crypto: fix broken crypto_register_instance() module handling
[linux-2.6-block.git] / arch / powerpc / kernel / setup_64.c
CommitLineData
40ef8cbc
PM
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
7191b615 13#define DEBUG
40ef8cbc 14
4b16f8e2 15#include <linux/export.h>
40ef8cbc
PM
16#include <linux/string.h>
17#include <linux/sched.h>
18#include <linux/init.h>
19#include <linux/kernel.h>
20#include <linux/reboot.h>
21#include <linux/delay.h>
22#include <linux/initrd.h>
40ef8cbc
PM
23#include <linux/seq_file.h>
24#include <linux/ioport.h>
25#include <linux/console.h>
26#include <linux/utsname.h>
27#include <linux/tty.h>
28#include <linux/root_dev.h>
29#include <linux/notifier.h>
30#include <linux/cpu.h>
31#include <linux/unistd.h>
32#include <linux/serial.h>
33#include <linux/serial_8250.h>
7a0268fa 34#include <linux/bootmem.h>
12d04eef 35#include <linux/pci.h>
945feb17 36#include <linux/lockdep.h>
95f72d1e 37#include <linux/memblock.h>
a6146888 38#include <linux/hugetlb.h>
a5d86257 39#include <linux/memory.h>
a6146888 40
40ef8cbc 41#include <asm/io.h>
0cc4746c 42#include <asm/kdump.h>
40ef8cbc
PM
43#include <asm/prom.h>
44#include <asm/processor.h>
45#include <asm/pgtable.h>
40ef8cbc
PM
46#include <asm/smp.h>
47#include <asm/elf.h>
48#include <asm/machdep.h>
49#include <asm/paca.h>
40ef8cbc
PM
50#include <asm/time.h>
51#include <asm/cputable.h>
52#include <asm/sections.h>
53#include <asm/btext.h>
54#include <asm/nvram.h>
55#include <asm/setup.h>
40ef8cbc
PM
56#include <asm/rtas.h>
57#include <asm/iommu.h>
58#include <asm/serial.h>
59#include <asm/cache.h>
60#include <asm/page.h>
61#include <asm/mmu.h>
40ef8cbc 62#include <asm/firmware.h>
f78541dc 63#include <asm/xmon.h>
dcad47fc 64#include <asm/udbg.h>
593e537b 65#include <asm/kexec.h>
25d21ad6 66#include <asm/mmu_context.h>
d36b4c4f 67#include <asm/code-patching.h>
aa04b4cc 68#include <asm/kvm_ppc.h>
a6146888 69#include <asm/hugetlb.h>
4e21b94c 70#include <asm/epapr_hcalls.h>
40ef8cbc
PM
71
72#ifdef DEBUG
73#define DBG(fmt...) udbg_printf(fmt)
74#else
75#define DBG(fmt...)
76#endif
77
8246aca7 78int spinning_secondaries;
40ef8cbc
PM
79u64 ppc64_pft_size;
80
dabcafd3
OJ
81/* Pick defaults since we might want to patch instructions
82 * before we've read this from the device tree.
83 */
84struct ppc64_caches ppc64_caches = {
5a2fe38d
OJ
85 .dline_size = 0x40,
86 .log_dline_size = 6,
87 .iline_size = 0x40,
88 .log_iline_size = 6
dabcafd3 89};
40ef8cbc
PM
90EXPORT_SYMBOL_GPL(ppc64_caches);
91
92/*
93 * These are used in binfmt_elf.c to put aux entries on the stack
94 * for each elf executable being started.
95 */
96int dcache_bsize;
97int icache_bsize;
98int ucache_bsize;
99
28efc35f
SW
100#if defined(CONFIG_PPC_BOOK3E) && defined(CONFIG_SMP)
101static void setup_tlb_core_data(void)
102{
103 int cpu;
104
82d86de2
SW
105 BUILD_BUG_ON(offsetof(struct tlb_core_data, lock) != 0);
106
28efc35f
SW
107 for_each_possible_cpu(cpu) {
108 int first = cpu_first_thread_sibling(cpu);
109
110 paca[cpu].tcd_ptr = &paca[first].tcd;
111
112 /*
113 * If we have threads, we need either tlbsrx.
114 * or e6500 tablewalk mode, or else TLB handlers
115 * will be racy and could produce duplicate entries.
116 */
117 if (smt_enabled_at_boot >= 2 &&
118 !mmu_has_feature(MMU_FTR_USE_TLBRSRV) &&
119 book3e_htw_mode != PPC_HTW_E6500) {
120 /* Should we panic instead? */
121 WARN_ONCE("%s: unsupported MMU configuration -- expect problems\n",
122 __func__);
123 }
124 }
125}
126#else
127static void setup_tlb_core_data(void)
128{
129}
130#endif
131
40ef8cbc
PM
132#ifdef CONFIG_SMP
133
954e6da5 134static char *smt_enabled_cmdline;
40ef8cbc
PM
135
136/* Look for ibm,smt-enabled OF option */
137static void check_smt_enabled(void)
138{
139 struct device_node *dn;
a7f67bdf 140 const char *smt_option;
40ef8cbc 141
954e6da5
NF
142 /* Default to enabling all threads */
143 smt_enabled_at_boot = threads_per_core;
40ef8cbc 144
954e6da5
NF
145 /* Allow the command line to overrule the OF option */
146 if (smt_enabled_cmdline) {
147 if (!strcmp(smt_enabled_cmdline, "on"))
148 smt_enabled_at_boot = threads_per_core;
149 else if (!strcmp(smt_enabled_cmdline, "off"))
150 smt_enabled_at_boot = 0;
151 else {
1618bd53 152 int smt;
954e6da5
NF
153 int rc;
154
1618bd53 155 rc = kstrtoint(smt_enabled_cmdline, 10, &smt);
954e6da5
NF
156 if (!rc)
157 smt_enabled_at_boot =
1618bd53 158 min(threads_per_core, smt);
954e6da5
NF
159 }
160 } else {
161 dn = of_find_node_by_path("/options");
162 if (dn) {
163 smt_option = of_get_property(dn, "ibm,smt-enabled",
164 NULL);
165
166 if (smt_option) {
167 if (!strcmp(smt_option, "on"))
168 smt_enabled_at_boot = threads_per_core;
169 else if (!strcmp(smt_option, "off"))
170 smt_enabled_at_boot = 0;
171 }
172
173 of_node_put(dn);
174 }
175 }
40ef8cbc
PM
176}
177
178/* Look for smt-enabled= cmdline option */
179static int __init early_smt_enabled(char *p)
180{
954e6da5 181 smt_enabled_cmdline = p;
40ef8cbc
PM
182 return 0;
183}
184early_param("smt-enabled", early_smt_enabled);
185
5ad57078
PM
186#else
187#define check_smt_enabled()
40ef8cbc
PM
188#endif /* CONFIG_SMP */
189
25e13814
ME
190/** Fix up paca fields required for the boot cpu */
191static void fixup_boot_paca(void)
192{
193 /* The boot cpu is started */
194 get_paca()->cpu_start = 1;
195 /* Allow percpu accesses to work until we setup percpu data */
196 get_paca()->data_offset = 0;
197}
198
8f619b54
BH
199static void cpu_ready_for_interrupts(void)
200{
201 /* Set IR and DR in PACA MSR */
202 get_paca()->kernel_msr = MSR_KERNEL;
203
633440f1
ME
204 /*
205 * Enable AIL if supported, and we are in hypervisor mode. If we are
206 * not in hypervisor mode, we enable relocation-on interrupts later
207 * in pSeries_setup_arch() using the H_SET_MODE hcall.
208 */
18aa0da3
PM
209 if (cpu_has_feature(CPU_FTR_HVMODE) &&
210 cpu_has_feature(CPU_FTR_ARCH_207S)) {
8f619b54
BH
211 unsigned long lpcr = mfspr(SPRN_LPCR);
212 mtspr(SPRN_LPCR, lpcr | LPCR_AIL_3);
213 }
214}
215
40ef8cbc
PM
216/*
217 * Early initialization entry point. This is called by head.S
218 * with MMU translation disabled. We rely on the "feature" of
219 * the CPU that ignores the top 2 bits of the address in real
220 * mode so we can access kernel globals normally provided we
221 * only toy with things in the RMO region. From here, we do
95f72d1e 222 * some early parsing of the device-tree to setup out MEMBLOCK
40ef8cbc
PM
223 * data structures, and allocate & initialize the hash table
224 * and segment tables so we can start running with translation
225 * enabled.
226 *
227 * It is this function which will call the probe() callback of
228 * the various platform types and copy the matching one to the
229 * global ppc_md structure. Your platform can eventually do
230 * some very early initializations from the probe() routine, but
231 * this is not recommended, be very careful as, for example, the
232 * device-tree is not accessible via normal means at this point.
233 */
234
235void __init early_setup(unsigned long dt_ptr)
236{
6a7e4064
GL
237 static __initdata struct paca_struct boot_paca;
238
24d96495
BH
239 /* -------- printk is _NOT_ safe to use here ! ------- */
240
42c4aaad 241 /* Identify CPU type */
974a76f5 242 identify_cpu(0, mfspr(SPRN_PVR));
42c4aaad 243
33dbcf72 244 /* Assume we're on cpu 0 for now. Don't write to the paca yet! */
1426d5a3
ME
245 initialise_paca(&boot_paca, 0);
246 setup_paca(&boot_paca);
25e13814 247 fixup_boot_paca();
33dbcf72 248
945feb17
BH
249 /* Initialize lockdep early or else spinlocks will blow */
250 lockdep_init();
251
24d96495
BH
252 /* -------- printk is now safe to use ------- */
253
f2fd2513
BH
254 /* Enable early debugging if any specified (see udbg.h) */
255 udbg_early_init();
256
e8222502 257 DBG(" -> early_setup(), dt_ptr: 0x%lx\n", dt_ptr);
40ef8cbc 258
40ef8cbc 259 /*
3c607ce2
LV
260 * Do early initialization using the flattened device
261 * tree, such as retrieving the physical memory map or
262 * calculating/retrieving the hash table size.
40ef8cbc
PM
263 */
264 early_init_devtree(__va(dt_ptr));
265
4e21b94c
LT
266 epapr_paravirt_early_init();
267
4df20460 268 /* Now we know the logical id of our boot cpu, setup the paca. */
1426d5a3 269 setup_paca(&paca[boot_cpuid]);
25e13814 270 fixup_boot_paca();
4df20460 271
e8222502
BH
272 /* Probe the machine type */
273 probe_machine();
40ef8cbc 274
47310413 275 setup_kdump_trampoline();
0cc4746c 276
40ef8cbc
PM
277 DBG("Found, Initializing memory management...\n");
278
757c74d2
BH
279 /* Initialize the hash table or TLB handling */
280 early_init_mmu();
40ef8cbc 281
a944a9c4
BH
282 /*
283 * At this point, we can let interrupts switch to virtual mode
284 * (the MMU has been setup), so adjust the MSR in the PACA to
8f619b54 285 * have IR and DR set and enable AIL if it exists
a944a9c4 286 */
8f619b54 287 cpu_ready_for_interrupts();
a944a9c4
BH
288
289 /* Reserve large chunks of memory for use by CMA for KVM */
fa61a4e3
AK
290 kvm_cma_reserve();
291
a6146888
BB
292 /*
293 * Reserve any gigantic pages requested on the command line.
294 * memblock needs to have been initialized by the time this is
295 * called since this will reserve memory.
296 */
297 reserve_hugetlb_gpages();
298
40ef8cbc 299 DBG(" <- early_setup()\n");
7191b615
BH
300
301#ifdef CONFIG_PPC_EARLY_DEBUG_BOOTX
302 /*
303 * This needs to be done *last* (after the above DBG() even)
304 *
305 * Right after we return from this function, we turn on the MMU
306 * which means the real-mode access trick that btext does will
307 * no longer work, it needs to switch to using a real MMU
308 * mapping. This call will ensure that it does
309 */
310 btext_map();
311#endif /* CONFIG_PPC_EARLY_DEBUG_BOOTX */
40ef8cbc
PM
312}
313
799d6046
PM
314#ifdef CONFIG_SMP
315void early_setup_secondary(void)
316{
d04c56f7 317 /* Mark interrupts enabled in PACA */
757c74d2 318 get_paca()->soft_enabled = 0;
799d6046 319
757c74d2
BH
320 /* Initialize the hash table or TLB handling */
321 early_init_mmu_secondary();
a944a9c4
BH
322
323 /*
324 * At this point, we can let interrupts switch to virtual mode
325 * (the MMU has been setup), so adjust the MSR in the PACA to
326 * have IR and DR set.
327 */
8f619b54 328 cpu_ready_for_interrupts();
799d6046
PM
329}
330
331#endif /* CONFIG_SMP */
40ef8cbc 332
b8f51021
ME
333#if defined(CONFIG_SMP) || defined(CONFIG_KEXEC)
334void smp_release_cpus(void)
335{
758438a7 336 unsigned long *ptr;
9d07bc84 337 int i;
b8f51021
ME
338
339 DBG(" -> smp_release_cpus()\n");
340
341 /* All secondary cpus are spinning on a common spinloop, release them
342 * all now so they can start to spin on their individual paca
343 * spinloops. For non SMP kernels, the secondary cpus never get out
344 * of the common spinloop.
1f6a93e4 345 */
b8f51021 346
758438a7
ME
347 ptr = (unsigned long *)((unsigned long)&__secondary_hold_spinloop
348 - PHYSICAL_START);
2751b628 349 *ptr = ppc_function_entry(generic_secondary_smp_init);
9d07bc84
BH
350
351 /* And wait a bit for them to catch up */
352 for (i = 0; i < 100000; i++) {
353 mb();
354 HMT_low();
7ac87abb 355 if (spinning_secondaries == 0)
9d07bc84
BH
356 break;
357 udelay(1);
358 }
7ac87abb 359 DBG("spinning_secondaries = %d\n", spinning_secondaries);
b8f51021
ME
360
361 DBG(" <- smp_release_cpus()\n");
362}
363#endif /* CONFIG_SMP || CONFIG_KEXEC */
364
40ef8cbc 365/*
799d6046
PM
366 * Initialize some remaining members of the ppc64_caches and systemcfg
367 * structures
40ef8cbc
PM
368 * (at least until we get rid of them completely). This is mostly some
369 * cache informations about the CPU that will be used by cache flush
370 * routines and/or provided to userland
371 */
372static void __init initialize_cache_info(void)
373{
374 struct device_node *np;
375 unsigned long num_cpus = 0;
376
377 DBG(" -> initialize_cache_info()\n");
378
94db7c5e 379 for_each_node_by_type(np, "cpu") {
40ef8cbc
PM
380 num_cpus += 1;
381
dfbe93a2
AB
382 /*
383 * We're assuming *all* of the CPUs have the same
40ef8cbc
PM
384 * d-cache and i-cache sizes... -Peter
385 */
dfbe93a2 386 if (num_cpus == 1) {
7946d5a5 387 const __be32 *sizep, *lsizep;
40ef8cbc 388 u32 size, lsize;
40ef8cbc
PM
389
390 size = 0;
391 lsize = cur_cpu_spec->dcache_bsize;
e2eb6392 392 sizep = of_get_property(np, "d-cache-size", NULL);
40ef8cbc 393 if (sizep != NULL)
7946d5a5 394 size = be32_to_cpu(*sizep);
dfbe93a2
AB
395 lsizep = of_get_property(np, "d-cache-block-size",
396 NULL);
20474abd
BH
397 /* fallback if block size missing */
398 if (lsizep == NULL)
dfbe93a2
AB
399 lsizep = of_get_property(np,
400 "d-cache-line-size",
401 NULL);
40ef8cbc 402 if (lsizep != NULL)
7946d5a5 403 lsize = be32_to_cpu(*lsizep);
b0d436c7 404 if (sizep == NULL || lsizep == NULL)
40ef8cbc
PM
405 DBG("Argh, can't find dcache properties ! "
406 "sizep: %p, lsizep: %p\n", sizep, lsizep);
407
a7f290da
BH
408 ppc64_caches.dsize = size;
409 ppc64_caches.dline_size = lsize;
40ef8cbc
PM
410 ppc64_caches.log_dline_size = __ilog2(lsize);
411 ppc64_caches.dlines_per_page = PAGE_SIZE / lsize;
412
413 size = 0;
414 lsize = cur_cpu_spec->icache_bsize;
e2eb6392 415 sizep = of_get_property(np, "i-cache-size", NULL);
40ef8cbc 416 if (sizep != NULL)
7946d5a5 417 size = be32_to_cpu(*sizep);
dfbe93a2
AB
418 lsizep = of_get_property(np, "i-cache-block-size",
419 NULL);
20474abd 420 if (lsizep == NULL)
dfbe93a2
AB
421 lsizep = of_get_property(np,
422 "i-cache-line-size",
423 NULL);
40ef8cbc 424 if (lsizep != NULL)
7946d5a5 425 lsize = be32_to_cpu(*lsizep);
b0d436c7 426 if (sizep == NULL || lsizep == NULL)
40ef8cbc
PM
427 DBG("Argh, can't find icache properties ! "
428 "sizep: %p, lsizep: %p\n", sizep, lsizep);
429
a7f290da
BH
430 ppc64_caches.isize = size;
431 ppc64_caches.iline_size = lsize;
40ef8cbc
PM
432 ppc64_caches.log_iline_size = __ilog2(lsize);
433 ppc64_caches.ilines_per_page = PAGE_SIZE / lsize;
434 }
435 }
436
40ef8cbc
PM
437 DBG(" <- initialize_cache_info()\n");
438}
439
40ef8cbc
PM
440
441/*
442 * Do some initial setup of the system. The parameters are those which
443 * were passed in from the bootloader.
444 */
445void __init setup_system(void)
446{
447 DBG(" -> setup_system()\n");
448
826ea8f2
TB
449 /* Apply the CPUs-specific and firmware specific fixups to kernel
450 * text (nop out sections not relevant to this CPU or this firmware)
42c4aaad 451 */
0909c8c2 452 do_feature_fixups(cur_cpu_spec->cpu_features,
42c4aaad 453 &__start___ftr_fixup, &__stop___ftr_fixup);
7c03d653
BH
454 do_feature_fixups(cur_cpu_spec->mmu_features,
455 &__start___mmu_ftr_fixup, &__stop___mmu_ftr_fixup);
826ea8f2
TB
456 do_feature_fixups(powerpc_firmware_features,
457 &__start___fw_ftr_fixup, &__stop___fw_ftr_fixup);
2d1b2027
KG
458 do_lwsync_fixups(cur_cpu_spec->cpu_features,
459 &__start___lwsync_fixup, &__stop___lwsync_fixup);
d715e433 460 do_final_fixups();
42c4aaad 461
40ef8cbc
PM
462 /*
463 * Unflatten the device-tree passed by prom_init or kexec
464 */
465 unflatten_device_tree();
466
467 /*
468 * Fill the ppc64_caches & systemcfg structures with informations
0ebfff14 469 * retrieved from the device-tree.
40ef8cbc
PM
470 */
471 initialize_cache_info();
472
473#ifdef CONFIG_PPC_RTAS
474 /*
475 * Initialize RTAS if available
476 */
477 rtas_initialize();
478#endif /* CONFIG_PPC_RTAS */
40ef8cbc
PM
479
480 /*
481 * Check if we have an initrd provided via the device-tree
482 */
483 check_for_initrd();
40ef8cbc
PM
484
485 /*
486 * Do some platform specific early initializations, that includes
487 * setting up the hash table pointers. It also sets up some interrupt-mapping
488 * related options that will be used by finish_device_tree()
489 */
57744ea9
GL
490 if (ppc_md.init_early)
491 ppc_md.init_early();
40ef8cbc 492
463ce0e1
BH
493 /*
494 * We can discover serial ports now since the above did setup the
495 * hash table management for us, thus ioremap works. We do that early
496 * so that further code can be debugged
497 */
463ce0e1 498 find_legacy_serial_ports();
463ce0e1 499
40ef8cbc
PM
500 /*
501 * Register early console
502 */
503 register_early_udbg_console();
40ef8cbc 504
47679283
ME
505 /*
506 * Initialize xmon
507 */
508 xmon_setup();
480f6f35 509
5ad57078 510 smp_setup_cpu_maps();
954e6da5 511 check_smt_enabled();
28efc35f 512 setup_tlb_core_data();
40ef8cbc 513
e16c8765
AF
514 /*
515 * Freescale Book3e parts spin in a loop provided by firmware,
516 * so smp_release_cpus() does nothing for them
517 */
518#if defined(CONFIG_SMP) && !defined(CONFIG_PPC_FSL_BOOK3E)
40ef8cbc
PM
519 /* Release secondary cpus out of their spinloops at 0x60 now that
520 * we can map physical -> logical CPU ids
521 */
522 smp_release_cpus();
f018b36f 523#endif
40ef8cbc 524
2c186e05 525 pr_info("Starting Linux PPC64 %s\n", init_utsname()->version);
40ef8cbc 526
2c186e05
AB
527 pr_info("-----------------------------------------------------\n");
528 pr_info("ppc64_pft_size = 0x%llx\n", ppc64_pft_size);
529 pr_info("phys_mem_size = 0x%llx\n", memblock_phys_mem_size());
bdce97e9 530
9697add0 531 if (ppc64_caches.dline_size != 0x80)
2c186e05 532 pr_info("dcache_line_size = 0x%x\n", ppc64_caches.dline_size);
9697add0 533 if (ppc64_caches.iline_size != 0x80)
2c186e05 534 pr_info("icache_line_size = 0x%x\n", ppc64_caches.iline_size);
bdce97e9 535
2c186e05
AB
536 pr_info("cpu_features = 0x%016lx\n", cur_cpu_spec->cpu_features);
537 pr_info(" possible = 0x%016lx\n", CPU_FTRS_POSSIBLE);
538 pr_info(" always = 0x%016lx\n", CPU_FTRS_ALWAYS);
539 pr_info("cpu_user_features = 0x%08x 0x%08x\n", cur_cpu_spec->cpu_user_features,
87d99c0e 540 cur_cpu_spec->cpu_user_features2);
2c186e05
AB
541 pr_info("mmu_features = 0x%08x\n", cur_cpu_spec->mmu_features);
542 pr_info("firmware_features = 0x%016lx\n", powerpc_firmware_features);
87d99c0e 543
94491685 544#ifdef CONFIG_PPC_STD_MMU_64
9697add0 545 if (htab_address)
2c186e05 546 pr_info("htab_address = 0x%p\n", htab_address);
bdce97e9 547
2c186e05 548 pr_info("htab_hash_mask = 0x%lx\n", htab_hash_mask);
bdce97e9
ME
549#endif
550
b160544c 551 if (PHYSICAL_START > 0)
2c186e05 552 pr_info("physical_start = 0x%llx\n",
e468455e 553 (unsigned long long)PHYSICAL_START);
2c186e05 554 pr_info("-----------------------------------------------------\n");
40ef8cbc 555
40ef8cbc
PM
556 DBG(" <- setup_system()\n");
557}
558
40bd587a
BH
559/* This returns the limit below which memory accesses to the linear
560 * mapping are guarnateed not to cause a TLB or SLB miss. This is
561 * used to allocate interrupt or emergency stacks for which our
562 * exception entry path doesn't deal with being interrupted.
563 */
564static u64 safe_stack_limit(void)
095c7965 565{
40bd587a
BH
566#ifdef CONFIG_PPC_BOOK3E
567 /* Freescale BookE bolts the entire linear mapping */
568 if (mmu_has_feature(MMU_FTR_TYPE_FSL_E))
569 return linear_map_top;
570 /* Other BookE, we assume the first GB is bolted */
571 return 1ul << 30;
572#else
573 /* BookS, the first segment is bolted */
574 if (mmu_has_feature(MMU_FTR_1T_SEGMENT))
095c7965 575 return 1UL << SID_SHIFT_1T;
095c7965 576 return 1UL << SID_SHIFT;
40bd587a 577#endif
095c7965
AB
578}
579
40ef8cbc
PM
580static void __init irqstack_early_init(void)
581{
40bd587a 582 u64 limit = safe_stack_limit();
40ef8cbc
PM
583 unsigned int i;
584
585 /*
8f4da26e
AB
586 * Interrupt stacks must be in the first segment since we
587 * cannot afford to take SLB misses on them.
40ef8cbc 588 */
0e551954 589 for_each_possible_cpu(i) {
3c726f8d 590 softirq_ctx[i] = (struct thread_info *)
95f72d1e 591 __va(memblock_alloc_base(THREAD_SIZE,
095c7965 592 THREAD_SIZE, limit));
3c726f8d 593 hardirq_ctx[i] = (struct thread_info *)
95f72d1e 594 __va(memblock_alloc_base(THREAD_SIZE,
095c7965 595 THREAD_SIZE, limit));
40ef8cbc
PM
596 }
597}
40ef8cbc 598
2d27cfd3
BH
599#ifdef CONFIG_PPC_BOOK3E
600static void __init exc_lvl_early_init(void)
601{
602 unsigned int i;
160c7324 603 unsigned long sp;
2d27cfd3
BH
604
605 for_each_possible_cpu(i) {
160c7324
TC
606 sp = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
607 critirq_ctx[i] = (struct thread_info *)__va(sp);
608 paca[i].crit_kstack = __va(sp + THREAD_SIZE);
609
610 sp = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
611 dbgirq_ctx[i] = (struct thread_info *)__va(sp);
612 paca[i].dbg_kstack = __va(sp + THREAD_SIZE);
613
614 sp = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
615 mcheckirq_ctx[i] = (struct thread_info *)__va(sp);
616 paca[i].mc_kstack = __va(sp + THREAD_SIZE);
2d27cfd3 617 }
d36b4c4f
KG
618
619 if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC))
565c2f24 620 patch_exception(0x040, exc_debug_debug_book3e);
2d27cfd3
BH
621}
622#else
623#define exc_lvl_early_init()
624#endif
625
40ef8cbc
PM
626/*
627 * Stack space used when we detect a bad kernel stack pointer, and
729b0f71
MS
628 * early in SMP boots before relocation is enabled. Exclusive emergency
629 * stack for machine checks.
40ef8cbc
PM
630 */
631static void __init emergency_stack_init(void)
632{
095c7965 633 u64 limit;
40ef8cbc
PM
634 unsigned int i;
635
636 /*
637 * Emergency stacks must be under 256MB, we cannot afford to take
638 * SLB misses on them. The ABI also requires them to be 128-byte
639 * aligned.
640 *
641 * Since we use these as temporary stacks during secondary CPU
642 * bringup, we need to get at them in real mode. This means they
643 * must also be within the RMO region.
644 */
40bd587a 645 limit = min(safe_stack_limit(), ppc64_rma_size);
40ef8cbc 646
3243d874
ME
647 for_each_possible_cpu(i) {
648 unsigned long sp;
95f72d1e 649 sp = memblock_alloc_base(THREAD_SIZE, THREAD_SIZE, limit);
3243d874
ME
650 sp += THREAD_SIZE;
651 paca[i].emergency_sp = __va(sp);
729b0f71
MS
652
653#ifdef CONFIG_PPC_BOOK3S_64
654 /* emergency stack for machine check exception handling. */
655 sp = memblock_alloc_base(THREAD_SIZE, THREAD_SIZE, limit);
656 sp += THREAD_SIZE;
657 paca[i].mc_emergency_sp = __va(sp);
658#endif
3243d874 659 }
40ef8cbc
PM
660}
661
40ef8cbc 662/*
e39f223f 663 * Called into from start_kernel this initializes memblock, which is used
0f6b77ca 664 * to manage page allocation until mem_init is called.
40ef8cbc
PM
665 */
666void __init setup_arch(char **cmdline_p)
667{
3e47d147 668 *cmdline_p = boot_command_line;
40ef8cbc
PM
669
670 /*
671 * Set cache line size based on type of cpu as a default.
672 * Systems with OF can look in the properties on the cpu node(s)
673 * for a possibly more accurate value.
674 */
675 dcache_bsize = ppc64_caches.dline_size;
676 icache_bsize = ppc64_caches.iline_size;
677
40ef8cbc 678 if (ppc_md.panic)
7e990266 679 setup_panic();
40ef8cbc 680
4846c5de 681 init_mm.start_code = (unsigned long)_stext;
40ef8cbc
PM
682 init_mm.end_code = (unsigned long) _etext;
683 init_mm.end_data = (unsigned long) _edata;
684 init_mm.brk = klimit;
5c1f6ee9
AK
685#ifdef CONFIG_PPC_64K_PAGES
686 init_mm.context.pte_frag = NULL;
687#endif
40ef8cbc 688 irqstack_early_init();
2d27cfd3 689 exc_lvl_early_init();
40ef8cbc
PM
690 emergency_stack_init();
691
10239733 692 initmem_init();
40ef8cbc 693
0458060c
PM
694#ifdef CONFIG_DUMMY_CONSOLE
695 conswitchp = &dummy_con;
696#endif
697
38db7e74
GL
698 if (ppc_md.setup_arch)
699 ppc_md.setup_arch();
40ef8cbc 700
40ef8cbc 701 paging_init();
6f0ef0f5
BH
702
703 /* Initialize the MMU context management stuff */
704 mmu_context_init();
705
61e2390e
MN
706 /* Interrupt code needs to be 64K-aligned */
707 if ((unsigned long)_stext & 0xffff)
708 panic("Kernelbase not 64K-aligned (0x%lx)!\n",
709 (unsigned long)_stext);
40ef8cbc
PM
710}
711
7a0268fa 712#ifdef CONFIG_SMP
c2a7e818
TH
713#define PCPU_DYN_SIZE ()
714
715static void * __init pcpu_fc_alloc(unsigned int cpu, size_t size, size_t align)
7a0268fa 716{
c2a7e818
TH
717 return __alloc_bootmem_node(NODE_DATA(cpu_to_node(cpu)), size, align,
718 __pa(MAX_DMA_ADDRESS));
719}
7a0268fa 720
c2a7e818
TH
721static void __init pcpu_fc_free(void *ptr, size_t size)
722{
723 free_bootmem(__pa(ptr), size);
724}
7a0268fa 725
c2a7e818
TH
726static int pcpu_cpu_distance(unsigned int from, unsigned int to)
727{
728 if (cpu_to_node(from) == cpu_to_node(to))
729 return LOCAL_DISTANCE;
730 else
731 return REMOTE_DISTANCE;
732}
733
ae01f84b
AB
734unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
735EXPORT_SYMBOL(__per_cpu_offset);
736
c2a7e818
TH
737void __init setup_per_cpu_areas(void)
738{
739 const size_t dyn_size = PERCPU_MODULE_RESERVE + PERCPU_DYNAMIC_RESERVE;
740 size_t atom_size;
741 unsigned long delta;
742 unsigned int cpu;
743 int rc;
744
745 /*
746 * Linear mapping is one of 4K, 1M and 16M. For 4K, no need
747 * to group units. For larger mappings, use 1M atom which
748 * should be large enough to contain a number of units.
749 */
750 if (mmu_linear_psize == MMU_PAGE_4K)
751 atom_size = PAGE_SIZE;
752 else
753 atom_size = 1 << 20;
754
755 rc = pcpu_embed_first_chunk(0, dyn_size, atom_size, pcpu_cpu_distance,
756 pcpu_fc_alloc, pcpu_fc_free);
757 if (rc < 0)
758 panic("cannot initialize percpu area (err=%d)", rc);
759
760 delta = (unsigned long)pcpu_base_addr - (unsigned long)__per_cpu_start;
ae01f84b
AB
761 for_each_possible_cpu(cpu) {
762 __per_cpu_offset[cpu] = delta + pcpu_unit_offsets[cpu];
763 paca[cpu].data_offset = __per_cpu_offset[cpu];
764 }
7a0268fa
AB
765}
766#endif
4cb3cee0 767
a5d86257
AB
768#ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
769unsigned long memory_block_size_bytes(void)
770{
771 if (ppc_md.memory_block_size)
772 return ppc_md.memory_block_size();
773
774 return MIN_MEMORY_BLOCK_SIZE;
775}
776#endif
4cb3cee0 777
ecd73cc5 778#if defined(CONFIG_PPC_INDIRECT_PIO) || defined(CONFIG_PPC_INDIRECT_MMIO)
4cb3cee0
BH
779struct ppc_pci_io ppc_pci_io;
780EXPORT_SYMBOL(ppc_pci_io);
ecd73cc5 781#endif