console("ttyS", 0, NULL);
  121. else if (CONSOLE_IS_3270)
  122. add_preferred_console("tty3270", 0, NULL);
  123. }
  124. static int __init conmode_setup(char *str)
  125. {
  126. #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
  127. if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
  128. SET_CONSOLE_SCLP;
  129. #endif
  130. #if defined(CONFIG_TN3215_CONSOLE)
  131. if (strncmp(str, "3215", 5) == 0)
  132. SET_CONSOLE_3215;
  133. #endif
  134. #if defined(CONFIG_TN3270_CONSOLE)
  135. if (strncmp(str, "3270", 5) == 0)
  136. SET_CONSOLE_3270;
  137. #endif
  138. set_preferred_console();
  139. return 1;
  140. }
  141. __setup("conmode=", conmode_setup);
  142. static void __init conmode_default(void)
  143. {
  144. char query_buffer[1024];
  145. char *ptr;
  146. if (MACHINE_IS_VM) {
  147. cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
  148. console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
  149. ptr = strstr(query_buffer, "SUBCHANNEL =");
  150. console_irq = simple_strtoul(ptr + 13, NULL, 16);
  151. cpcmd("QUERY TERM", query_buffer, 1024, NULL);
  152. ptr = strstr(query_buffer, "CONMODE");
  153. /*
  154. * Set the conmode to 3215 so that the device recognition
  155. * will set the cu_type of the console to 3215. If the
  156. * conmode is 3270 and we don't set it back then both
  157. * 3215 and the 3270 driver will try to access the console
  158. * device (3215 as console and 3270 as normal tty).
  159. */
  160. cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
  161. if (ptr == NULL) {
  162. #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
  163. SET_CONSOLE_SCLP;
  164. #endif
  165. return;
  166. }
  167. if (strncmp(ptr + 8, "3270", 4) == 0) {
  168. #if defined(CONFIG_TN3270_CONSOLE)
  169. SET_CONSOLE_3270;
  170. #elif defined(CONFIG_TN3215_CONSOLE)
  171. SET_CONSOLE_3215;
  172. #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
  173. SET_CONSOLE_SCLP;
  174. #endif
  175. } else if (strncmp(ptr + 8, "3215", 4) == 0) {
  176. #if defined(CONFIG_TN3215_CONSOLE)
  177. SET_CONSOLE_3215;
  178. #elif defined(CONFIG_TN3270_CONSOLE)
  179. SET_CONSOLE_3270;
  180. #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
  181. SET_CONSOLE_SCLP;
  182. #endif
  183. }
  184. } else {
  185. #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
  186. SET_CONSOLE_SCLP;
  187. #endif
  188. }
  189. }
  190. #ifdef CONFIG_CRASH_DUMP
  191. static void __init setup_zfcpdump(void)
  192. {
  193. if (ipl_info.type != IPL_TYPE_FCP_DUMP)
  194. return;
  195. if (OLDMEM_BASE)
  196. return;
  197. strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
  198. console_loglevel = 2;
  199. }
  200. #else
  201. static inline void setup_zfcpdump(void) {}
  202. #endif /* CONFIG_CRASH_DUMP */
  203. /*
  204. * Reboot, halt and power_off stubs. They just call _machine_restart,
  205. * _machine_halt or _machine_power_off.
  206. */
  207. void machine_restart(char *command)
  208. {
  209. if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
  210. /*
  211. * Only unblank the console if we are called in enabled
  212. * context or a bust_spinlocks cleared the way for us.
  213. */
  214. console_unblank();
  215. _machine_restart(command);
  216. }
  217. void machine_halt(void)
  218. {
  219. if (!in_interrupt() || oops_in_progress)
  220. /*
  221. * Only unblank the console if we are called in enabled
  222. * context or a bust_spinlocks cleared the way for us.
  223. */
  224. console_unblank();
  225. _machine_halt();
  226. }
  227. void machine_power_off(void)
  228. {
  229. if (!in_interrupt() || oops_in_progress)
  230. /*
  231. * Only unblank the console if we are called in enabled
  232. * context or a bust_spinlocks cleared the way for us.
  233. */
  234. console_unblank();
  235. _machine_power_off();
  236. }
  237. /*
  238. * Dummy power off function.
  239. */
  240. void (*pm_power_off)(void) = machine_power_off;
  241. EXPORT_SYMBOL_GPL(pm_power_off);
  242. static int __init early_parse_mem(char *p)
  243. {
  244. memory_end = memparse(p, &p);
  245. memory_end &= PAGE_MASK;
  246. memory_end_set = 1;
  247. return 0;
  248. }
  249. early_param("mem", early_parse_mem);
  250. static int __init parse_vmalloc(char *arg)
  251. {
  252. if (!arg)
  253. return -EINVAL;
  254. VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
  255. return 0;
  256. }
  257. early_param("vmalloc", parse_vmalloc);
  258. void *restart_stack __attribute__((__section__(".data")));
  259. static void __init setup_lowcore(void)
  260. {
  261. struct _lowcore *lc;
  262. /*
  263. * Setup lowcore for boot cpu
  264. */
  265. BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
  266. lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
  267. lc->restart_psw.mask = PSW_KERNEL_BITS;
  268. lc->restart_psw.addr =
  269. PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
  270. lc->external_new_psw.mask = PSW_KERNEL_BITS |
  271. PSW_MASK_DAT | PSW_MASK_MCHECK;
  272. lc->external_new_psw.addr =
  273. PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
  274. lc->svc_new_psw.mask = PSW_KERNEL_BITS |
  275. PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
  276. lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
  277. lc->program_new_psw.mask = PSW_KERNEL_BITS |
  278. PSW_MASK_DAT | PSW_MASK_MCHECK;
  279. lc->program_new_psw.addr =
  280. PSW_ADDR_AMODE | (unsigned long) pgm_check_handler;
  281. lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
  282. lc->mcck_new_psw.addr =
  283. PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
  284. lc->io_new_psw.mask = PSW_KERNEL_BITS |
  285. PSW_MASK_DAT | PSW_MASK_MCHECK;
  286. lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
  287. lc->clock_comparator = -1ULL;
  288. lc->kernel_stack = ((unsigned long) &init_thread_union)
  289. + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
  290. lc->async_stack = (unsigned long)
  291. __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0)
  292. + ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
  293. lc->panic_stack = (unsigned long)
  294. __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0)
  295. + PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
  296. lc->current_task = (unsigned long) init_thread_union.thread_info.task;
  297. lc->thread_info = (unsigned long) &init_thread_union;
  298. lc->machine_flags = S390_lowcore.machine_flags;
  299. lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
  300. memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
  301. MAX_FACILITY_BIT/8);
  302. #ifndef CONFIG_64BIT
  303. if (MACHINE_HAS_IEEE) {
  304. lc->extended_save_area_addr = (__u32)
  305. __alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
  306. /* enable extended save area */
  307. __ctl_set_bit(14, 29);
  308. }
  309. #else
  310. lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
  311. #endif
  312. lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
  313. lc->async_enter_timer = S390_lowcore.async_enter_timer;
  314. lc->exit_timer = S390_lowcore.exit_timer;
  315. lc->user_timer = S390_lowcore.user_timer;
  316. lc->system_timer = S390_lowcore.system_timer;
  317. lc->steal_timer = S390_lowcore.steal_timer;
  318. lc->last_update_timer = S390_lowcore.last_update_timer;
  319. lc->last_update_clock = S390_lowcore.last_update_clock;
  320. lc->ftrace_func = S390_lowcore.ftrace_func;
  321. restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
  322. restart_stack += ASYNC_SIZE;
  323. /*
  324. * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
  325. * restart data to the absolute zero lowcore. This is necessary if
  326. * PSW restart is done on an offline CPU that has lowcore zero.
  327. */
  328. lc->restart_stack = (unsigned long) restart_stack;
  329. lc->restart_fn = (unsigned long) do_restart;
  330. lc->restart_data = 0;
  331. lc->restart_source = -1UL;
  332. /* Setup absolute zero lowcore */
  333. mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
  334. mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
  335. mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
  336. mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
  337. mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
  338. #ifdef CONFIG_SMP
  339. lc->spinlock_lockval = arch_spin_lockval(0);
  340. #endif
  341. set_prefix((u32)(unsigned long) lc);
  342. lowcore_ptr[0] = lc;
  343. }
  344. static struct resource code_resource = {
  345. .name = "Kernel code",
  346. .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
  347. };
  348. static struct resource data_resource = {
  349. .name = "Kernel data",
  350. .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
  351. };
  352. static struct resource bss_resource = {
  353. .name = "Kernel bss",
  354. .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
  355. };
  356. static struct resource __initdata *standard_resources[] = {
  357. &code_resource,
  358. &data_resource,
  359. &bss_resource,
  360. };
  361. static void __init setup_resources(void)
  362. {
  363. struct resource *res, *std_res, *sub_res;
  364. struct memblock_region *reg;
  365. int j;
  366. code_resource.start = (unsigned long) &_text;
  367. code_resource.end = (unsigned long) &_etext - 1;
  368. data_resource.start = (unsigned long) &_etext;
  369. data_resource.end = (unsigned long) &_edata - 1;
  370. bss_resource.start = (unsigned long) &__bss_start;
  371. bss_resource.end = (unsigned long) &__bss_stop - 1;
  372. for_each_memblock(memory, reg) {
  373. res = alloc_bootmem_low(sizeof(*res));
  374. res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
  375. res->name = "System RAM";
  376. res->start = reg->base;
  377. res->end = reg->base + reg->size - 1;
  378. request_resource(&iomem_resource, res);
  379. for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
  380. std_res = standard_resources[j];
  381. if (std_res->start < res->start ||
  382. std_res->start > res->end)
  383. continue;
  384. if (std_res->end > res->end) {
  385. sub_res = alloc_bootmem_low(sizeof(*sub_res));
  386. *sub_res = *std_res;
  387. sub_res->end = res->end;
  388. std_res->start = res->end + 1;
  389. request_resource(res, sub_res);
  390. } else {
  391. request_resource(res, std_res);
  392. }
  393. }
  394. }
  395. }
  396. static void __init setup_memory_end(void)
  397. {
  398. unsigned long vmax, vmalloc_size, tmp;
  399. /* Choose kernel address space layout: 2, 3, or 4 levels. */
  400. #ifdef CONFIG_64BIT
  401. vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
  402. tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
  403. tmp = tmp * (sizeof(struct page) + PAGE_SIZE) + vmalloc_size;
  404. if (tmp <= (1UL << 42))
  405. vmax = 1UL << 42; /* 3-level kernel page table */
  406. else
  407. vmax = 1UL << 53; /* 4-level kernel page table */
  408. /* module area is at the end of the kernel address space. */
  409. MODULES_END = vmax;
  410. MODULES_VADDR = MODULES_END - MODULES_LEN;
  411. VMALLOC_END = MODULES_VADDR;
  412. #else
  413. vmalloc_size = VMALLOC_END ?: 96UL << 20;
  414. vmax = 1UL << 31; /* 2-level kernel page table */
  415. /* vmalloc area is at the end of the kernel address space. */
  416. VMALLOC_END = vmax;
  417. #endif
  418. VMALLOC_START = vmax - vmalloc_size;
  419. /* Split remaining virtual space between 1:1 mapping & vmemmap array */
  420. tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
  421. /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
  422. tmp = SECTION_ALIGN_UP(tmp);
  423. tmp = VMALLOC_START - tmp * sizeof(struct page);
  424. tmp &= ~((vmax >> 11) - 1); /* align to page table level */
  425. tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
  426. vmemmap = (struct page *) tmp;
  427. /* Take care that memory_end is set and <= vmemmap */
  428. memory_end = min(memory_end ?: max_physmem_end, tmp);
  429. max_pfn = max_low_pfn = PFN_DOWN(memory_end);
  430. memblock_remove(memory_end, ULONG_MAX);
  431. pr_notice("Max memory size: %luMB\n", memory_end >> 20);
  432. }
  433. static void __init setup_vmcoreinfo(void)
  434. {
  435. mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note());
  436. }
  437. #ifdef CONFIG_CRASH_DUMP
  438. /*
  439. * When kdump is enabled, we have to ensure that no memory from
  440. * the area [0 - crashkernel memory size] and
  441. * [crashk_res.start - crashk_res.end] is set offline.
  442. */
  443. static int kdump_mem_notifier(struct notifier_block *nb,
  444. unsigned long action, void *data)
  445. {
  446. struct memory_notify *arg = data;
  447. if (action != MEM_GOING_OFFLINE)
  448. return NOTIFY_OK;
  449. if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
  450. return NOTIFY_BAD;
  451. if (arg->start_pfn > PFN_DOWN(crashk_res.end))
  452. return NOTIFY_OK;
  453. if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
  454. return NOTIFY_OK;
  455. return NOTIFY_BAD;
  456. }
  457. static struct notifier_block kdump_mem_nb = {
  458. .notifier_call = kdump_mem_notifier,
  459. };
  460. #endif
  461. /*
  462. * Make sure that the area behind memory_end is protected
  463. */
  464. static void reserve_memory_end(void)
  465. {
  466. #ifdef CONFIG_CRASH_DUMP
  467. if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
  468. !OLDMEM_BASE && sclp_get_hsa_size()) {
  469. memory_end = sclp_get_hsa_size();
  470. memory_end &= PAGE_MASK;
  471. memory_end_set = 1;
  472. }
  473. #endif
  474. if (!memory_end_set)
  475. return;
  476. memblock_reserve(memory_end, ULONG_MAX);
  477. }
  478. /*
  479. * Make sure that oldmem, where the dump is stored, is protected
  480. */
  481. static void reserve_oldmem(void)
  482. {
  483. #ifdef CONFIG_CRASH_DUMP
  484. if (OLDMEM_BASE)
  485. /* Forget all memory above the running kdump system */
  486. memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
  487. #endif
  488. }
  489. /*
  490. * Make sure that oldmem, where the dump is stored, is protected
  491. */
  492. static void remove_oldmem(void)
  493. {
  494. #ifdef CONFIG_CRASH_DUMP
  495. if (OLDMEM_BASE)
  496. /* Forget all memory above the running kdump system */
  497. memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
  498. #endif
  499. }
  500. /*
  501. * Reserve memory for kdump kernel to be loaded with kexec
  502. */
  503. static void __init reserve_crashkernel(void)
  504. {
  505. #ifdef CONFIG_CRASH_DUMP
  506. unsigned long long crash_base, crash_size;
  507. phys_addr_t low, high;
  508. int rc;
  509. rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
  510. &crash_base);
  511. crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
  512. crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
  513. if (rc || crash_size == 0)
  514. return;
  515. if (memblock.memory.regions[0].size < crash_size) {
  516. pr_info("crashkernel reservation failed: %s\n",
  517. "first memory chunk must be at least crashkernel size");
  518. return;
  519. }
  520. low = crash_base ?: OLDMEM_BASE;
  521. high = low + crash_size;
  522. if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
  523. /* The crashkernel fits into OLDMEM, reuse OLDMEM */
  524. crash_base = low;
  525. } else {
  526. /* Find suitable area in free memory */
  527. low = max_t(unsigned long, crash_size, sclp_get_hsa_size());
  528. high = crash_base ? crash_base + crash_size : ULONG_MAX;
  529. if (crash_base && crash_base < low) {
  530. pr_info("crashkernel reservation failed: %s\n",
  531. "crash_base too low");
  532. return;
  533. }
  534. low = crash_base ?: low;
  535. crash_base = memblock_find_in_range(low, high, crash_size,
  536. KEXEC_CRASH_MEM_ALIGN);
  537. }
  538. if (!crash_base) {
  539. pr_info("crashkernel reservation failed: %s\n",
  540. "no suitable area found");
  541. return;
  542. }
  543. if (register_memory_notifier(&kdump_mem_nb))
  544. return;
  545. if (!OLDMEM_BASE && MACHINE_IS_VM)
  546. diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
  547. crashk_res.start = crash_base;
  548. crashk_res.end = crash_base + crash_size - 1;
  549. insert_resource(&iomem_resource, &crashk_res);
  550. memblock_remove(crash_base, crash_size);
  551. pr_info("Reserving %lluMB of memory at %lluMB "
  552. "for crashkernel (System RAM: %luMB)\n",
  553. crash_size >> 20, crash_base >> 20,
  554. (unsigned long)memblock.memory.total_size >> 20);
  555. os_info_crashkernel_add(crash_base, crash_size);
  556. #endif
  557. }
  558. /*
  559. * Reserve the initrd from being used by memblock
  560. */
  561. static void __init reserve_initrd(void)
  562. {
  563. #ifdef CONFIG_BLK_DEV_INITRD
  564. initrd_start = INITRD_START;
  565. initrd_end = initrd_start + INITRD_SIZE;
  566. memblock_reserve(INITRD_START, INITRD_SIZE);
  567. #endif
  568. }
  569. /*
  570. * Check for initrd being in usable memory
  571. */
  572. static void __init check_initrd(void)
  573. {
  574. #ifdef CONFIG_BLK_DEV_INITRD
  575. if (INITRD_START && INITRD_SIZE &&
  576. !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
  577. pr_err("initrd does not fit memory.\n");
  578. memblock_free(INITRD_START, INITRD_SIZE);
  579. initrd_start = initrd_end = 0;
  580. }
  581. #endif
  582. }
  583. /*
  584. * Reserve all kernel text
  585. */
  586. static void __init reserve_kernel(void)
  587. {
  588. unsigned long start_pfn;
  589. start_pfn = PFN_UP(__pa(&_end));
  590. /*
  591. * Reserve memory used for lowcore/command line/kernel image.
  592. */
  593. memblock_reserve(0, (unsigned long)_ehead);
  594. memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
  595. - (unsigned long)_stext);
  596. }
  597. static void __init reserve_elfcorehdr(void)
  598. {
  599. #ifdef CONFIG_CRASH_DUMP
  600. if (is_kdump_kernel())
  601. memblock_reserve(elfcorehdr_addr - OLDMEM_BASE,
  602. PAGE_ALIGN(elfcorehdr_size));
  603. #endif
  604. }
  605. static void __init setup_memory(void)
  606. {
  607. struct memblock_region *reg;
  608. /*
  609. * Init storage key for present memory
  610. */
  611. for_each_memblock(memory, reg) {
  612. storage_key_init_range(reg->base, reg->base + reg->size);
  613. }
  614. psw_set_key(PAGE_DEFAULT_KEY);
  615. /* Only cosmetics */
  616. memblock_enforce_memory_limit(memblock_end_of_DRAM());
  617. }
  618. /*
  619. * Setup hardware capabilities.
  620. */
  621. static void __init setup_hwcaps(void)
  622. {
  623. static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
  624. struct cpuid cpu_id;
  625. int i;
  626. /*
  627. * The store facility list bits numbers as found in the principles
  628. * of operation are numbered with bit 1UL<<31 as number 0 to
  629. * bit 1UL<<0 as number 31.
  630. * Bit 0: instructions named N3, "backported" to esa-mode
  631. * Bit 2: z/Architecture mode is active
  632. * Bit 7: the store-facility-list-extended facility is installed
  633. * Bit 17: the message-security assist is installed
  634. * Bit 19: the long-displacement facility is installed
  635. * Bit 21: the extended-immediate facility is installed
  636. * Bit 22: extended-translation facility 3 is installed
  637. * Bit 30: extended-translation facility 3 enhancement facility
  638. * These get translated to:
  639. * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
  640. * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
  641. * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
  642. * HWCAP_S390_ETF3EH bit 8 (22 && 30).
  643. */
  644. for (i = 0; i < 6; i++)
  645. if (test_facility(stfl_bits[i]))
  646. elf_hwcap |= 1UL << i;
  647. if (test_facility(22) && test_facility(30))
  648. elf_hwcap |= HWCAP_S390_ETF3EH;
  649. /*
  650. * Check for additional facilities with store-facility-list-extended.
  651. * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
  652. * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
  653. * as stored by stfl, bits 32-xxx contain additional facilities.
  654. * How many facility words are stored depends on the number of
  655. * doublewords passed to the instruction. The additional facilities
  656. * are:
  657. * Bit 42: decimal floating point facility is installed
  658. * Bit 44: perform floating point operation facility is installed
  659. * translated to:
  660. * HWCAP_S390_DFP bit 6 (42 && 44).
  661. */
  662. if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
  663. elf_hwcap |= HWCAP_S390_DFP;
  664. /*
  665. * Huge page support HWCAP_S390_HPAGE is bit 7.
  666. */
  667. if (MACHINE_HAS_HPAGE)
  668. elf_hwcap |= HWCAP_S390_HPAGE;
  669. #if defined(CONFIG_64BIT)
  670. /*
  671. * 64-bit register support for 31-bit processes
  672. * HWCAP_S390_HIGH_GPRS is bit 9.
  673. */
  674. elf_hwcap |= HWCAP_S390_HIGH_GPRS;
  675. /*
  676. * Transactional execution support HWCAP_S390_TE is bit 10.
  677. */
  678. if (test_facility(50) && test_facility(73))
  679. elf_hwcap |= HWCAP_S390_TE;
  680. #endif
  681. get_cpu_id(&cpu_id);
  682. add_device_randomness(&cpu_id, sizeof(cpu_id));
  683. switch (cpu_id.machine) {
  684. case 0x9672:
  685. #if !defined(CONFIG_64BIT)
  686. default: /* Use "g5" as default for 31 bit kernels. */
  687. #endif
  688. strcpy(elf_platform, "g5");
  689. break;
  690. case 0x2064:
  691. case 0x2066:
  692. #if defined(CONFIG_64BIT)
  693. default: /* Use "z900" as default for 64 bit kernels. */
  694. #endif
  695. strcpy(elf_platform, "z900");
  696. break;
  697. case 0x2084:
  698. case 0x2086:
  699. strcpy(elf_platform, "z990");
  700. break;
  701. case 0x2094:
  702. case 0x2096:
  703. strcpy(elf_platform, "z9-109");
  704. break;
  705. case 0x2097:
  706. case 0x2098:
  707. strcpy(elf_platform, "z10");
  708. break;
  709. case 0x2817:
  710. case 0x2818:
  711. strcpy(elf_platform, "z196");
  712. break;
  713. case 0x2827:
  714. case 0x2828:
  715. strcpy(elf_platform, "zEC12");
  716. break;
  717. }
  718. }
  719. /*
  720. * Add system information as device randomness
  721. */
  722. static void __init setup_randomness(void)
  723. {
  724. struct sysinfo_3_2_2 *vmms;
  725. vmms = (struct sysinfo_3_2_2 *) alloc_page(GFP_KERNEL);
  726. if (vmms && stsi(vmms, 3, 2, 2) == 0 && vmms->count)
  727. add_device_randomness(&vmms, vmms->count);
  728. free_page((unsigned long) vmms);
  729. }
  730. /*
  731. * Setup function called from init/main.c just after the banner
  732. * was printed.
  733. */
  734. void __init setup_arch(char **cmdline_p)
  735. {
  736. /*
  737. * print what head.S has found out about the machine
  738. */
  739. #ifndef CONFIG_64BIT
  740. if (MACHINE_IS_VM)
  741. pr_info("Linux is running as a z/VM "
  742. "guest operating system in 31-bit mode\n");
  743. else if (MACHINE_IS_LPAR)
  744. pr_info("Linux is running natively in 31-bit mode\n");
  745. if (MACHINE_HAS_IEEE)
  746. pr_info("The hardware system has IEEE compatible "
  747. "floating point units\n");
  748. else
  749. pr_info("The hardware system has no IEEE compatible "
  750. "floating point units\n");
  751. #else /* CONFIG_64BIT */
  752. if (MACHINE_IS_VM)
  753. pr_info("Linux is running as a z/VM "
  754. "guest operating system in 64-bit mode\n");
  755. else if (MACHINE_IS_KVM)
  756. pr_info("Linux is running under KVM in 64-bit mode\n");
  757. else if (MACHINE_IS_LPAR)
  758. pr_info("Linux is running natively in 64-bit mode\n");
  759. #endif /* CONFIG_64BIT */
  760. /* Have one command line that is parsed and saved in /proc/cmdline */
  761. /* boot_command_line has been already set up in early.c */
  762. *cmdline_p = boot_command_line;
  763. ROOT_DEV = Root_RAM0;
  764. /* Is init_mm really needed? */
  765. init_mm.start_code = PAGE_OFFSET;
  766. init_mm.end_code = (unsigned long) &_etext;
  767. init_mm.end_data = (unsigned long) &_edata;
  768. init_mm.brk = (unsigned long) &_end;
  769. parse_early_param();
  770. os_info_init();
  771. setup_ipl();
  772. /* Do some memory reservations *before* memory is added to memblock */
  773. reserve_memory_end();
  774. reserve_oldmem();
  775. reserve_kernel();
  776. reserve_initrd();
  777. reserve_elfcorehdr();
  778. memblock_allow_resize();
  779. /* Get information about *all* installed memory */
  780. detect_memory_memblock();
  781. remove_oldmem();
  782. /*
  783. * Make sure all chunks are MAX_ORDER aligned so we don't need the
  784. * extra checks that HOLES_IN_ZONE would require.
  785. *
  786. * Is this still required?
  787. */
  788. memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
  789. setup_memory_end();
  790. setup_memory();
  791. check_initrd();
  792. reserve_crashkernel();
  793. setup_resources();
  794. setup_vmcoreinfo();
  795. setup_lowcore();
  796. smp_fill_possible_mask();
  797. cpu_init();
  798. s390_init_cpu_topology();
  799. /*
  800. * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
  801. */
  802. setup_hwcaps();
  803. /*
  804. * Create kernel page tables and switch to virtual addressing.
  805. */
  806. paging_init();
  807. /* Setup default console */
  808. conmode_default();
  809. set_preferred_console();
  810. /* Setup zfcpdump support */
  811. setup_zfcpdump();
  812. /* Add system specific data to the random pool */
  813. setup_randomness();
  814. }
  815. #ifdef CONFIG_32BIT
  816. static int no_removal_warning __initdata;
  817. static int __init parse_no_removal_warning(char *str)
  818. {
  819. no_removal_warning = 1;
  820. return 0;
  821. }
  822. __setup("no_removal_warning", parse_no_removal_warning);
  823. static int __init removal_warning(void)
  824. {
  825. if (no_removal_warning)
  826. return 0;
  827. printk(KERN_ALERT "\n\n");
  828. printk(KERN_CONT "Warning - you are using a 31 bit kernel!\n\n");
  829. printk(KERN_CONT "We plan to remove 31 bit kernel support from the kernel sources in March 2015.\n");
  830. printk(KERN_CONT "Currently we assume that nobody is using the 31 bit kernel on old 31 bit\n");
  831. printk(KERN_CONT "hardware anymore. If you think that the code should not be removed and also\n");
  832. printk(KERN_CONT "future versions of the Linux kernel should be able to run in 31 bit mode\n");
  833. printk(KERN_CONT "please let us know. Please write to:\n");
  834. printk(KERN_CONT "linux390@de.ibm.com (mail address) and/or\n");
  835. printk(KERN_CONT "linux-s390@vger.kernel.org (mailing list).\n\n");
  836. printk(KERN_CONT "Thank you!\n\n");
  837. printk(KERN_CONT "If this kernel runs on a 64 bit machine you may consider using a 64 bit kernel.\n");
  838. printk(KERN_CONT "This message can be disabled with the \"no_removal_warning\" kernel parameter.\n");
  839. schedule_timeout_uninterruptible(300 * HZ);
  840. return 0;
  841. }
  842. early_initcall(removal_warning);
  843. #endif