setup_64.c 17.9 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14
/*
 * 
 * Common boot and setup code.
 *
 * Copyright (C) 2001 PPC64 Team, IBM Corp
 *
 *      This program is free software; you can redistribute it and/or
 *      modify it under the terms of the GNU General Public License
 *      as published by the Free Software Foundation; either version
 *      2 of the License, or (at your option) any later version.
 */

#undef DEBUG

15
#include <linux/export.h>
16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
#include <linux/string.h>
#include <linux/sched.h>
#include <linux/init.h>
#include <linux/kernel.h>
#include <linux/reboot.h>
#include <linux/delay.h>
#include <linux/initrd.h>
#include <linux/seq_file.h>
#include <linux/ioport.h>
#include <linux/console.h>
#include <linux/utsname.h>
#include <linux/tty.h>
#include <linux/root_dev.h>
#include <linux/notifier.h>
#include <linux/cpu.h>
#include <linux/unistd.h>
#include <linux/serial.h>
#include <linux/serial_8250.h>
34
#include <linux/bootmem.h>
35
#include <linux/pci.h>
36
#include <linux/lockdep.h>
Y
Yinghai Lu 已提交
37
#include <linux/memblock.h>
38 39
#include <linux/hugetlb.h>

40
#include <asm/io.h>
41
#include <asm/kdump.h>
42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61
#include <asm/prom.h>
#include <asm/processor.h>
#include <asm/pgtable.h>
#include <asm/smp.h>
#include <asm/elf.h>
#include <asm/machdep.h>
#include <asm/paca.h>
#include <asm/time.h>
#include <asm/cputable.h>
#include <asm/sections.h>
#include <asm/btext.h>
#include <asm/nvram.h>
#include <asm/setup.h>
#include <asm/rtas.h>
#include <asm/iommu.h>
#include <asm/serial.h>
#include <asm/cache.h>
#include <asm/page.h>
#include <asm/mmu.h>
#include <asm/firmware.h>
P
Paul Mackerras 已提交
62
#include <asm/xmon.h>
D
David Gibson 已提交
63
#include <asm/udbg.h>
64
#include <asm/kexec.h>
65
#include <asm/mmu_context.h>
66
#include <asm/code-patching.h>
67
#include <asm/kvm_ppc.h>
68
#include <asm/hugetlb.h>
69

S
Stephen Rothwell 已提交
70 71
#include "setup.h"

72 73 74 75 76 77 78
#ifdef DEBUG
#define DBG(fmt...) udbg_printf(fmt)
#else
#define DBG(fmt...)
#endif

int boot_cpuid = 0;
79
int __initdata spinning_secondaries;
80 81
u64 ppc64_pft_size;

82 83 84 85
/* Pick defaults since we might want to patch instructions
 * before we've read this from the device tree.
 */
struct ppc64_caches ppc64_caches = {
86 87 88 89
	.dline_size = 0x40,
	.log_dline_size = 6,
	.iline_size = 0x40,
	.log_iline_size = 6
90
};
91 92 93 94 95 96 97 98 99 100 101 102
EXPORT_SYMBOL_GPL(ppc64_caches);

/*
 * These are used in binfmt_elf.c to put aux entries on the stack
 * for each elf executable being started.
 */
int dcache_bsize;
int icache_bsize;
int ucache_bsize;

#ifdef CONFIG_SMP

103
static char *smt_enabled_cmdline;
104 105 106 107 108

/* Look for ibm,smt-enabled OF option */
static void check_smt_enabled(void)
{
	struct device_node *dn;
109
	const char *smt_option;
110

111 112
	/* Default to enabling all threads */
	smt_enabled_at_boot = threads_per_core;
113

114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144
	/* Allow the command line to overrule the OF option */
	if (smt_enabled_cmdline) {
		if (!strcmp(smt_enabled_cmdline, "on"))
			smt_enabled_at_boot = threads_per_core;
		else if (!strcmp(smt_enabled_cmdline, "off"))
			smt_enabled_at_boot = 0;
		else {
			long smt;
			int rc;

			rc = strict_strtol(smt_enabled_cmdline, 10, &smt);
			if (!rc)
				smt_enabled_at_boot =
					min(threads_per_core, (int)smt);
		}
	} else {
		dn = of_find_node_by_path("/options");
		if (dn) {
			smt_option = of_get_property(dn, "ibm,smt-enabled",
						     NULL);

			if (smt_option) {
				if (!strcmp(smt_option, "on"))
					smt_enabled_at_boot = threads_per_core;
				else if (!strcmp(smt_option, "off"))
					smt_enabled_at_boot = 0;
			}

			of_node_put(dn);
		}
	}
145 146 147 148 149
}

/* Look for smt-enabled= cmdline option */
static int __init early_smt_enabled(char *p)
{
150
	smt_enabled_cmdline = p;
151 152 153 154
	return 0;
}
early_param("smt-enabled", early_smt_enabled);

P
Paul Mackerras 已提交
155 156
#else
#define check_smt_enabled()
157 158
#endif /* CONFIG_SMP */

159 160 161 162 163 164 165 166 167
/** Fix up paca fields required for the boot cpu */
static void fixup_boot_paca(void)
{
	/* The boot cpu is started */
	get_paca()->cpu_start = 1;
	/* Allow percpu accesses to work until we setup percpu data */
	get_paca()->data_offset = 0;
}

168 169 170 171 172 173
/*
 * Early initialization entry point. This is called by head.S
 * with MMU translation disabled. We rely on the "feature" of
 * the CPU that ignores the top 2 bits of the address in real
 * mode so we can access kernel globals normally provided we
 * only toy with things in the RMO region. From here, we do
Y
Yinghai Lu 已提交
174
 * some early parsing of the device-tree to setup out MEMBLOCK
175 176 177 178 179 180 181 182 183 184 185 186 187 188
 * data structures, and allocate & initialize the hash table
 * and segment tables so we can start running with translation
 * enabled.
 *
 * It is this function which will call the probe() callback of
 * the various platform types and copy the matching one to the
 * global ppc_md structure. Your platform can eventually do
 * some very early initializations from the probe() routine, but
 * this is not recommended, be very careful as, for example, the
 * device-tree is not accessible via normal means at this point.
 */

void __init early_setup(unsigned long dt_ptr)
{
189 190
	static __initdata struct paca_struct boot_paca;

191 192
	/* -------- printk is _NOT_ safe to use here ! ------- */

193
	/* Identify CPU type */
194
	identify_cpu(0, mfspr(SPRN_PVR));
195

196
	/* Assume we're on cpu 0 for now. Don't write to the paca yet! */
197 198
	initialise_paca(&boot_paca, 0);
	setup_paca(&boot_paca);
199
	fixup_boot_paca();
200

201 202 203
	/* Initialize lockdep early or else spinlocks will blow */
	lockdep_init();

204 205
	/* -------- printk is now safe to use ------- */

206 207 208
	/* Enable early debugging if any specified (see udbg.h) */
	udbg_early_init();

209
 	DBG(" -> early_setup(), dt_ptr: 0x%lx\n", dt_ptr);
210 211

	/*
212 213 214
	 * Do early initialization using the flattened device
	 * tree, such as retrieving the physical memory map or
	 * calculating/retrieving the hash table size.
215 216 217
	 */
	early_init_devtree(__va(dt_ptr));

218
	/* Now we know the logical id of our boot cpu, setup the paca. */
219
	setup_paca(&paca[boot_cpuid]);
220
	fixup_boot_paca();
221

222 223
	/* Probe the machine type */
	probe_machine();
224

225
	setup_kdump_trampoline();
226

227 228
	DBG("Found, Initializing memory management...\n");

229 230
	/* Initialize the hash table or TLB handling */
	early_init_mmu();
231

232 233 234 235 236 237 238
	/*
	 * Reserve any gigantic pages requested on the command line.
	 * memblock needs to have been initialized by the time this is
	 * called since this will reserve memory.
	 */
	reserve_hugetlb_gpages();

239 240 241
	DBG(" <- early_setup()\n");
}

242 243 244
#ifdef CONFIG_SMP
void early_setup_secondary(void)
{
245
	/* Mark interrupts enabled in PACA */
246
	get_paca()->soft_enabled = 0;
247

248 249
	/* Initialize the hash table or TLB handling */
	early_init_mmu_secondary();
250 251 252
}

#endif /* CONFIG_SMP */
253

254 255 256
#if defined(CONFIG_SMP) || defined(CONFIG_KEXEC)
void smp_release_cpus(void)
{
257
	unsigned long *ptr;
258
	int i;
259 260 261 262 263 264 265

	DBG(" -> smp_release_cpus()\n");

	/* All secondary cpus are spinning on a common spinloop, release them
	 * all now so they can start to spin on their individual paca
	 * spinloops. For non SMP kernels, the secondary cpus never get out
	 * of the common spinloop.
266
	 */
267

268 269
	ptr  = (unsigned long *)((unsigned long)&__secondary_hold_spinloop
			- PHYSICAL_START);
270
	*ptr = __pa(generic_secondary_smp_init);
271 272 273 274 275

	/* And wait a bit for them to catch up */
	for (i = 0; i < 100000; i++) {
		mb();
		HMT_low();
276
		if (spinning_secondaries == 0)
277 278 279
			break;
		udelay(1);
	}
280
	DBG("spinning_secondaries = %d\n", spinning_secondaries);
281 282 283 284 285

	DBG(" <- smp_release_cpus()\n");
}
#endif /* CONFIG_SMP || CONFIG_KEXEC */

286
/*
287 288
 * Initialize some remaining members of the ppc64_caches and systemcfg
 * structures
289 290 291 292 293 294 295 296 297 298 299
 * (at least until we get rid of them completely). This is mostly some
 * cache informations about the CPU that will be used by cache flush
 * routines and/or provided to userland
 */
static void __init initialize_cache_info(void)
{
	struct device_node *np;
	unsigned long num_cpus = 0;

	DBG(" -> initialize_cache_info()\n");

300
	for_each_node_by_type(np, "cpu") {
301 302
		num_cpus += 1;

A
Anton Blanchard 已提交
303 304
		/*
		 * We're assuming *all* of the CPUs have the same
305 306
		 * d-cache and i-cache sizes... -Peter
		 */
A
Anton Blanchard 已提交
307
		if (num_cpus == 1) {
308
			const u32 *sizep, *lsizep;
309 310 311 312
			u32 size, lsize;

			size = 0;
			lsize = cur_cpu_spec->dcache_bsize;
313
			sizep = of_get_property(np, "d-cache-size", NULL);
314 315
			if (sizep != NULL)
				size = *sizep;
A
Anton Blanchard 已提交
316 317
			lsizep = of_get_property(np, "d-cache-block-size",
						 NULL);
318 319
			/* fallback if block size missing */
			if (lsizep == NULL)
A
Anton Blanchard 已提交
320 321 322
				lsizep = of_get_property(np,
							 "d-cache-line-size",
							 NULL);
323 324 325 326 327 328
			if (lsizep != NULL)
				lsize = *lsizep;
			if (sizep == 0 || lsizep == 0)
				DBG("Argh, can't find dcache properties ! "
				    "sizep: %p, lsizep: %p\n", sizep, lsizep);

329 330
			ppc64_caches.dsize = size;
			ppc64_caches.dline_size = lsize;
331 332 333 334 335
			ppc64_caches.log_dline_size = __ilog2(lsize);
			ppc64_caches.dlines_per_page = PAGE_SIZE / lsize;

			size = 0;
			lsize = cur_cpu_spec->icache_bsize;
336
			sizep = of_get_property(np, "i-cache-size", NULL);
337 338
			if (sizep != NULL)
				size = *sizep;
A
Anton Blanchard 已提交
339 340
			lsizep = of_get_property(np, "i-cache-block-size",
						 NULL);
341
			if (lsizep == NULL)
A
Anton Blanchard 已提交
342 343 344
				lsizep = of_get_property(np,
							 "i-cache-line-size",
							 NULL);
345 346 347 348 349 350
			if (lsizep != NULL)
				lsize = *lsizep;
			if (sizep == 0 || lsizep == 0)
				DBG("Argh, can't find icache properties ! "
				    "sizep: %p, lsizep: %p\n", sizep, lsizep);

351 352
			ppc64_caches.isize = size;
			ppc64_caches.iline_size = lsize;
353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369
			ppc64_caches.log_iline_size = __ilog2(lsize);
			ppc64_caches.ilines_per_page = PAGE_SIZE / lsize;
		}
	}

	DBG(" <- initialize_cache_info()\n");
}


/*
 * Do some initial setup of the system.  The parameters are those which 
 * were passed in from the bootloader.
 */
void __init setup_system(void)
{
	DBG(" -> setup_system()\n");

370 371
	/* Apply the CPUs-specific and firmware specific fixups to kernel
	 * text (nop out sections not relevant to this CPU or this firmware)
372
	 */
373
	do_feature_fixups(cur_cpu_spec->cpu_features,
374
			  &__start___ftr_fixup, &__stop___ftr_fixup);
375 376
	do_feature_fixups(cur_cpu_spec->mmu_features,
			  &__start___mmu_ftr_fixup, &__stop___mmu_ftr_fixup);
377 378
	do_feature_fixups(powerpc_firmware_features,
			  &__start___fw_ftr_fixup, &__stop___fw_ftr_fixup);
K
Kumar Gala 已提交
379 380
	do_lwsync_fixups(cur_cpu_spec->cpu_features,
			 &__start___lwsync_fixup, &__stop___lwsync_fixup);
381
	do_final_fixups();
382

383 384 385 386 387 388 389
	/*
	 * Unflatten the device-tree passed by prom_init or kexec
	 */
	unflatten_device_tree();

	/*
	 * Fill the ppc64_caches & systemcfg structures with informations
390
 	 * retrieved from the device-tree.
391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410
	 */
	initialize_cache_info();

#ifdef CONFIG_PPC_RTAS
	/*
	 * Initialize RTAS if available
	 */
	rtas_initialize();
#endif /* CONFIG_PPC_RTAS */

	/*
	 * Check if we have an initrd provided via the device-tree
	 */
	check_for_initrd();

	/*
	 * Do some platform specific early initializations, that includes
	 * setting up the hash table pointers. It also sets up some interrupt-mapping
	 * related options that will be used by finish_device_tree()
	 */
411 412
	if (ppc_md.init_early)
		ppc_md.init_early();
413

414 415 416 417 418 419 420
 	/*
	 * We can discover serial ports now since the above did setup the
	 * hash table management for us, thus ioremap works. We do that early
	 * so that further code can be debugged
	 */
	find_legacy_serial_ports();

421 422 423 424 425
	/*
	 * Register early console
	 */
	register_early_udbg_console();

426 427 428 429
	/*
	 * Initialize xmon
	 */
	xmon_setup();
430

P
Paul Mackerras 已提交
431
	smp_setup_cpu_maps();
432
	check_smt_enabled();
433

434
#ifdef CONFIG_SMP
435 436 437 438
	/* Release secondary cpus out of their spinloops at 0x60 now that
	 * we can map physical -> logical CPU ids
	 */
	smp_release_cpus();
439
#endif
440

441
	printk("Starting Linux PPC64 %s\n", init_utsname()->version);
442 443

	printk("-----------------------------------------------------\n");
444
	printk("ppc64_pft_size                = 0x%llx\n", ppc64_pft_size);
Y
Yinghai Lu 已提交
445
	printk("physicalMemorySize            = 0x%llx\n", memblock_phys_mem_size());
446 447 448 449 450 451
	if (ppc64_caches.dline_size != 0x80)
		printk("ppc64_caches.dcache_line_size = 0x%x\n",
		       ppc64_caches.dline_size);
	if (ppc64_caches.iline_size != 0x80)
		printk("ppc64_caches.icache_line_size = 0x%x\n",
		       ppc64_caches.iline_size);
452
#ifdef CONFIG_PPC_STD_MMU_64
453 454
	if (htab_address)
		printk("htab_address                  = 0x%p\n", htab_address);
455
	printk("htab_hash_mask                = 0x%lx\n", htab_hash_mask);
456
#endif /* CONFIG_PPC_STD_MMU_64 */
457
	if (PHYSICAL_START > 0)
458 459
		printk("physical_start                = 0x%llx\n",
		       (unsigned long long)PHYSICAL_START);
460 461 462 463 464
	printk("-----------------------------------------------------\n");

	DBG(" <- setup_system()\n");
}

465 466 467 468 469 470
/* This returns the limit below which memory accesses to the linear
 * mapping are guarnateed not to cause a TLB or SLB miss. This is
 * used to allocate interrupt or emergency stacks for which our
 * exception entry path doesn't deal with being interrupted.
 */
static u64 safe_stack_limit(void)
471
{
472 473 474 475 476 477 478 479 480
#ifdef CONFIG_PPC_BOOK3E
	/* Freescale BookE bolts the entire linear mapping */
	if (mmu_has_feature(MMU_FTR_TYPE_FSL_E))
		return linear_map_top;
	/* Other BookE, we assume the first GB is bolted */
	return 1ul << 30;
#else
	/* BookS, the first segment is bolted */
	if (mmu_has_feature(MMU_FTR_1T_SEGMENT))
481 482
		return 1UL << SID_SHIFT_1T;
	return 1UL << SID_SHIFT;
483
#endif
484 485
}

486 487
static void __init irqstack_early_init(void)
{
488
	u64 limit = safe_stack_limit();
489 490 491
	unsigned int i;

	/*
492 493
	 * Interrupt stacks must be in the first segment since we
	 * cannot afford to take SLB misses on them.
494
	 */
495
	for_each_possible_cpu(i) {
496
		softirq_ctx[i] = (struct thread_info *)
Y
Yinghai Lu 已提交
497
			__va(memblock_alloc_base(THREAD_SIZE,
498
					    THREAD_SIZE, limit));
499
		hardirq_ctx[i] = (struct thread_info *)
Y
Yinghai Lu 已提交
500
			__va(memblock_alloc_base(THREAD_SIZE,
501
					    THREAD_SIZE, limit));
502 503 504
	}
}

505 506 507
#ifdef CONFIG_PPC_BOOK3E
static void __init exc_lvl_early_init(void)
{
508 509 510
	extern unsigned int interrupt_base_book3e;
	extern unsigned int exc_debug_debug_book3e;

511 512 513 514
	unsigned int i;

	for_each_possible_cpu(i) {
		critirq_ctx[i] = (struct thread_info *)
Y
Yinghai Lu 已提交
515
			__va(memblock_alloc(THREAD_SIZE, THREAD_SIZE));
516
		dbgirq_ctx[i] = (struct thread_info *)
Y
Yinghai Lu 已提交
517
			__va(memblock_alloc(THREAD_SIZE, THREAD_SIZE));
518
		mcheckirq_ctx[i] = (struct thread_info *)
Y
Yinghai Lu 已提交
519
			__va(memblock_alloc(THREAD_SIZE, THREAD_SIZE));
520
	}
521 522 523 524

	if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC))
		patch_branch(&interrupt_base_book3e + (0x040 / 4) + 1,
			     (unsigned long)&exc_debug_debug_book3e, 0);
525 526 527 528 529
}
#else
#define exc_lvl_early_init()
#endif

530 531 532 533 534 535
/*
 * Stack space used when we detect a bad kernel stack pointer, and
 * early in SMP boots before relocation is enabled.
 */
static void __init emergency_stack_init(void)
{
536
	u64 limit;
537 538 539 540 541 542 543 544 545 546 547
	unsigned int i;

	/*
	 * Emergency stacks must be under 256MB, we cannot afford to take
	 * SLB misses on them. The ABI also requires them to be 128-byte
	 * aligned.
	 *
	 * Since we use these as temporary stacks during secondary CPU
	 * bringup, we need to get at them in real mode. This means they
	 * must also be within the RMO region.
	 */
548
	limit = min(safe_stack_limit(), ppc64_rma_size);
549

550 551
	for_each_possible_cpu(i) {
		unsigned long sp;
Y
Yinghai Lu 已提交
552
		sp  = memblock_alloc_base(THREAD_SIZE, THREAD_SIZE, limit);
553 554 555
		sp += THREAD_SIZE;
		paca[i].emergency_sp = __va(sp);
	}
556 557 558
}

/*
559 560
 * Called into from start_kernel this initializes bootmem, which is used
 * to manage page allocation until mem_init is called.
561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579
 */
void __init setup_arch(char **cmdline_p)
{
	ppc64_boot_msg(0x12, "Setup Arch");

	*cmdline_p = cmd_line;

	/*
	 * Set cache line size based on type of cpu as a default.
	 * Systems with OF can look in the properties on the cpu node(s)
	 * for a possibly more accurate value.
	 */
	dcache_bsize = ppc64_caches.dline_size;
	icache_bsize = ppc64_caches.iline_size;

	/* reboot on panic */
	panic_timeout = 180;

	if (ppc_md.panic)
580
		setup_panic();
581

582
	init_mm.start_code = (unsigned long)_stext;
583 584 585 586 587
	init_mm.end_code = (unsigned long) _etext;
	init_mm.end_data = (unsigned long) _edata;
	init_mm.brk = klimit;
	
	irqstack_early_init();
588
	exc_lvl_early_init();
589 590
	emergency_stack_init();

591
#ifdef CONFIG_PPC_STD_MMU_64
592
	stabs_alloc();
593
#endif
594 595 596 597
	/* set up the bootmem stuff with available memory */
	do_init_bootmem();
	sparse_init();

598 599 600 601
#ifdef CONFIG_DUMMY_CONSOLE
	conswitchp = &dummy_con;
#endif

602 603
	if (ppc_md.setup_arch)
		ppc_md.setup_arch();
604 605

	paging_init();
606 607 608 609

	/* Initialize the MMU context management stuff */
	mmu_context_init();

610
	kvm_linear_init();
611

612 613 614 615 616
	/* Interrupt code needs to be 64K-aligned */
	if ((unsigned long)_stext & 0xffff)
		panic("Kernelbase not 64K-aligned (0x%lx)!\n",
		      (unsigned long)_stext);

617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644
	ppc64_boot_msg(0x15, "Setup Done");
}


/* ToDo: do something useful if ppc_md is not yet setup. */
#define PPC64_LINUX_FUNCTION 0x0f000000
#define PPC64_IPL_MESSAGE 0xc0000000
#define PPC64_TERM_MESSAGE 0xb0000000

static void ppc64_do_msg(unsigned int src, const char *msg)
{
	if (ppc_md.progress) {
		char buf[128];

		sprintf(buf, "%08X\n", src);
		ppc_md.progress(buf, 0);
		snprintf(buf, 128, "%s", msg);
		ppc_md.progress(buf, 0);
	}
}

/* Print a boot progress message. */
void ppc64_boot_msg(unsigned int src, const char *msg)
{
	ppc64_do_msg(PPC64_LINUX_FUNCTION|PPC64_IPL_MESSAGE|src, msg);
	printk("[boot]%04x %s\n", src, msg);
}

645
#ifdef CONFIG_SMP
646 647 648
#define PCPU_DYN_SIZE		()

static void * __init pcpu_fc_alloc(unsigned int cpu, size_t size, size_t align)
649
{
650 651 652
	return __alloc_bootmem_node(NODE_DATA(cpu_to_node(cpu)), size, align,
				    __pa(MAX_DMA_ADDRESS));
}
653

654 655 656 657
static void __init pcpu_fc_free(void *ptr, size_t size)
{
	free_bootmem(__pa(ptr), size);
}
658

659 660 661 662 663 664 665 666
static int pcpu_cpu_distance(unsigned int from, unsigned int to)
{
	if (cpu_to_node(from) == cpu_to_node(to))
		return LOCAL_DISTANCE;
	else
		return REMOTE_DISTANCE;
}

667 668 669
unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
EXPORT_SYMBOL(__per_cpu_offset);

670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693
void __init setup_per_cpu_areas(void)
{
	const size_t dyn_size = PERCPU_MODULE_RESERVE + PERCPU_DYNAMIC_RESERVE;
	size_t atom_size;
	unsigned long delta;
	unsigned int cpu;
	int rc;

	/*
	 * Linear mapping is one of 4K, 1M and 16M.  For 4K, no need
	 * to group units.  For larger mappings, use 1M atom which
	 * should be large enough to contain a number of units.
	 */
	if (mmu_linear_psize == MMU_PAGE_4K)
		atom_size = PAGE_SIZE;
	else
		atom_size = 1 << 20;

	rc = pcpu_embed_first_chunk(0, dyn_size, atom_size, pcpu_cpu_distance,
				    pcpu_fc_alloc, pcpu_fc_free);
	if (rc < 0)
		panic("cannot initialize percpu area (err=%d)", rc);

	delta = (unsigned long)pcpu_base_addr - (unsigned long)__per_cpu_start;
694 695 696 697
	for_each_possible_cpu(cpu) {
                __per_cpu_offset[cpu] = delta + pcpu_unit_offsets[cpu];
		paca[cpu].data_offset = __per_cpu_offset[cpu];
	}
698 699
}
#endif
700 701 702 703 704 705 706


#ifdef CONFIG_PPC_INDIRECT_IO
struct ppc_pci_io ppc_pci_io;
EXPORT_SYMBOL(ppc_pci_io);
#endif /* CONFIG_PPC_INDIRECT_IO */