setup_64.c 21 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806
  1. /*
  2. *
  3. * Common boot and setup code.
  4. *
  5. * Copyright (C) 2001 PPC64 Team, IBM Corp
  6. *
  7. * This program is free software; you can redistribute it and/or
  8. * modify it under the terms of the GNU General Public License
  9. * as published by the Free Software Foundation; either version
  10. * 2 of the License, or (at your option) any later version.
  11. */
  12. #define DEBUG
  13. #include <linux/export.h>
  14. #include <linux/string.h>
  15. #include <linux/sched.h>
  16. #include <linux/init.h>
  17. #include <linux/kernel.h>
  18. #include <linux/reboot.h>
  19. #include <linux/delay.h>
  20. #include <linux/initrd.h>
  21. #include <linux/seq_file.h>
  22. #include <linux/ioport.h>
  23. #include <linux/console.h>
  24. #include <linux/utsname.h>
  25. #include <linux/tty.h>
  26. #include <linux/root_dev.h>
  27. #include <linux/notifier.h>
  28. #include <linux/cpu.h>
  29. #include <linux/unistd.h>
  30. #include <linux/serial.h>
  31. #include <linux/serial_8250.h>
  32. #include <linux/bootmem.h>
  33. #include <linux/pci.h>
  34. #include <linux/lockdep.h>
  35. #include <linux/memblock.h>
  36. #include <linux/hugetlb.h>
  37. #include <linux/memory.h>
  38. #include <linux/nmi.h>
  39. #include <asm/io.h>
  40. #include <asm/kdump.h>
  41. #include <asm/prom.h>
  42. #include <asm/processor.h>
  43. #include <asm/pgtable.h>
  44. #include <asm/smp.h>
  45. #include <asm/elf.h>
  46. #include <asm/machdep.h>
  47. #include <asm/paca.h>
  48. #include <asm/time.h>
  49. #include <asm/cputable.h>
  50. #include <asm/sections.h>
  51. #include <asm/btext.h>
  52. #include <asm/nvram.h>
  53. #include <asm/setup.h>
  54. #include <asm/rtas.h>
  55. #include <asm/iommu.h>
  56. #include <asm/serial.h>
  57. #include <asm/cache.h>
  58. #include <asm/page.h>
  59. #include <asm/mmu.h>
  60. #include <asm/firmware.h>
  61. #include <asm/xmon.h>
  62. #include <asm/udbg.h>
  63. #include <asm/kexec.h>
  64. #include <asm/mmu_context.h>
  65. #include <asm/code-patching.h>
  66. #include <asm/kvm_ppc.h>
  67. #include <asm/hugetlb.h>
  68. #include <asm/epapr_hcalls.h>
  69. #ifdef DEBUG
  70. #define DBG(fmt...) udbg_printf(fmt)
  71. #else
  72. #define DBG(fmt...)
  73. #endif
  74. int spinning_secondaries;
  75. u64 ppc64_pft_size;
  76. /* Pick defaults since we might want to patch instructions
  77. * before we've read this from the device tree.
  78. */
  79. struct ppc64_caches ppc64_caches = {
  80. .dline_size = 0x40,
  81. .log_dline_size = 6,
  82. .iline_size = 0x40,
  83. .log_iline_size = 6
  84. };
  85. EXPORT_SYMBOL_GPL(ppc64_caches);
  86. /*
  87. * These are used in binfmt_elf.c to put aux entries on the stack
  88. * for each elf executable being started.
  89. */
  90. int dcache_bsize;
  91. int icache_bsize;
  92. int ucache_bsize;
  93. #if defined(CONFIG_PPC_BOOK3E) && defined(CONFIG_SMP)
  94. static void setup_tlb_core_data(void)
  95. {
  96. int cpu;
  97. BUILD_BUG_ON(offsetof(struct tlb_core_data, lock) != 0);
  98. for_each_possible_cpu(cpu) {
  99. int first = cpu_first_thread_sibling(cpu);
  100. paca[cpu].tcd_ptr = &paca[first].tcd;
  101. /*
  102. * If we have threads, we need either tlbsrx.
  103. * or e6500 tablewalk mode, or else TLB handlers
  104. * will be racy and could produce duplicate entries.
  105. */
  106. if (smt_enabled_at_boot >= 2 &&
  107. !mmu_has_feature(MMU_FTR_USE_TLBRSRV) &&
  108. book3e_htw_mode != PPC_HTW_E6500) {
  109. /* Should we panic instead? */
  110. WARN_ONCE("%s: unsupported MMU configuration -- expect problems\n",
  111. __func__);
  112. }
  113. }
  114. }
  115. #else
  116. static void setup_tlb_core_data(void)
  117. {
  118. }
  119. #endif
  120. #ifdef CONFIG_SMP
  121. static char *smt_enabled_cmdline;
  122. /* Look for ibm,smt-enabled OF option */
  123. static void check_smt_enabled(void)
  124. {
  125. struct device_node *dn;
  126. const char *smt_option;
  127. /* Default to enabling all threads */
  128. smt_enabled_at_boot = threads_per_core;
  129. /* Allow the command line to overrule the OF option */
  130. if (smt_enabled_cmdline) {
  131. if (!strcmp(smt_enabled_cmdline, "on"))
  132. smt_enabled_at_boot = threads_per_core;
  133. else if (!strcmp(smt_enabled_cmdline, "off"))
  134. smt_enabled_at_boot = 0;
  135. else {
  136. int smt;
  137. int rc;
  138. rc = kstrtoint(smt_enabled_cmdline, 10, &smt);
  139. if (!rc)
  140. smt_enabled_at_boot =
  141. min(threads_per_core, smt);
  142. }
  143. } else {
  144. dn = of_find_node_by_path("/options");
  145. if (dn) {
  146. smt_option = of_get_property(dn, "ibm,smt-enabled",
  147. NULL);
  148. if (smt_option) {
  149. if (!strcmp(smt_option, "on"))
  150. smt_enabled_at_boot = threads_per_core;
  151. else if (!strcmp(smt_option, "off"))
  152. smt_enabled_at_boot = 0;
  153. }
  154. of_node_put(dn);
  155. }
  156. }
  157. }
  158. /* Look for smt-enabled= cmdline option */
  159. static int __init early_smt_enabled(char *p)
  160. {
  161. smt_enabled_cmdline = p;
  162. return 0;
  163. }
  164. early_param("smt-enabled", early_smt_enabled);
  165. #else
  166. #define check_smt_enabled()
  167. #endif /* CONFIG_SMP */
  168. /** Fix up paca fields required for the boot cpu */
  169. static void fixup_boot_paca(void)
  170. {
  171. /* The boot cpu is started */
  172. get_paca()->cpu_start = 1;
  173. /* Allow percpu accesses to work until we setup percpu data */
  174. get_paca()->data_offset = 0;
  175. }
  176. static void cpu_ready_for_interrupts(void)
  177. {
  178. /* Set IR and DR in PACA MSR */
  179. get_paca()->kernel_msr = MSR_KERNEL;
  180. /*
  181. * Enable AIL if supported, and we are in hypervisor mode. If we are
  182. * not in hypervisor mode, we enable relocation-on interrupts later
  183. * in pSeries_setup_arch() using the H_SET_MODE hcall.
  184. */
  185. if (cpu_has_feature(CPU_FTR_HVMODE) &&
  186. cpu_has_feature(CPU_FTR_ARCH_207S)) {
  187. unsigned long lpcr = mfspr(SPRN_LPCR);
  188. mtspr(SPRN_LPCR, lpcr | LPCR_AIL_3);
  189. }
  190. }
  191. /*
  192. * Early initialization entry point. This is called by head.S
  193. * with MMU translation disabled. We rely on the "feature" of
  194. * the CPU that ignores the top 2 bits of the address in real
  195. * mode so we can access kernel globals normally provided we
  196. * only toy with things in the RMO region. From here, we do
  197. * some early parsing of the device-tree to setup out MEMBLOCK
  198. * data structures, and allocate & initialize the hash table
  199. * and segment tables so we can start running with translation
  200. * enabled.
  201. *
  202. * It is this function which will call the probe() callback of
  203. * the various platform types and copy the matching one to the
  204. * global ppc_md structure. Your platform can eventually do
  205. * some very early initializations from the probe() routine, but
  206. * this is not recommended, be very careful as, for example, the
  207. * device-tree is not accessible via normal means at this point.
  208. */
  209. void __init early_setup(unsigned long dt_ptr)
  210. {
  211. static __initdata struct paca_struct boot_paca;
  212. /* -------- printk is _NOT_ safe to use here ! ------- */
  213. /* Identify CPU type */
  214. identify_cpu(0, mfspr(SPRN_PVR));
  215. /* Assume we're on cpu 0 for now. Don't write to the paca yet! */
  216. initialise_paca(&boot_paca, 0);
  217. setup_paca(&boot_paca);
  218. fixup_boot_paca();
  219. /* Initialize lockdep early or else spinlocks will blow */
  220. lockdep_init();
  221. /* -------- printk is now safe to use ------- */
  222. /* Enable early debugging if any specified (see udbg.h) */
  223. udbg_early_init();
  224. DBG(" -> early_setup(), dt_ptr: 0x%lx\n", dt_ptr);
  225. /*
  226. * Do early initialization using the flattened device
  227. * tree, such as retrieving the physical memory map or
  228. * calculating/retrieving the hash table size.
  229. */
  230. early_init_devtree(__va(dt_ptr));
  231. epapr_paravirt_early_init();
  232. /* Now we know the logical id of our boot cpu, setup the paca. */
  233. setup_paca(&paca[boot_cpuid]);
  234. fixup_boot_paca();
  235. /* Probe the machine type */
  236. probe_machine();
  237. setup_kdump_trampoline();
  238. DBG("Found, Initializing memory management...\n");
  239. /* Initialize the hash table or TLB handling */
  240. early_init_mmu();
  241. /*
  242. * At this point, we can let interrupts switch to virtual mode
  243. * (the MMU has been setup), so adjust the MSR in the PACA to
  244. * have IR and DR set and enable AIL if it exists
  245. */
  246. cpu_ready_for_interrupts();
  247. /* Reserve large chunks of memory for use by CMA for KVM */
  248. kvm_cma_reserve();
  249. /*
  250. * Reserve any gigantic pages requested on the command line.
  251. * memblock needs to have been initialized by the time this is
  252. * called since this will reserve memory.
  253. */
  254. reserve_hugetlb_gpages();
  255. DBG(" <- early_setup()\n");
  256. #ifdef CONFIG_PPC_EARLY_DEBUG_BOOTX
  257. /*
  258. * This needs to be done *last* (after the above DBG() even)
  259. *
  260. * Right after we return from this function, we turn on the MMU
  261. * which means the real-mode access trick that btext does will
  262. * no longer work, it needs to switch to using a real MMU
  263. * mapping. This call will ensure that it does
  264. */
  265. btext_map();
  266. #endif /* CONFIG_PPC_EARLY_DEBUG_BOOTX */
  267. }
  268. #ifdef CONFIG_SMP
  269. void early_setup_secondary(void)
  270. {
  271. /* Mark interrupts enabled in PACA */
  272. get_paca()->soft_enabled = 0;
  273. /* Initialize the hash table or TLB handling */
  274. early_init_mmu_secondary();
  275. /*
  276. * At this point, we can let interrupts switch to virtual mode
  277. * (the MMU has been setup), so adjust the MSR in the PACA to
  278. * have IR and DR set.
  279. */
  280. cpu_ready_for_interrupts();
  281. }
  282. #endif /* CONFIG_SMP */
  283. #if defined(CONFIG_SMP) || defined(CONFIG_KEXEC)
  284. void smp_release_cpus(void)
  285. {
  286. unsigned long *ptr;
  287. int i;
  288. DBG(" -> smp_release_cpus()\n");
  289. /* All secondary cpus are spinning on a common spinloop, release them
  290. * all now so they can start to spin on their individual paca
  291. * spinloops. For non SMP kernels, the secondary cpus never get out
  292. * of the common spinloop.
  293. */
  294. ptr = (unsigned long *)((unsigned long)&__secondary_hold_spinloop
  295. - PHYSICAL_START);
  296. *ptr = ppc_function_entry(generic_secondary_smp_init);
  297. /* And wait a bit for them to catch up */
  298. for (i = 0; i < 100000; i++) {
  299. mb();
  300. HMT_low();
  301. if (spinning_secondaries == 0)
  302. break;
  303. udelay(1);
  304. }
  305. DBG("spinning_secondaries = %d\n", spinning_secondaries);
  306. DBG(" <- smp_release_cpus()\n");
  307. }
  308. #endif /* CONFIG_SMP || CONFIG_KEXEC */
  309. /*
  310. * Initialize some remaining members of the ppc64_caches and systemcfg
  311. * structures
  312. * (at least until we get rid of them completely). This is mostly some
  313. * cache informations about the CPU that will be used by cache flush
  314. * routines and/or provided to userland
  315. */
  316. static void __init initialize_cache_info(void)
  317. {
  318. struct device_node *np;
  319. unsigned long num_cpus = 0;
  320. DBG(" -> initialize_cache_info()\n");
  321. for_each_node_by_type(np, "cpu") {
  322. num_cpus += 1;
  323. /*
  324. * We're assuming *all* of the CPUs have the same
  325. * d-cache and i-cache sizes... -Peter
  326. */
  327. if (num_cpus == 1) {
  328. const __be32 *sizep, *lsizep;
  329. u32 size, lsize;
  330. size = 0;
  331. lsize = cur_cpu_spec->dcache_bsize;
  332. sizep = of_get_property(np, "d-cache-size", NULL);
  333. if (sizep != NULL)
  334. size = be32_to_cpu(*sizep);
  335. lsizep = of_get_property(np, "d-cache-block-size",
  336. NULL);
  337. /* fallback if block size missing */
  338. if (lsizep == NULL)
  339. lsizep = of_get_property(np,
  340. "d-cache-line-size",
  341. NULL);
  342. if (lsizep != NULL)
  343. lsize = be32_to_cpu(*lsizep);
  344. if (sizep == NULL || lsizep == NULL)
  345. DBG("Argh, can't find dcache properties ! "
  346. "sizep: %p, lsizep: %p\n", sizep, lsizep);
  347. ppc64_caches.dsize = size;
  348. ppc64_caches.dline_size = lsize;
  349. ppc64_caches.log_dline_size = __ilog2(lsize);
  350. ppc64_caches.dlines_per_page = PAGE_SIZE / lsize;
  351. size = 0;
  352. lsize = cur_cpu_spec->icache_bsize;
  353. sizep = of_get_property(np, "i-cache-size", NULL);
  354. if (sizep != NULL)
  355. size = be32_to_cpu(*sizep);
  356. lsizep = of_get_property(np, "i-cache-block-size",
  357. NULL);
  358. if (lsizep == NULL)
  359. lsizep = of_get_property(np,
  360. "i-cache-line-size",
  361. NULL);
  362. if (lsizep != NULL)
  363. lsize = be32_to_cpu(*lsizep);
  364. if (sizep == NULL || lsizep == NULL)
  365. DBG("Argh, can't find icache properties ! "
  366. "sizep: %p, lsizep: %p\n", sizep, lsizep);
  367. ppc64_caches.isize = size;
  368. ppc64_caches.iline_size = lsize;
  369. ppc64_caches.log_iline_size = __ilog2(lsize);
  370. ppc64_caches.ilines_per_page = PAGE_SIZE / lsize;
  371. }
  372. }
  373. DBG(" <- initialize_cache_info()\n");
  374. }
  375. /*
  376. * Do some initial setup of the system. The parameters are those which
  377. * were passed in from the bootloader.
  378. */
  379. void __init setup_system(void)
  380. {
  381. DBG(" -> setup_system()\n");
  382. /* Apply the CPUs-specific and firmware specific fixups to kernel
  383. * text (nop out sections not relevant to this CPU or this firmware)
  384. */
  385. do_feature_fixups(cur_cpu_spec->cpu_features,
  386. &__start___ftr_fixup, &__stop___ftr_fixup);
  387. do_feature_fixups(cur_cpu_spec->mmu_features,
  388. &__start___mmu_ftr_fixup, &__stop___mmu_ftr_fixup);
  389. do_feature_fixups(powerpc_firmware_features,
  390. &__start___fw_ftr_fixup, &__stop___fw_ftr_fixup);
  391. do_lwsync_fixups(cur_cpu_spec->cpu_features,
  392. &__start___lwsync_fixup, &__stop___lwsync_fixup);
  393. do_final_fixups();
  394. /*
  395. * Unflatten the device-tree passed by prom_init or kexec
  396. */
  397. unflatten_device_tree();
  398. /*
  399. * Fill the ppc64_caches & systemcfg structures with informations
  400. * retrieved from the device-tree.
  401. */
  402. initialize_cache_info();
  403. #ifdef CONFIG_PPC_RTAS
  404. /*
  405. * Initialize RTAS if available
  406. */
  407. rtas_initialize();
  408. #endif /* CONFIG_PPC_RTAS */
  409. /*
  410. * Check if we have an initrd provided via the device-tree
  411. */
  412. check_for_initrd();
  413. /*
  414. * Do some platform specific early initializations, that includes
  415. * setting up the hash table pointers. It also sets up some interrupt-mapping
  416. * related options that will be used by finish_device_tree()
  417. */
  418. if (ppc_md.init_early)
  419. ppc_md.init_early();
  420. /*
  421. * We can discover serial ports now since the above did setup the
  422. * hash table management for us, thus ioremap works. We do that early
  423. * so that further code can be debugged
  424. */
  425. find_legacy_serial_ports();
  426. /*
  427. * Register early console
  428. */
  429. register_early_udbg_console();
  430. /*
  431. * Initialize xmon
  432. */
  433. xmon_setup();
  434. smp_setup_cpu_maps();
  435. check_smt_enabled();
  436. setup_tlb_core_data();
  437. /*
  438. * Freescale Book3e parts spin in a loop provided by firmware,
  439. * so smp_release_cpus() does nothing for them
  440. */
  441. #if defined(CONFIG_SMP) && !defined(CONFIG_PPC_FSL_BOOK3E)
  442. /* Release secondary cpus out of their spinloops at 0x60 now that
  443. * we can map physical -> logical CPU ids
  444. */
  445. smp_release_cpus();
  446. #endif
  447. pr_info("Starting Linux %s %s\n", init_utsname()->machine,
  448. init_utsname()->version);
  449. pr_info("-----------------------------------------------------\n");
  450. pr_info("ppc64_pft_size = 0x%llx\n", ppc64_pft_size);
  451. pr_info("phys_mem_size = 0x%llx\n", memblock_phys_mem_size());
  452. if (ppc64_caches.dline_size != 0x80)
  453. pr_info("dcache_line_size = 0x%x\n", ppc64_caches.dline_size);
  454. if (ppc64_caches.iline_size != 0x80)
  455. pr_info("icache_line_size = 0x%x\n", ppc64_caches.iline_size);
  456. pr_info("cpu_features = 0x%016lx\n", cur_cpu_spec->cpu_features);
  457. pr_info(" possible = 0x%016lx\n", CPU_FTRS_POSSIBLE);
  458. pr_info(" always = 0x%016lx\n", CPU_FTRS_ALWAYS);
  459. pr_info("cpu_user_features = 0x%08x 0x%08x\n", cur_cpu_spec->cpu_user_features,
  460. cur_cpu_spec->cpu_user_features2);
  461. pr_info("mmu_features = 0x%08x\n", cur_cpu_spec->mmu_features);
  462. pr_info("firmware_features = 0x%016lx\n", powerpc_firmware_features);
  463. #ifdef CONFIG_PPC_STD_MMU_64
  464. if (htab_address)
  465. pr_info("htab_address = 0x%p\n", htab_address);
  466. pr_info("htab_hash_mask = 0x%lx\n", htab_hash_mask);
  467. #endif
  468. if (PHYSICAL_START > 0)
  469. pr_info("physical_start = 0x%llx\n",
  470. (unsigned long long)PHYSICAL_START);
  471. pr_info("-----------------------------------------------------\n");
  472. DBG(" <- setup_system()\n");
  473. }
  474. /* This returns the limit below which memory accesses to the linear
  475. * mapping are guarnateed not to cause a TLB or SLB miss. This is
  476. * used to allocate interrupt or emergency stacks for which our
  477. * exception entry path doesn't deal with being interrupted.
  478. */
  479. static u64 safe_stack_limit(void)
  480. {
  481. #ifdef CONFIG_PPC_BOOK3E
  482. /* Freescale BookE bolts the entire linear mapping */
  483. if (mmu_has_feature(MMU_FTR_TYPE_FSL_E))
  484. return linear_map_top;
  485. /* Other BookE, we assume the first GB is bolted */
  486. return 1ul << 30;
  487. #else
  488. /* BookS, the first segment is bolted */
  489. if (mmu_has_feature(MMU_FTR_1T_SEGMENT))
  490. return 1UL << SID_SHIFT_1T;
  491. return 1UL << SID_SHIFT;
  492. #endif
  493. }
  494. static void __init irqstack_early_init(void)
  495. {
  496. u64 limit = safe_stack_limit();
  497. unsigned int i;
  498. /*
  499. * Interrupt stacks must be in the first segment since we
  500. * cannot afford to take SLB misses on them.
  501. */
  502. for_each_possible_cpu(i) {
  503. softirq_ctx[i] = (struct thread_info *)
  504. __va(memblock_alloc_base(THREAD_SIZE,
  505. THREAD_SIZE, limit));
  506. hardirq_ctx[i] = (struct thread_info *)
  507. __va(memblock_alloc_base(THREAD_SIZE,
  508. THREAD_SIZE, limit));
  509. }
  510. }
  511. #ifdef CONFIG_PPC_BOOK3E
  512. static void __init exc_lvl_early_init(void)
  513. {
  514. unsigned int i;
  515. unsigned long sp;
  516. for_each_possible_cpu(i) {
  517. sp = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
  518. critirq_ctx[i] = (struct thread_info *)__va(sp);
  519. paca[i].crit_kstack = __va(sp + THREAD_SIZE);
  520. sp = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
  521. dbgirq_ctx[i] = (struct thread_info *)__va(sp);
  522. paca[i].dbg_kstack = __va(sp + THREAD_SIZE);
  523. sp = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
  524. mcheckirq_ctx[i] = (struct thread_info *)__va(sp);
  525. paca[i].mc_kstack = __va(sp + THREAD_SIZE);
  526. }
  527. if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC))
  528. patch_exception(0x040, exc_debug_debug_book3e);
  529. }
  530. #else
  531. #define exc_lvl_early_init()
  532. #endif
  533. /*
  534. * Stack space used when we detect a bad kernel stack pointer, and
  535. * early in SMP boots before relocation is enabled. Exclusive emergency
  536. * stack for machine checks.
  537. */
  538. static void __init emergency_stack_init(void)
  539. {
  540. u64 limit;
  541. unsigned int i;
  542. /*
  543. * Emergency stacks must be under 256MB, we cannot afford to take
  544. * SLB misses on them. The ABI also requires them to be 128-byte
  545. * aligned.
  546. *
  547. * Since we use these as temporary stacks during secondary CPU
  548. * bringup, we need to get at them in real mode. This means they
  549. * must also be within the RMO region.
  550. */
  551. limit = min(safe_stack_limit(), ppc64_rma_size);
  552. for_each_possible_cpu(i) {
  553. unsigned long sp;
  554. sp = memblock_alloc_base(THREAD_SIZE, THREAD_SIZE, limit);
  555. sp += THREAD_SIZE;
  556. paca[i].emergency_sp = __va(sp);
  557. #ifdef CONFIG_PPC_BOOK3S_64
  558. /* emergency stack for machine check exception handling. */
  559. sp = memblock_alloc_base(THREAD_SIZE, THREAD_SIZE, limit);
  560. sp += THREAD_SIZE;
  561. paca[i].mc_emergency_sp = __va(sp);
  562. #endif
  563. }
  564. }
  565. /*
  566. * Called into from start_kernel this initializes memblock, which is used
  567. * to manage page allocation until mem_init is called.
  568. */
  569. void __init setup_arch(char **cmdline_p)
  570. {
  571. *cmdline_p = boot_command_line;
  572. /*
  573. * Set cache line size based on type of cpu as a default.
  574. * Systems with OF can look in the properties on the cpu node(s)
  575. * for a possibly more accurate value.
  576. */
  577. dcache_bsize = ppc64_caches.dline_size;
  578. icache_bsize = ppc64_caches.iline_size;
  579. if (ppc_md.panic)
  580. setup_panic();
  581. init_mm.start_code = (unsigned long)_stext;
  582. init_mm.end_code = (unsigned long) _etext;
  583. init_mm.end_data = (unsigned long) _edata;
  584. init_mm.brk = klimit;
  585. #ifdef CONFIG_PPC_64K_PAGES
  586. init_mm.context.pte_frag = NULL;
  587. #endif
  588. #ifdef CONFIG_SPAPR_TCE_IOMMU
  589. mm_iommu_init(&init_mm.context);
  590. #endif
  591. irqstack_early_init();
  592. exc_lvl_early_init();
  593. emergency_stack_init();
  594. initmem_init();
  595. #ifdef CONFIG_DUMMY_CONSOLE
  596. conswitchp = &dummy_con;
  597. #endif
  598. if (ppc_md.setup_arch)
  599. ppc_md.setup_arch();
  600. paging_init();
  601. /* Initialize the MMU context management stuff */
  602. mmu_context_init();
  603. /* Interrupt code needs to be 64K-aligned */
  604. if ((unsigned long)_stext & 0xffff)
  605. panic("Kernelbase not 64K-aligned (0x%lx)!\n",
  606. (unsigned long)_stext);
  607. }
  608. #ifdef CONFIG_SMP
  609. #define PCPU_DYN_SIZE ()
  610. static void * __init pcpu_fc_alloc(unsigned int cpu, size_t size, size_t align)
  611. {
  612. return __alloc_bootmem_node(NODE_DATA(cpu_to_node(cpu)), size, align,
  613. __pa(MAX_DMA_ADDRESS));
  614. }
  615. static void __init pcpu_fc_free(void *ptr, size_t size)
  616. {
  617. free_bootmem(__pa(ptr), size);
  618. }
  619. static int pcpu_cpu_distance(unsigned int from, unsigned int to)
  620. {
  621. if (cpu_to_node(from) == cpu_to_node(to))
  622. return LOCAL_DISTANCE;
  623. else
  624. return REMOTE_DISTANCE;
  625. }
  626. unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
  627. EXPORT_SYMBOL(__per_cpu_offset);
  628. void __init setup_per_cpu_areas(void)
  629. {
  630. const size_t dyn_size = PERCPU_MODULE_RESERVE + PERCPU_DYNAMIC_RESERVE;
  631. size_t atom_size;
  632. unsigned long delta;
  633. unsigned int cpu;
  634. int rc;
  635. /*
  636. * Linear mapping is one of 4K, 1M and 16M. For 4K, no need
  637. * to group units. For larger mappings, use 1M atom which
  638. * should be large enough to contain a number of units.
  639. */
  640. if (mmu_linear_psize == MMU_PAGE_4K)
  641. atom_size = PAGE_SIZE;
  642. else
  643. atom_size = 1 << 20;
  644. rc = pcpu_embed_first_chunk(0, dyn_size, atom_size, pcpu_cpu_distance,
  645. pcpu_fc_alloc, pcpu_fc_free);
  646. if (rc < 0)
  647. panic("cannot initialize percpu area (err=%d)", rc);
  648. delta = (unsigned long)pcpu_base_addr - (unsigned long)__per_cpu_start;
  649. for_each_possible_cpu(cpu) {
  650. __per_cpu_offset[cpu] = delta + pcpu_unit_offsets[cpu];
  651. paca[cpu].data_offset = __per_cpu_offset[cpu];
  652. }
  653. }
  654. #endif
  655. #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
  656. unsigned long memory_block_size_bytes(void)
  657. {
  658. if (ppc_md.memory_block_size)
  659. return ppc_md.memory_block_size();
  660. return MIN_MEMORY_BLOCK_SIZE;
  661. }
  662. #endif
  663. #if defined(CONFIG_PPC_INDIRECT_PIO) || defined(CONFIG_PPC_INDIRECT_MMIO)
  664. struct ppc_pci_io ppc_pci_io;
  665. EXPORT_SYMBOL(ppc_pci_io);
  666. #endif
  667. #ifdef CONFIG_HARDLOCKUP_DETECTOR
  668. u64 hw_nmi_get_sample_period(int watchdog_thresh)
  669. {
  670. return ppc_proc_freq * watchdog_thresh;
  671. }
  672. /*
  673. * The hardlockup detector breaks PMU event based branches and is likely
  674. * to get false positives in KVM guests, so disable it by default.
  675. */
  676. static int __init disable_hardlockup_detector(void)
  677. {
  678. hardlockup_detector_disable();
  679. return 0;
  680. }
  681. early_initcall(disable_hardlockup_detector);
  682. #endif