setup.c 25 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934
  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/user.h>
  25. #include <linux/tty.h>
  26. #include <linux/ioport.h>
  27. #include <linux/delay.h>
  28. #include <linux/init.h>
  29. #include <linux/initrd.h>
  30. #include <linux/bootmem.h>
  31. #include <linux/root_dev.h>
  32. #include <linux/console.h>
  33. #include <linux/kernel_stat.h>
  34. #include <linux/device.h>
  35. #include <linux/notifier.h>
  36. #include <linux/pfn.h>
  37. #include <linux/ctype.h>
  38. #include <linux/reboot.h>
  39. #include <linux/topology.h>
  40. #include <linux/ftrace.h>
  41. #include <linux/kexec.h>
  42. #include <linux/crash_dump.h>
  43. #include <linux/memory.h>
  44. #include <linux/compat.h>
  45. #include <asm/ipl.h>
  46. #include <asm/facility.h>
  47. #include <asm/smp.h>
  48. #include <asm/mmu_context.h>
  49. #include <asm/cpcmd.h>
  50. #include <asm/lowcore.h>
  51. #include <asm/irq.h>
  52. #include <asm/page.h>
  53. #include <asm/ptrace.h>
  54. #include <asm/sections.h>
  55. #include <asm/ebcdic.h>
  56. #include <asm/kvm_virtio.h>
  57. #include <asm/diag.h>
  58. #include <asm/os_info.h>
  59. #include <asm/sclp.h>
  60. #include "entry.h"
  61. /*
  62. * Machine setup..
  63. */
  64. unsigned int console_mode = 0;
  65. EXPORT_SYMBOL(console_mode);
  66. unsigned int console_devno = -1;
  67. EXPORT_SYMBOL(console_devno);
  68. unsigned int console_irq = -1;
  69. EXPORT_SYMBOL(console_irq);
  70. unsigned long elf_hwcap = 0;
  71. char elf_platform[ELF_PLATFORM_SIZE];
  72. int __initdata memory_end_set;
  73. unsigned long __initdata memory_end;
  74. unsigned long __initdata max_physmem_end;
  75. unsigned long VMALLOC_START;
  76. EXPORT_SYMBOL(VMALLOC_START);
  77. unsigned long VMALLOC_END;
  78. EXPORT_SYMBOL(VMALLOC_END);
  79. struct page *vmemmap;
  80. EXPORT_SYMBOL(vmemmap);
  81. #ifdef CONFIG_64BIT
  82. unsigned long MODULES_VADDR;
  83. unsigned long MODULES_END;
  84. #endif
  85. /* An array with a pointer to the lowcore of every CPU. */
  86. struct _lowcore *lowcore_ptr[NR_CPUS];
  87. EXPORT_SYMBOL(lowcore_ptr);
  88. /*
  89. * This is set up by the setup-routine at boot-time
  90. * for S390 need to find out, what we have to setup
  91. * using address 0x10400 ...
  92. */
  93. #include <asm/setup.h>
  94. /*
  95. * condev= and conmode= setup parameter.
  96. */
  97. static int __init condev_setup(char *str)
  98. {
  99. int vdev;
  100. vdev = simple_strtoul(str, &str, 0);
  101. if (vdev >= 0 && vdev < 65536) {
  102. console_devno = vdev;
  103. console_irq = -1;
  104. }
  105. return 1;
  106. }
  107. __setup("condev=", condev_setup);
  108. static void __init set_preferred_console(void)
  109. {
  110. if (MACHINE_IS_KVM) {
  111. if (sclp_has_vt220())
  112. add_preferred_console("ttyS", 1, NULL);
  113. else if (sclp_has_linemode())
  114. add_preferred_console("ttyS", 0, NULL);
  115. else
  116. add_preferred_console("hvc", 0, NULL);
  117. } else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
  118. add_preferred_console("ttyS", 0, NULL);
  119. else if (CONSOLE_IS_3270)
  120. add_preferred_console("tty3270", 0, NULL);
  121. }
  122. static int __init conmode_setup(char *str)
  123. {
  124. #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
  125. if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
  126. SET_CONSOLE_SCLP;
  127. #endif
  128. #if defined(CONFIG_TN3215_CONSOLE)
  129. if (strncmp(str, "3215", 5) == 0)
  130. SET_CONSOLE_3215;
  131. #endif
  132. #if defined(CONFIG_TN3270_CONSOLE)
  133. if (strncmp(str, "3270", 5) == 0)
  134. SET_CONSOLE_3270;
  135. #endif
  136. set_preferred_console();
  137. return 1;
  138. }
  139. __setup("conmode=", conmode_setup);
  140. static void __init conmode_default(void)
  141. {
  142. char query_buffer[1024];
  143. char *ptr;
  144. if (MACHINE_IS_VM) {
  145. cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
  146. console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
  147. ptr = strstr(query_buffer, "SUBCHANNEL =");
  148. console_irq = simple_strtoul(ptr + 13, NULL, 16);
  149. cpcmd("QUERY TERM", query_buffer, 1024, NULL);
  150. ptr = strstr(query_buffer, "CONMODE");
  151. /*
  152. * Set the conmode to 3215 so that the device recognition
  153. * will set the cu_type of the console to 3215. If the
  154. * conmode is 3270 and we don't set it back then both
  155. * 3215 and the 3270 driver will try to access the console
  156. * device (3215 as console and 3270 as normal tty).
  157. */
  158. cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
  159. if (ptr == NULL) {
  160. #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
  161. SET_CONSOLE_SCLP;
  162. #endif
  163. return;
  164. }
  165. if (strncmp(ptr + 8, "3270", 4) == 0) {
  166. #if defined(CONFIG_TN3270_CONSOLE)
  167. SET_CONSOLE_3270;
  168. #elif defined(CONFIG_TN3215_CONSOLE)
  169. SET_CONSOLE_3215;
  170. #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
  171. SET_CONSOLE_SCLP;
  172. #endif
  173. } else if (strncmp(ptr + 8, "3215", 4) == 0) {
  174. #if defined(CONFIG_TN3215_CONSOLE)
  175. SET_CONSOLE_3215;
  176. #elif defined(CONFIG_TN3270_CONSOLE)
  177. SET_CONSOLE_3270;
  178. #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
  179. SET_CONSOLE_SCLP;
  180. #endif
  181. }
  182. } else {
  183. #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
  184. SET_CONSOLE_SCLP;
  185. #endif
  186. }
  187. }
  188. #ifdef CONFIG_CRASH_DUMP
  189. static void __init setup_zfcpdump(void)
  190. {
  191. if (ipl_info.type != IPL_TYPE_FCP_DUMP)
  192. return;
  193. if (OLDMEM_BASE)
  194. return;
  195. strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
  196. console_loglevel = 2;
  197. }
  198. #else
  199. static inline void setup_zfcpdump(void) {}
  200. #endif /* CONFIG_CRASH_DUMP */
  201. /*
  202. * Reboot, halt and power_off stubs. They just call _machine_restart,
  203. * _machine_halt or _machine_power_off.
  204. */
  205. void machine_restart(char *command)
  206. {
  207. if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
  208. /*
  209. * Only unblank the console if we are called in enabled
  210. * context or a bust_spinlocks cleared the way for us.
  211. */
  212. console_unblank();
  213. _machine_restart(command);
  214. }
  215. void machine_halt(void)
  216. {
  217. if (!in_interrupt() || oops_in_progress)
  218. /*
  219. * Only unblank the console if we are called in enabled
  220. * context or a bust_spinlocks cleared the way for us.
  221. */
  222. console_unblank();
  223. _machine_halt();
  224. }
  225. void machine_power_off(void)
  226. {
  227. if (!in_interrupt() || oops_in_progress)
  228. /*
  229. * Only unblank the console if we are called in enabled
  230. * context or a bust_spinlocks cleared the way for us.
  231. */
  232. console_unblank();
  233. _machine_power_off();
  234. }
  235. /*
  236. * Dummy power off function.
  237. */
  238. void (*pm_power_off)(void) = machine_power_off;
  239. EXPORT_SYMBOL_GPL(pm_power_off);
  240. static int __init early_parse_mem(char *p)
  241. {
  242. memory_end = memparse(p, &p);
  243. memory_end &= PAGE_MASK;
  244. memory_end_set = 1;
  245. return 0;
  246. }
  247. early_param("mem", early_parse_mem);
  248. static int __init parse_vmalloc(char *arg)
  249. {
  250. if (!arg)
  251. return -EINVAL;
  252. VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
  253. return 0;
  254. }
  255. early_param("vmalloc", parse_vmalloc);
  256. void *restart_stack __attribute__((__section__(".data")));
  257. static void __init setup_lowcore(void)
  258. {
  259. struct _lowcore *lc;
  260. /*
  261. * Setup lowcore for boot cpu
  262. */
  263. BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
  264. lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
  265. lc->restart_psw.mask = PSW_KERNEL_BITS;
  266. lc->restart_psw.addr =
  267. PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
  268. lc->external_new_psw.mask = PSW_KERNEL_BITS |
  269. PSW_MASK_DAT | PSW_MASK_MCHECK;
  270. lc->external_new_psw.addr =
  271. PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
  272. lc->svc_new_psw.mask = PSW_KERNEL_BITS |
  273. PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
  274. lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
  275. lc->program_new_psw.mask = PSW_KERNEL_BITS |
  276. PSW_MASK_DAT | PSW_MASK_MCHECK;
  277. lc->program_new_psw.addr =
  278. PSW_ADDR_AMODE | (unsigned long) pgm_check_handler;
  279. lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
  280. lc->mcck_new_psw.addr =
  281. PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
  282. lc->io_new_psw.mask = PSW_KERNEL_BITS |
  283. PSW_MASK_DAT | PSW_MASK_MCHECK;
  284. lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
  285. lc->clock_comparator = -1ULL;
  286. lc->kernel_stack = ((unsigned long) &init_thread_union)
  287. + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
  288. lc->async_stack = (unsigned long)
  289. __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0)
  290. + ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
  291. lc->panic_stack = (unsigned long)
  292. __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0)
  293. + PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
  294. lc->current_task = (unsigned long) init_thread_union.thread_info.task;
  295. lc->thread_info = (unsigned long) &init_thread_union;
  296. lc->machine_flags = S390_lowcore.machine_flags;
  297. lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
  298. memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
  299. MAX_FACILITY_BIT/8);
  300. #ifndef CONFIG_64BIT
  301. if (MACHINE_HAS_IEEE) {
  302. lc->extended_save_area_addr = (__u32)
  303. __alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
  304. /* enable extended save area */
  305. __ctl_set_bit(14, 29);
  306. }
  307. #else
  308. lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
  309. #endif
  310. lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
  311. lc->async_enter_timer = S390_lowcore.async_enter_timer;
  312. lc->exit_timer = S390_lowcore.exit_timer;
  313. lc->user_timer = S390_lowcore.user_timer;
  314. lc->system_timer = S390_lowcore.system_timer;
  315. lc->steal_timer = S390_lowcore.steal_timer;
  316. lc->last_update_timer = S390_lowcore.last_update_timer;
  317. lc->last_update_clock = S390_lowcore.last_update_clock;
  318. lc->ftrace_func = S390_lowcore.ftrace_func;
  319. restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
  320. restart_stack += ASYNC_SIZE;
  321. /*
  322. * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
  323. * restart data to the absolute zero lowcore. This is necessary if
  324. * PSW restart is done on an offline CPU that has lowcore zero.
  325. */
  326. lc->restart_stack = (unsigned long) restart_stack;
  327. lc->restart_fn = (unsigned long) do_restart;
  328. lc->restart_data = 0;
  329. lc->restart_source = -1UL;
  330. /* Setup absolute zero lowcore */
  331. mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
  332. mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
  333. mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
  334. mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
  335. mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
  336. #ifdef CONFIG_SMP
  337. lc->spinlock_lockval = arch_spin_lockval(0);
  338. #endif
  339. set_prefix((u32)(unsigned long) lc);
  340. lowcore_ptr[0] = lc;
  341. }
  342. static struct resource code_resource = {
  343. .name = "Kernel code",
  344. .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
  345. };
  346. static struct resource data_resource = {
  347. .name = "Kernel data",
  348. .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
  349. };
  350. static struct resource bss_resource = {
  351. .name = "Kernel bss",
  352. .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
  353. };
  354. static struct resource __initdata *standard_resources[] = {
  355. &code_resource,
  356. &data_resource,
  357. &bss_resource,
  358. };
  359. static void __init setup_resources(void)
  360. {
  361. struct resource *res, *std_res, *sub_res;
  362. struct memblock_region *reg;
  363. int j;
  364. code_resource.start = (unsigned long) &_text;
  365. code_resource.end = (unsigned long) &_etext - 1;
  366. data_resource.start = (unsigned long) &_etext;
  367. data_resource.end = (unsigned long) &_edata - 1;
  368. bss_resource.start = (unsigned long) &__bss_start;
  369. bss_resource.end = (unsigned long) &__bss_stop - 1;
  370. for_each_memblock(memory, reg) {
  371. res = alloc_bootmem_low(sizeof(*res));
  372. res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
  373. res->name = "System RAM";
  374. res->start = reg->base;
  375. res->end = reg->base + reg->size - 1;
  376. request_resource(&iomem_resource, res);
  377. for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
  378. std_res = standard_resources[j];
  379. if (std_res->start < res->start ||
  380. std_res->start > res->end)
  381. continue;
  382. if (std_res->end > res->end) {
  383. sub_res = alloc_bootmem_low(sizeof(*sub_res));
  384. *sub_res = *std_res;
  385. sub_res->end = res->end;
  386. std_res->start = res->end + 1;
  387. request_resource(res, sub_res);
  388. } else {
  389. request_resource(res, std_res);
  390. }
  391. }
  392. }
  393. }
  394. static void __init setup_memory_end(void)
  395. {
  396. unsigned long vmax, vmalloc_size, tmp;
  397. /* Choose kernel address space layout: 2, 3, or 4 levels. */
  398. #ifdef CONFIG_64BIT
  399. vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
  400. tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
  401. tmp = tmp * (sizeof(struct page) + PAGE_SIZE) + vmalloc_size;
  402. if (tmp <= (1UL << 42))
  403. vmax = 1UL << 42; /* 3-level kernel page table */
  404. else
  405. vmax = 1UL << 53; /* 4-level kernel page table */
  406. /* module area is at the end of the kernel address space. */
  407. MODULES_END = vmax;
  408. MODULES_VADDR = MODULES_END - MODULES_LEN;
  409. VMALLOC_END = MODULES_VADDR;
  410. #else
  411. vmalloc_size = VMALLOC_END ?: 96UL << 20;
  412. vmax = 1UL << 31; /* 2-level kernel page table */
  413. /* vmalloc area is at the end of the kernel address space. */
  414. VMALLOC_END = vmax;
  415. #endif
  416. VMALLOC_START = vmax - vmalloc_size;
  417. /* Split remaining virtual space between 1:1 mapping & vmemmap array */
  418. tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
  419. /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
  420. tmp = SECTION_ALIGN_UP(tmp);
  421. tmp = VMALLOC_START - tmp * sizeof(struct page);
  422. tmp &= ~((vmax >> 11) - 1); /* align to page table level */
  423. tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
  424. vmemmap = (struct page *) tmp;
  425. /* Take care that memory_end is set and <= vmemmap */
  426. memory_end = min(memory_end ?: max_physmem_end, tmp);
  427. max_pfn = max_low_pfn = PFN_DOWN(memory_end);
  428. memblock_remove(memory_end, ULONG_MAX);
  429. pr_notice("Max memory size: %luMB\n", memory_end >> 20);
  430. }
  431. static void __init setup_vmcoreinfo(void)
  432. {
  433. mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note());
  434. }
  435. #ifdef CONFIG_CRASH_DUMP
  436. /*
  437. * When kdump is enabled, we have to ensure that no memory from
  438. * the area [0 - crashkernel memory size] and
  439. * [crashk_res.start - crashk_res.end] is set offline.
  440. */
  441. static int kdump_mem_notifier(struct notifier_block *nb,
  442. unsigned long action, void *data)
  443. {
  444. struct memory_notify *arg = data;
  445. if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
  446. return NOTIFY_BAD;
  447. if (arg->start_pfn > PFN_DOWN(crashk_res.end))
  448. return NOTIFY_OK;
  449. if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
  450. return NOTIFY_OK;
  451. return NOTIFY_BAD;
  452. }
  453. static struct notifier_block kdump_mem_nb = {
  454. .notifier_call = kdump_mem_notifier,
  455. };
  456. #endif
  457. /*
  458. * Make sure that the area behind memory_end is protected
  459. */
  460. static void reserve_memory_end(void)
  461. {
  462. #ifdef CONFIG_CRASH_DUMP
  463. if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
  464. !OLDMEM_BASE && sclp_get_hsa_size()) {
  465. memory_end = sclp_get_hsa_size();
  466. memory_end &= PAGE_MASK;
  467. memory_end_set = 1;
  468. }
  469. #endif
  470. if (!memory_end_set)
  471. return;
  472. memblock_reserve(memory_end, ULONG_MAX);
  473. }
  474. /*
  475. * Make sure that oldmem, where the dump is stored, is protected
  476. */
  477. static void reserve_oldmem(void)
  478. {
  479. #ifdef CONFIG_CRASH_DUMP
  480. if (OLDMEM_BASE)
  481. /* Forget all memory above the running kdump system */
  482. memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
  483. #endif
  484. }
  485. /*
  486. * Make sure that oldmem, where the dump is stored, is protected
  487. */
  488. static void remove_oldmem(void)
  489. {
  490. #ifdef CONFIG_CRASH_DUMP
  491. if (OLDMEM_BASE)
  492. /* Forget all memory above the running kdump system */
  493. memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
  494. #endif
  495. }
  496. /*
  497. * Reserve memory for kdump kernel to be loaded with kexec
  498. */
  499. static void __init reserve_crashkernel(void)
  500. {
  501. #ifdef CONFIG_CRASH_DUMP
  502. unsigned long long crash_base, crash_size;
  503. phys_addr_t low, high;
  504. int rc;
  505. rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
  506. &crash_base);
  507. crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
  508. crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
  509. if (rc || crash_size == 0)
  510. return;
  511. if (memblock.memory.regions[0].size < crash_size) {
  512. pr_info("crashkernel reservation failed: %s\n",
  513. "first memory chunk must be at least crashkernel size");
  514. return;
  515. }
  516. low = crash_base ?: OLDMEM_BASE;
  517. high = low + crash_size;
  518. if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
  519. /* The crashkernel fits into OLDMEM, reuse OLDMEM */
  520. crash_base = low;
  521. } else {
  522. /* Find suitable area in free memory */
  523. low = max_t(unsigned long, crash_size, sclp_get_hsa_size());
  524. high = crash_base ? crash_base + crash_size : ULONG_MAX;
  525. if (crash_base && crash_base < low) {
  526. pr_info("crashkernel reservation failed: %s\n",
  527. "crash_base too low");
  528. return;
  529. }
  530. low = crash_base ?: low;
  531. crash_base = memblock_find_in_range(low, high, crash_size,
  532. KEXEC_CRASH_MEM_ALIGN);
  533. }
  534. if (!crash_base) {
  535. pr_info("crashkernel reservation failed: %s\n",
  536. "no suitable area found");
  537. return;
  538. }
  539. if (register_memory_notifier(&kdump_mem_nb))
  540. return;
  541. if (!OLDMEM_BASE && MACHINE_IS_VM)
  542. diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
  543. crashk_res.start = crash_base;
  544. crashk_res.end = crash_base + crash_size - 1;
  545. insert_resource(&iomem_resource, &crashk_res);
  546. memblock_remove(crash_base, crash_size);
  547. pr_info("Reserving %lluMB of memory at %lluMB "
  548. "for crashkernel (System RAM: %luMB)\n",
  549. crash_size >> 20, crash_base >> 20,
  550. (unsigned long)memblock.memory.total_size >> 20);
  551. os_info_crashkernel_add(crash_base, crash_size);
  552. #endif
  553. }
  554. /*
  555. * Reserve the initrd from being used by memblock
  556. */
  557. static void __init reserve_initrd(void)
  558. {
  559. #ifdef CONFIG_BLK_DEV_INITRD
  560. initrd_start = INITRD_START;
  561. initrd_end = initrd_start + INITRD_SIZE;
  562. memblock_reserve(INITRD_START, INITRD_SIZE);
  563. #endif
  564. }
  565. /*
  566. * Check for initrd being in usable memory
  567. */
  568. static void __init check_initrd(void)
  569. {
  570. #ifdef CONFIG_BLK_DEV_INITRD
  571. if (INITRD_START && INITRD_SIZE &&
  572. !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
  573. pr_err("initrd does not fit memory.\n");
  574. memblock_free(INITRD_START, INITRD_SIZE);
  575. initrd_start = initrd_end = 0;
  576. }
  577. #endif
  578. }
  579. /*
  580. * Reserve all kernel text
  581. */
  582. static void __init reserve_kernel(void)
  583. {
  584. unsigned long start_pfn;
  585. start_pfn = PFN_UP(__pa(&_end));
  586. /*
  587. * Reserve memory used for lowcore/command line/kernel image.
  588. */
  589. memblock_reserve(0, (unsigned long)_ehead);
  590. memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
  591. - (unsigned long)_stext);
  592. }
  593. static void __init reserve_elfcorehdr(void)
  594. {
  595. #ifdef CONFIG_CRASH_DUMP
  596. if (is_kdump_kernel())
  597. memblock_reserve(elfcorehdr_addr - OLDMEM_BASE,
  598. PAGE_ALIGN(elfcorehdr_size));
  599. #endif
  600. }
  601. static void __init setup_memory(void)
  602. {
  603. struct memblock_region *reg;
  604. /*
  605. * Init storage key for present memory
  606. */
  607. for_each_memblock(memory, reg) {
  608. storage_key_init_range(reg->base, reg->base + reg->size);
  609. }
  610. psw_set_key(PAGE_DEFAULT_KEY);
  611. /* Only cosmetics */
  612. memblock_enforce_memory_limit(memblock_end_of_DRAM());
  613. }
  614. /*
  615. * Setup hardware capabilities.
  616. */
  617. static void __init setup_hwcaps(void)
  618. {
  619. static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
  620. struct cpuid cpu_id;
  621. int i;
  622. /*
  623. * The store facility list bits numbers as found in the principles
  624. * of operation are numbered with bit 1UL<<31 as number 0 to
  625. * bit 1UL<<0 as number 31.
  626. * Bit 0: instructions named N3, "backported" to esa-mode
  627. * Bit 2: z/Architecture mode is active
  628. * Bit 7: the store-facility-list-extended facility is installed
  629. * Bit 17: the message-security assist is installed
  630. * Bit 19: the long-displacement facility is installed
  631. * Bit 21: the extended-immediate facility is installed
  632. * Bit 22: extended-translation facility 3 is installed
  633. * Bit 30: extended-translation facility 3 enhancement facility
  634. * These get translated to:
  635. * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
  636. * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
  637. * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
  638. * HWCAP_S390_ETF3EH bit 8 (22 && 30).
  639. */
  640. for (i = 0; i < 6; i++)
  641. if (test_facility(stfl_bits[i]))
  642. elf_hwcap |= 1UL << i;
  643. if (test_facility(22) && test_facility(30))
  644. elf_hwcap |= HWCAP_S390_ETF3EH;
  645. /*
  646. * Check for additional facilities with store-facility-list-extended.
  647. * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
  648. * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
  649. * as stored by stfl, bits 32-xxx contain additional facilities.
  650. * How many facility words are stored depends on the number of
  651. * doublewords passed to the instruction. The additional facilities
  652. * are:
  653. * Bit 42: decimal floating point facility is installed
  654. * Bit 44: perform floating point operation facility is installed
  655. * translated to:
  656. * HWCAP_S390_DFP bit 6 (42 && 44).
  657. */
  658. if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
  659. elf_hwcap |= HWCAP_S390_DFP;
  660. /*
  661. * Huge page support HWCAP_S390_HPAGE is bit 7.
  662. */
  663. if (MACHINE_HAS_HPAGE)
  664. elf_hwcap |= HWCAP_S390_HPAGE;
  665. #if defined(CONFIG_64BIT)
  666. /*
  667. * 64-bit register support for 31-bit processes
  668. * HWCAP_S390_HIGH_GPRS is bit 9.
  669. */
  670. elf_hwcap |= HWCAP_S390_HIGH_GPRS;
  671. /*
  672. * Transactional execution support HWCAP_S390_TE is bit 10.
  673. */
  674. if (test_facility(50) && test_facility(73))
  675. elf_hwcap |= HWCAP_S390_TE;
  676. #endif
  677. get_cpu_id(&cpu_id);
  678. switch (cpu_id.machine) {
  679. case 0x9672:
  680. #if !defined(CONFIG_64BIT)
  681. default: /* Use "g5" as default for 31 bit kernels. */
  682. #endif
  683. strcpy(elf_platform, "g5");
  684. break;
  685. case 0x2064:
  686. case 0x2066:
  687. #if defined(CONFIG_64BIT)
  688. default: /* Use "z900" as default for 64 bit kernels. */
  689. #endif
  690. strcpy(elf_platform, "z900");
  691. break;
  692. case 0x2084:
  693. case 0x2086:
  694. strcpy(elf_platform, "z990");
  695. break;
  696. case 0x2094:
  697. case 0x2096:
  698. strcpy(elf_platform, "z9-109");
  699. break;
  700. case 0x2097:
  701. case 0x2098:
  702. strcpy(elf_platform, "z10");
  703. break;
  704. case 0x2817:
  705. case 0x2818:
  706. strcpy(elf_platform, "z196");
  707. break;
  708. case 0x2827:
  709. case 0x2828:
  710. strcpy(elf_platform, "zEC12");
  711. break;
  712. }
  713. }
  714. /*
  715. * Setup function called from init/main.c just after the banner
  716. * was printed.
  717. */
  718. void __init setup_arch(char **cmdline_p)
  719. {
  720. /*
  721. * print what head.S has found out about the machine
  722. */
  723. #ifndef CONFIG_64BIT
  724. if (MACHINE_IS_VM)
  725. pr_info("Linux is running as a z/VM "
  726. "guest operating system in 31-bit mode\n");
  727. else if (MACHINE_IS_LPAR)
  728. pr_info("Linux is running natively in 31-bit mode\n");
  729. if (MACHINE_HAS_IEEE)
  730. pr_info("The hardware system has IEEE compatible "
  731. "floating point units\n");
  732. else
  733. pr_info("The hardware system has no IEEE compatible "
  734. "floating point units\n");
  735. #else /* CONFIG_64BIT */
  736. if (MACHINE_IS_VM)
  737. pr_info("Linux is running as a z/VM "
  738. "guest operating system in 64-bit mode\n");
  739. else if (MACHINE_IS_KVM)
  740. pr_info("Linux is running under KVM in 64-bit mode\n");
  741. else if (MACHINE_IS_LPAR)
  742. pr_info("Linux is running natively in 64-bit mode\n");
  743. #endif /* CONFIG_64BIT */
  744. /* Have one command line that is parsed and saved in /proc/cmdline */
  745. /* boot_command_line has been already set up in early.c */
  746. *cmdline_p = boot_command_line;
  747. ROOT_DEV = Root_RAM0;
  748. /* Is init_mm really needed? */
  749. init_mm.start_code = PAGE_OFFSET;
  750. init_mm.end_code = (unsigned long) &_etext;
  751. init_mm.end_data = (unsigned long) &_edata;
  752. init_mm.brk = (unsigned long) &_end;
  753. parse_early_param();
  754. os_info_init();
  755. setup_ipl();
  756. /* Do some memory reservations *before* memory is added to memblock */
  757. reserve_memory_end();
  758. reserve_oldmem();
  759. reserve_kernel();
  760. reserve_initrd();
  761. reserve_elfcorehdr();
  762. memblock_allow_resize();
  763. /* Get information about *all* installed memory */
  764. detect_memory_memblock();
  765. remove_oldmem();
  766. /*
  767. * Make sure all chunks are MAX_ORDER aligned so we don't need the
  768. * extra checks that HOLES_IN_ZONE would require.
  769. *
  770. * Is this still required?
  771. */
  772. memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
  773. setup_memory_end();
  774. setup_memory();
  775. check_initrd();
  776. reserve_crashkernel();
  777. setup_resources();
  778. setup_vmcoreinfo();
  779. setup_lowcore();
  780. smp_fill_possible_mask();
  781. cpu_init();
  782. s390_init_cpu_topology();
  783. /*
  784. * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
  785. */
  786. setup_hwcaps();
  787. /*
  788. * Create kernel page tables and switch to virtual addressing.
  789. */
  790. paging_init();
  791. /* Setup default console */
  792. conmode_default();
  793. set_preferred_console();
  794. /* Setup zfcpdump support */
  795. setup_zfcpdump();
  796. }
  797. #ifdef CONFIG_32BIT
  798. static int no_removal_warning __initdata;
  799. static int __init parse_no_removal_warning(char *str)
  800. {
  801. no_removal_warning = 1;
  802. return 0;
  803. }
  804. __setup("no_removal_warning", parse_no_removal_warning);
  805. static int __init removal_warning(void)
  806. {
  807. if (no_removal_warning)
  808. return 0;
  809. printk(KERN_ALERT "\n\n");
  810. printk(KERN_CONT "Warning - you are using a 31 bit kernel!\n\n");
  811. printk(KERN_CONT "We plan to remove 31 bit kernel support from the kernel sources in March 2015.\n");
  812. printk(KERN_CONT "Currently we assume that nobody is using the 31 bit kernel on old 31 bit\n");
  813. printk(KERN_CONT "hardware anymore. If you think that the code should not be removed and also\n");
  814. printk(KERN_CONT "future versions of the Linux kernel should be able to run in 31 bit mode\n");
  815. printk(KERN_CONT "please let us know. Please write to:\n");
  816. printk(KERN_CONT "linux390@de.ibm.com (mail address) and/or\n");
  817. printk(KERN_CONT "linux-s390@vger.kernel.org (mailing list).\n\n");
  818. printk(KERN_CONT "Thank you!\n\n");
  819. printk(KERN_CONT "If this kernel runs on a 64 bit machine you may consider using a 64 bit kernel.\n");
  820. printk(KERN_CONT "This message can be disabled with the \"no_removal_warning\" kernel parameter.\n");
  821. schedule_timeout_uninterruptible(300 * HZ);
  822. return 0;
  823. }
  824. early_initcall(removal_warning);
  825. #endif