]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - arch/powerpc/kernel/setup_64.c
Merge tag 'pm+acpi-3.15-rc1-3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafae...
[mirror_ubuntu-hirsute-kernel.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
BB
38#include <linux/hugetlb.h>
39
40ef8cbc 40#include <asm/io.h>
0cc4746c 41#include <asm/kdump.h>
40ef8cbc
PM
42#include <asm/prom.h>
43#include <asm/processor.h>
44#include <asm/pgtable.h>
40ef8cbc
PM
45#include <asm/smp.h>
46#include <asm/elf.h>
47#include <asm/machdep.h>
48#include <asm/paca.h>
40ef8cbc
PM
49#include <asm/time.h>
50#include <asm/cputable.h>
51#include <asm/sections.h>
52#include <asm/btext.h>
53#include <asm/nvram.h>
54#include <asm/setup.h>
40ef8cbc
PM
55#include <asm/rtas.h>
56#include <asm/iommu.h>
57#include <asm/serial.h>
58#include <asm/cache.h>
59#include <asm/page.h>
60#include <asm/mmu.h>
40ef8cbc 61#include <asm/firmware.h>
f78541dc 62#include <asm/xmon.h>
dcad47fc 63#include <asm/udbg.h>
593e537b 64#include <asm/kexec.h>
25d21ad6 65#include <asm/mmu_context.h>
d36b4c4f 66#include <asm/code-patching.h>
aa04b4cc 67#include <asm/kvm_ppc.h>
a6146888 68#include <asm/hugetlb.h>
4e21b94c 69#include <asm/epapr_hcalls.h>
40ef8cbc
PM
70
71#ifdef DEBUG
72#define DBG(fmt...) udbg_printf(fmt)
73#else
74#define DBG(fmt...)
75#endif
76
8246aca7 77int spinning_secondaries;
40ef8cbc
PM
78u64 ppc64_pft_size;
79
dabcafd3
OJ
80/* Pick defaults since we might want to patch instructions
81 * before we've read this from the device tree.
82 */
83struct ppc64_caches ppc64_caches = {
5a2fe38d
OJ
84 .dline_size = 0x40,
85 .log_dline_size = 6,
86 .iline_size = 0x40,
87 .log_iline_size = 6
dabcafd3 88};
40ef8cbc
PM
89EXPORT_SYMBOL_GPL(ppc64_caches);
90
91/*
92 * These are used in binfmt_elf.c to put aux entries on the stack
93 * for each elf executable being started.
94 */
95int dcache_bsize;
96int icache_bsize;
97int ucache_bsize;
98
28efc35f
SW
99#if defined(CONFIG_PPC_BOOK3E) && defined(CONFIG_SMP)
100static void setup_tlb_core_data(void)
101{
102 int cpu;
103
82d86de2
SW
104 BUILD_BUG_ON(offsetof(struct tlb_core_data, lock) != 0);
105
28efc35f
SW
106 for_each_possible_cpu(cpu) {
107 int first = cpu_first_thread_sibling(cpu);
108
109 paca[cpu].tcd_ptr = &paca[first].tcd;
110
111 /*
112 * If we have threads, we need either tlbsrx.
113 * or e6500 tablewalk mode, or else TLB handlers
114 * will be racy and could produce duplicate entries.
115 */
116 if (smt_enabled_at_boot >= 2 &&
117 !mmu_has_feature(MMU_FTR_USE_TLBRSRV) &&
118 book3e_htw_mode != PPC_HTW_E6500) {
119 /* Should we panic instead? */
120 WARN_ONCE("%s: unsupported MMU configuration -- expect problems\n",
121 __func__);
122 }
123 }
124}
125#else
126static void setup_tlb_core_data(void)
127{
128}
129#endif
130
40ef8cbc
PM
131#ifdef CONFIG_SMP
132
954e6da5 133static char *smt_enabled_cmdline;
40ef8cbc
PM
134
135/* Look for ibm,smt-enabled OF option */
136static void check_smt_enabled(void)
137{
138 struct device_node *dn;
a7f67bdf 139 const char *smt_option;
40ef8cbc 140
954e6da5
NF
141 /* Default to enabling all threads */
142 smt_enabled_at_boot = threads_per_core;
40ef8cbc 143
954e6da5
NF
144 /* Allow the command line to overrule the OF option */
145 if (smt_enabled_cmdline) {
146 if (!strcmp(smt_enabled_cmdline, "on"))
147 smt_enabled_at_boot = threads_per_core;
148 else if (!strcmp(smt_enabled_cmdline, "off"))
149 smt_enabled_at_boot = 0;
150 else {
151 long smt;
152 int rc;
153
154 rc = strict_strtol(smt_enabled_cmdline, 10, &smt);
155 if (!rc)
156 smt_enabled_at_boot =
157 min(threads_per_core, (int)smt);
158 }
159 } else {
160 dn = of_find_node_by_path("/options");
161 if (dn) {
162 smt_option = of_get_property(dn, "ibm,smt-enabled",
163 NULL);
164
165 if (smt_option) {
166 if (!strcmp(smt_option, "on"))
167 smt_enabled_at_boot = threads_per_core;
168 else if (!strcmp(smt_option, "off"))
169 smt_enabled_at_boot = 0;
170 }
171
172 of_node_put(dn);
173 }
174 }
40ef8cbc
PM
175}
176
177/* Look for smt-enabled= cmdline option */
178static int __init early_smt_enabled(char *p)
179{
954e6da5 180 smt_enabled_cmdline = p;
40ef8cbc
PM
181 return 0;
182}
183early_param("smt-enabled", early_smt_enabled);
184
5ad57078
PM
185#else
186#define check_smt_enabled()
40ef8cbc
PM
187#endif /* CONFIG_SMP */
188
25e13814
ME
189/** Fix up paca fields required for the boot cpu */
190static void fixup_boot_paca(void)
191{
192 /* The boot cpu is started */
193 get_paca()->cpu_start = 1;
194 /* Allow percpu accesses to work until we setup percpu data */
195 get_paca()->data_offset = 0;
196}
197
8f619b54
BH
198static void cpu_ready_for_interrupts(void)
199{
200 /* Set IR and DR in PACA MSR */
201 get_paca()->kernel_msr = MSR_KERNEL;
202
203 /* Enable AIL if supported */
204 if (cpu_has_feature(CPU_FTR_ARCH_207S)) {
205 unsigned long lpcr = mfspr(SPRN_LPCR);
206 mtspr(SPRN_LPCR, lpcr | LPCR_AIL_3);
207 }
208}
209
40ef8cbc
PM
210/*
211 * Early initialization entry point. This is called by head.S
212 * with MMU translation disabled. We rely on the "feature" of
213 * the CPU that ignores the top 2 bits of the address in real
214 * mode so we can access kernel globals normally provided we
215 * only toy with things in the RMO region. From here, we do
95f72d1e 216 * some early parsing of the device-tree to setup out MEMBLOCK
40ef8cbc
PM
217 * data structures, and allocate & initialize the hash table
218 * and segment tables so we can start running with translation
219 * enabled.
220 *
221 * It is this function which will call the probe() callback of
222 * the various platform types and copy the matching one to the
223 * global ppc_md structure. Your platform can eventually do
224 * some very early initializations from the probe() routine, but
225 * this is not recommended, be very careful as, for example, the
226 * device-tree is not accessible via normal means at this point.
227 */
228
229void __init early_setup(unsigned long dt_ptr)
230{
6a7e4064
GL
231 static __initdata struct paca_struct boot_paca;
232
24d96495
BH
233 /* -------- printk is _NOT_ safe to use here ! ------- */
234
42c4aaad 235 /* Identify CPU type */
974a76f5 236 identify_cpu(0, mfspr(SPRN_PVR));
42c4aaad 237
33dbcf72 238 /* Assume we're on cpu 0 for now. Don't write to the paca yet! */
1426d5a3
ME
239 initialise_paca(&boot_paca, 0);
240 setup_paca(&boot_paca);
25e13814 241 fixup_boot_paca();
33dbcf72 242
945feb17
BH
243 /* Initialize lockdep early or else spinlocks will blow */
244 lockdep_init();
245
24d96495
BH
246 /* -------- printk is now safe to use ------- */
247
f2fd2513
BH
248 /* Enable early debugging if any specified (see udbg.h) */
249 udbg_early_init();
250
e8222502 251 DBG(" -> early_setup(), dt_ptr: 0x%lx\n", dt_ptr);
40ef8cbc 252
40ef8cbc 253 /*
3c607ce2
LV
254 * Do early initialization using the flattened device
255 * tree, such as retrieving the physical memory map or
256 * calculating/retrieving the hash table size.
40ef8cbc
PM
257 */
258 early_init_devtree(__va(dt_ptr));
259
4e21b94c
LT
260 epapr_paravirt_early_init();
261
4df20460 262 /* Now we know the logical id of our boot cpu, setup the paca. */
1426d5a3 263 setup_paca(&paca[boot_cpuid]);
25e13814 264 fixup_boot_paca();
4df20460 265
e8222502
BH
266 /* Probe the machine type */
267 probe_machine();
40ef8cbc 268
47310413 269 setup_kdump_trampoline();
0cc4746c 270
40ef8cbc
PM
271 DBG("Found, Initializing memory management...\n");
272
757c74d2
BH
273 /* Initialize the hash table or TLB handling */
274 early_init_mmu();
40ef8cbc 275
a944a9c4
BH
276 /*
277 * At this point, we can let interrupts switch to virtual mode
278 * (the MMU has been setup), so adjust the MSR in the PACA to
8f619b54 279 * have IR and DR set and enable AIL if it exists
a944a9c4 280 */
8f619b54 281 cpu_ready_for_interrupts();
a944a9c4
BH
282
283 /* Reserve large chunks of memory for use by CMA for KVM */
fa61a4e3
AK
284 kvm_cma_reserve();
285
a6146888
BB
286 /*
287 * Reserve any gigantic pages requested on the command line.
288 * memblock needs to have been initialized by the time this is
289 * called since this will reserve memory.
290 */
291 reserve_hugetlb_gpages();
292
40ef8cbc 293 DBG(" <- early_setup()\n");
7191b615
BH
294
295#ifdef CONFIG_PPC_EARLY_DEBUG_BOOTX
296 /*
297 * This needs to be done *last* (after the above DBG() even)
298 *
299 * Right after we return from this function, we turn on the MMU
300 * which means the real-mode access trick that btext does will
301 * no longer work, it needs to switch to using a real MMU
302 * mapping. This call will ensure that it does
303 */
304 btext_map();
305#endif /* CONFIG_PPC_EARLY_DEBUG_BOOTX */
40ef8cbc
PM
306}
307
799d6046
PM
308#ifdef CONFIG_SMP
309void early_setup_secondary(void)
310{
d04c56f7 311 /* Mark interrupts enabled in PACA */
757c74d2 312 get_paca()->soft_enabled = 0;
799d6046 313
757c74d2
BH
314 /* Initialize the hash table or TLB handling */
315 early_init_mmu_secondary();
a944a9c4
BH
316
317 /*
318 * At this point, we can let interrupts switch to virtual mode
319 * (the MMU has been setup), so adjust the MSR in the PACA to
320 * have IR and DR set.
321 */
8f619b54 322 cpu_ready_for_interrupts();
799d6046
PM
323}
324
325#endif /* CONFIG_SMP */
40ef8cbc 326
b8f51021
ME
327#if defined(CONFIG_SMP) || defined(CONFIG_KEXEC)
328void smp_release_cpus(void)
329{
758438a7 330 unsigned long *ptr;
9d07bc84 331 int i;
b8f51021
ME
332
333 DBG(" -> smp_release_cpus()\n");
334
335 /* All secondary cpus are spinning on a common spinloop, release them
336 * all now so they can start to spin on their individual paca
337 * spinloops. For non SMP kernels, the secondary cpus never get out
338 * of the common spinloop.
1f6a93e4 339 */
b8f51021 340
758438a7
ME
341 ptr = (unsigned long *)((unsigned long)&__secondary_hold_spinloop
342 - PHYSICAL_START);
1f6a93e4 343 *ptr = __pa(generic_secondary_smp_init);
9d07bc84
BH
344
345 /* And wait a bit for them to catch up */
346 for (i = 0; i < 100000; i++) {
347 mb();
348 HMT_low();
7ac87abb 349 if (spinning_secondaries == 0)
9d07bc84
BH
350 break;
351 udelay(1);
352 }
7ac87abb 353 DBG("spinning_secondaries = %d\n", spinning_secondaries);
b8f51021
ME
354
355 DBG(" <- smp_release_cpus()\n");
356}
357#endif /* CONFIG_SMP || CONFIG_KEXEC */
358
40ef8cbc 359/*
799d6046
PM
360 * Initialize some remaining members of the ppc64_caches and systemcfg
361 * structures
40ef8cbc
PM
362 * (at least until we get rid of them completely). This is mostly some
363 * cache informations about the CPU that will be used by cache flush
364 * routines and/or provided to userland
365 */
366static void __init initialize_cache_info(void)
367{
368 struct device_node *np;
369 unsigned long num_cpus = 0;
370
371 DBG(" -> initialize_cache_info()\n");
372
94db7c5e 373 for_each_node_by_type(np, "cpu") {
40ef8cbc
PM
374 num_cpus += 1;
375
dfbe93a2
AB
376 /*
377 * We're assuming *all* of the CPUs have the same
40ef8cbc
PM
378 * d-cache and i-cache sizes... -Peter
379 */
dfbe93a2 380 if (num_cpus == 1) {
7946d5a5 381 const __be32 *sizep, *lsizep;
40ef8cbc 382 u32 size, lsize;
40ef8cbc
PM
383
384 size = 0;
385 lsize = cur_cpu_spec->dcache_bsize;
e2eb6392 386 sizep = of_get_property(np, "d-cache-size", NULL);
40ef8cbc 387 if (sizep != NULL)
7946d5a5 388 size = be32_to_cpu(*sizep);
dfbe93a2
AB
389 lsizep = of_get_property(np, "d-cache-block-size",
390 NULL);
20474abd
BH
391 /* fallback if block size missing */
392 if (lsizep == NULL)
dfbe93a2
AB
393 lsizep = of_get_property(np,
394 "d-cache-line-size",
395 NULL);
40ef8cbc 396 if (lsizep != NULL)
7946d5a5 397 lsize = be32_to_cpu(*lsizep);
b0d436c7 398 if (sizep == NULL || lsizep == NULL)
40ef8cbc
PM
399 DBG("Argh, can't find dcache properties ! "
400 "sizep: %p, lsizep: %p\n", sizep, lsizep);
401
a7f290da
BH
402 ppc64_caches.dsize = size;
403 ppc64_caches.dline_size = lsize;
40ef8cbc
PM
404 ppc64_caches.log_dline_size = __ilog2(lsize);
405 ppc64_caches.dlines_per_page = PAGE_SIZE / lsize;
406
407 size = 0;
408 lsize = cur_cpu_spec->icache_bsize;
e2eb6392 409 sizep = of_get_property(np, "i-cache-size", NULL);
40ef8cbc 410 if (sizep != NULL)
7946d5a5 411 size = be32_to_cpu(*sizep);
dfbe93a2
AB
412 lsizep = of_get_property(np, "i-cache-block-size",
413 NULL);
20474abd 414 if (lsizep == NULL)
dfbe93a2
AB
415 lsizep = of_get_property(np,
416 "i-cache-line-size",
417 NULL);
40ef8cbc 418 if (lsizep != NULL)
7946d5a5 419 lsize = be32_to_cpu(*lsizep);
b0d436c7 420 if (sizep == NULL || lsizep == NULL)
40ef8cbc
PM
421 DBG("Argh, can't find icache properties ! "
422 "sizep: %p, lsizep: %p\n", sizep, lsizep);
423
a7f290da
BH
424 ppc64_caches.isize = size;
425 ppc64_caches.iline_size = lsize;
40ef8cbc
PM
426 ppc64_caches.log_iline_size = __ilog2(lsize);
427 ppc64_caches.ilines_per_page = PAGE_SIZE / lsize;
428 }
429 }
430
40ef8cbc
PM
431 DBG(" <- initialize_cache_info()\n");
432}
433
40ef8cbc
PM
434
435/*
436 * Do some initial setup of the system. The parameters are those which
437 * were passed in from the bootloader.
438 */
439void __init setup_system(void)
440{
441 DBG(" -> setup_system()\n");
442
826ea8f2
TB
443 /* Apply the CPUs-specific and firmware specific fixups to kernel
444 * text (nop out sections not relevant to this CPU or this firmware)
42c4aaad 445 */
0909c8c2 446 do_feature_fixups(cur_cpu_spec->cpu_features,
42c4aaad 447 &__start___ftr_fixup, &__stop___ftr_fixup);
7c03d653
BH
448 do_feature_fixups(cur_cpu_spec->mmu_features,
449 &__start___mmu_ftr_fixup, &__stop___mmu_ftr_fixup);
826ea8f2
TB
450 do_feature_fixups(powerpc_firmware_features,
451 &__start___fw_ftr_fixup, &__stop___fw_ftr_fixup);
2d1b2027
KG
452 do_lwsync_fixups(cur_cpu_spec->cpu_features,
453 &__start___lwsync_fixup, &__stop___lwsync_fixup);
d715e433 454 do_final_fixups();
42c4aaad 455
40ef8cbc
PM
456 /*
457 * Unflatten the device-tree passed by prom_init or kexec
458 */
459 unflatten_device_tree();
460
461 /*
462 * Fill the ppc64_caches & systemcfg structures with informations
0ebfff14 463 * retrieved from the device-tree.
40ef8cbc
PM
464 */
465 initialize_cache_info();
466
467#ifdef CONFIG_PPC_RTAS
468 /*
469 * Initialize RTAS if available
470 */
471 rtas_initialize();
472#endif /* CONFIG_PPC_RTAS */
40ef8cbc
PM
473
474 /*
475 * Check if we have an initrd provided via the device-tree
476 */
477 check_for_initrd();
40ef8cbc
PM
478
479 /*
480 * Do some platform specific early initializations, that includes
481 * setting up the hash table pointers. It also sets up some interrupt-mapping
482 * related options that will be used by finish_device_tree()
483 */
57744ea9
GL
484 if (ppc_md.init_early)
485 ppc_md.init_early();
40ef8cbc 486
463ce0e1
BH
487 /*
488 * We can discover serial ports now since the above did setup the
489 * hash table management for us, thus ioremap works. We do that early
490 * so that further code can be debugged
491 */
463ce0e1 492 find_legacy_serial_ports();
463ce0e1 493
40ef8cbc
PM
494 /*
495 * Register early console
496 */
497 register_early_udbg_console();
40ef8cbc 498
47679283
ME
499 /*
500 * Initialize xmon
501 */
502 xmon_setup();
480f6f35 503
5ad57078 504 smp_setup_cpu_maps();
954e6da5 505 check_smt_enabled();
28efc35f 506 setup_tlb_core_data();
40ef8cbc 507
f018b36f 508#ifdef CONFIG_SMP
40ef8cbc
PM
509 /* Release secondary cpus out of their spinloops at 0x60 now that
510 * we can map physical -> logical CPU ids
511 */
512 smp_release_cpus();
f018b36f 513#endif
40ef8cbc 514
96b644bd 515 printk("Starting Linux PPC64 %s\n", init_utsname()->version);
40ef8cbc
PM
516
517 printk("-----------------------------------------------------\n");
fe333321 518 printk("ppc64_pft_size = 0x%llx\n", ppc64_pft_size);
95f72d1e 519 printk("physicalMemorySize = 0x%llx\n", memblock_phys_mem_size());
9697add0
AB
520 if (ppc64_caches.dline_size != 0x80)
521 printk("ppc64_caches.dcache_line_size = 0x%x\n",
522 ppc64_caches.dline_size);
523 if (ppc64_caches.iline_size != 0x80)
524 printk("ppc64_caches.icache_line_size = 0x%x\n",
525 ppc64_caches.iline_size);
94491685 526#ifdef CONFIG_PPC_STD_MMU_64
9697add0
AB
527 if (htab_address)
528 printk("htab_address = 0x%p\n", htab_address);
40ef8cbc 529 printk("htab_hash_mask = 0x%lx\n", htab_hash_mask);
94491685 530#endif /* CONFIG_PPC_STD_MMU_64 */
b160544c 531 if (PHYSICAL_START > 0)
e468455e
ME
532 printk("physical_start = 0x%llx\n",
533 (unsigned long long)PHYSICAL_START);
40ef8cbc 534 printk("-----------------------------------------------------\n");
40ef8cbc 535
40ef8cbc
PM
536 DBG(" <- setup_system()\n");
537}
538
40bd587a
BH
539/* This returns the limit below which memory accesses to the linear
540 * mapping are guarnateed not to cause a TLB or SLB miss. This is
541 * used to allocate interrupt or emergency stacks for which our
542 * exception entry path doesn't deal with being interrupted.
543 */
544static u64 safe_stack_limit(void)
095c7965 545{
40bd587a
BH
546#ifdef CONFIG_PPC_BOOK3E
547 /* Freescale BookE bolts the entire linear mapping */
548 if (mmu_has_feature(MMU_FTR_TYPE_FSL_E))
549 return linear_map_top;
550 /* Other BookE, we assume the first GB is bolted */
551 return 1ul << 30;
552#else
553 /* BookS, the first segment is bolted */
554 if (mmu_has_feature(MMU_FTR_1T_SEGMENT))
095c7965 555 return 1UL << SID_SHIFT_1T;
095c7965 556 return 1UL << SID_SHIFT;
40bd587a 557#endif
095c7965
AB
558}
559
40ef8cbc
PM
560static void __init irqstack_early_init(void)
561{
40bd587a 562 u64 limit = safe_stack_limit();
40ef8cbc
PM
563 unsigned int i;
564
565 /*
8f4da26e
AB
566 * Interrupt stacks must be in the first segment since we
567 * cannot afford to take SLB misses on them.
40ef8cbc 568 */
0e551954 569 for_each_possible_cpu(i) {
3c726f8d 570 softirq_ctx[i] = (struct thread_info *)
95f72d1e 571 __va(memblock_alloc_base(THREAD_SIZE,
095c7965 572 THREAD_SIZE, limit));
3c726f8d 573 hardirq_ctx[i] = (struct thread_info *)
95f72d1e 574 __va(memblock_alloc_base(THREAD_SIZE,
095c7965 575 THREAD_SIZE, limit));
40ef8cbc
PM
576 }
577}
40ef8cbc 578
2d27cfd3
BH
579#ifdef CONFIG_PPC_BOOK3E
580static void __init exc_lvl_early_init(void)
581{
582 unsigned int i;
160c7324 583 unsigned long sp;
2d27cfd3
BH
584
585 for_each_possible_cpu(i) {
160c7324
TC
586 sp = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
587 critirq_ctx[i] = (struct thread_info *)__va(sp);
588 paca[i].crit_kstack = __va(sp + THREAD_SIZE);
589
590 sp = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
591 dbgirq_ctx[i] = (struct thread_info *)__va(sp);
592 paca[i].dbg_kstack = __va(sp + THREAD_SIZE);
593
594 sp = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
595 mcheckirq_ctx[i] = (struct thread_info *)__va(sp);
596 paca[i].mc_kstack = __va(sp + THREAD_SIZE);
2d27cfd3 597 }
d36b4c4f
KG
598
599 if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC))
565c2f24 600 patch_exception(0x040, exc_debug_debug_book3e);
2d27cfd3
BH
601}
602#else
603#define exc_lvl_early_init()
604#endif
605
40ef8cbc
PM
606/*
607 * Stack space used when we detect a bad kernel stack pointer, and
729b0f71
MS
608 * early in SMP boots before relocation is enabled. Exclusive emergency
609 * stack for machine checks.
40ef8cbc
PM
610 */
611static void __init emergency_stack_init(void)
612{
095c7965 613 u64 limit;
40ef8cbc
PM
614 unsigned int i;
615
616 /*
617 * Emergency stacks must be under 256MB, we cannot afford to take
618 * SLB misses on them. The ABI also requires them to be 128-byte
619 * aligned.
620 *
621 * Since we use these as temporary stacks during secondary CPU
622 * bringup, we need to get at them in real mode. This means they
623 * must also be within the RMO region.
624 */
40bd587a 625 limit = min(safe_stack_limit(), ppc64_rma_size);
40ef8cbc 626
3243d874
ME
627 for_each_possible_cpu(i) {
628 unsigned long sp;
95f72d1e 629 sp = memblock_alloc_base(THREAD_SIZE, THREAD_SIZE, limit);
3243d874
ME
630 sp += THREAD_SIZE;
631 paca[i].emergency_sp = __va(sp);
729b0f71
MS
632
633#ifdef CONFIG_PPC_BOOK3S_64
634 /* emergency stack for machine check exception handling. */
635 sp = memblock_alloc_base(THREAD_SIZE, THREAD_SIZE, limit);
636 sp += THREAD_SIZE;
637 paca[i].mc_emergency_sp = __va(sp);
638#endif
3243d874 639 }
40ef8cbc
PM
640}
641
40ef8cbc 642/*
0f6b77ca
AIB
643 * Called into from start_kernel this initializes bootmem, which is used
644 * to manage page allocation until mem_init is called.
40ef8cbc
PM
645 */
646void __init setup_arch(char **cmdline_p)
647{
40ef8cbc
PM
648 ppc64_boot_msg(0x12, "Setup Arch");
649
650 *cmdline_p = cmd_line;
651
652 /*
653 * Set cache line size based on type of cpu as a default.
654 * Systems with OF can look in the properties on the cpu node(s)
655 * for a possibly more accurate value.
656 */
657 dcache_bsize = ppc64_caches.dline_size;
658 icache_bsize = ppc64_caches.iline_size;
659
40ef8cbc 660 if (ppc_md.panic)
7e990266 661 setup_panic();
40ef8cbc 662
4846c5de 663 init_mm.start_code = (unsigned long)_stext;
40ef8cbc
PM
664 init_mm.end_code = (unsigned long) _etext;
665 init_mm.end_data = (unsigned long) _edata;
666 init_mm.brk = klimit;
5c1f6ee9
AK
667#ifdef CONFIG_PPC_64K_PAGES
668 init_mm.context.pte_frag = NULL;
669#endif
40ef8cbc 670 irqstack_early_init();
2d27cfd3 671 exc_lvl_early_init();
40ef8cbc
PM
672 emergency_stack_init();
673
94491685 674#ifdef CONFIG_PPC_STD_MMU_64
40ef8cbc 675 stabs_alloc();
94491685 676#endif
40ef8cbc
PM
677 /* set up the bootmem stuff with available memory */
678 do_init_bootmem();
679 sparse_init();
680
0458060c
PM
681#ifdef CONFIG_DUMMY_CONSOLE
682 conswitchp = &dummy_con;
683#endif
684
38db7e74
GL
685 if (ppc_md.setup_arch)
686 ppc_md.setup_arch();
40ef8cbc 687
40ef8cbc 688 paging_init();
6f0ef0f5
BH
689
690 /* Initialize the MMU context management stuff */
691 mmu_context_init();
692
61e2390e
MN
693 /* Interrupt code needs to be 64K-aligned */
694 if ((unsigned long)_stext & 0xffff)
695 panic("Kernelbase not 64K-aligned (0x%lx)!\n",
696 (unsigned long)_stext);
697
40ef8cbc
PM
698 ppc64_boot_msg(0x15, "Setup Done");
699}
700
701
702/* ToDo: do something useful if ppc_md is not yet setup. */
703#define PPC64_LINUX_FUNCTION 0x0f000000
704#define PPC64_IPL_MESSAGE 0xc0000000
705#define PPC64_TERM_MESSAGE 0xb0000000
706
707static void ppc64_do_msg(unsigned int src, const char *msg)
708{
709 if (ppc_md.progress) {
710 char buf[128];
711
712 sprintf(buf, "%08X\n", src);
713 ppc_md.progress(buf, 0);
714 snprintf(buf, 128, "%s", msg);
715 ppc_md.progress(buf, 0);
716 }
717}
718
719/* Print a boot progress message. */
720void ppc64_boot_msg(unsigned int src, const char *msg)
721{
722 ppc64_do_msg(PPC64_LINUX_FUNCTION|PPC64_IPL_MESSAGE|src, msg);
723 printk("[boot]%04x %s\n", src, msg);
724}
725
7a0268fa 726#ifdef CONFIG_SMP
c2a7e818
TH
727#define PCPU_DYN_SIZE ()
728
729static void * __init pcpu_fc_alloc(unsigned int cpu, size_t size, size_t align)
7a0268fa 730{
c2a7e818
TH
731 return __alloc_bootmem_node(NODE_DATA(cpu_to_node(cpu)), size, align,
732 __pa(MAX_DMA_ADDRESS));
733}
7a0268fa 734
c2a7e818
TH
735static void __init pcpu_fc_free(void *ptr, size_t size)
736{
737 free_bootmem(__pa(ptr), size);
738}
7a0268fa 739
c2a7e818
TH
740static int pcpu_cpu_distance(unsigned int from, unsigned int to)
741{
742 if (cpu_to_node(from) == cpu_to_node(to))
743 return LOCAL_DISTANCE;
744 else
745 return REMOTE_DISTANCE;
746}
747
ae01f84b
AB
748unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
749EXPORT_SYMBOL(__per_cpu_offset);
750
c2a7e818
TH
751void __init setup_per_cpu_areas(void)
752{
753 const size_t dyn_size = PERCPU_MODULE_RESERVE + PERCPU_DYNAMIC_RESERVE;
754 size_t atom_size;
755 unsigned long delta;
756 unsigned int cpu;
757 int rc;
758
759 /*
760 * Linear mapping is one of 4K, 1M and 16M. For 4K, no need
761 * to group units. For larger mappings, use 1M atom which
762 * should be large enough to contain a number of units.
763 */
764 if (mmu_linear_psize == MMU_PAGE_4K)
765 atom_size = PAGE_SIZE;
766 else
767 atom_size = 1 << 20;
768
769 rc = pcpu_embed_first_chunk(0, dyn_size, atom_size, pcpu_cpu_distance,
770 pcpu_fc_alloc, pcpu_fc_free);
771 if (rc < 0)
772 panic("cannot initialize percpu area (err=%d)", rc);
773
774 delta = (unsigned long)pcpu_base_addr - (unsigned long)__per_cpu_start;
ae01f84b
AB
775 for_each_possible_cpu(cpu) {
776 __per_cpu_offset[cpu] = delta + pcpu_unit_offsets[cpu];
777 paca[cpu].data_offset = __per_cpu_offset[cpu];
778 }
7a0268fa
AB
779}
780#endif
4cb3cee0
BH
781
782
ecd73cc5 783#if defined(CONFIG_PPC_INDIRECT_PIO) || defined(CONFIG_PPC_INDIRECT_MMIO)
4cb3cee0
BH
784struct ppc_pci_io ppc_pci_io;
785EXPORT_SYMBOL(ppc_pci_io);
ecd73cc5 786#endif