smp.c 24.1 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
/*
 * SMP support for power macintosh.
 *
 * We support both the old "powersurge" SMP architecture
 * and the current Core99 (G4 PowerMac) machines.
 *
 * Note that we don't support the very first rev. of
 * Apple/DayStar 2 CPUs board, the one with the funky
 * watchdog. Hopefully, none of these should be there except
 * maybe internally to Apple. I should probably still add some
 * code to detect this card though and disable SMP. --BenH.
 *
 * Support Macintosh G4 SMP by Troy Benjegerdes (hozer@drgw.net)
 * and Ben Herrenschmidt <benh@kernel.crashing.org>.
 *
 * Support for DayStar quad CPU cards
 * Copyright (C) XLR8, Inc. 1994-2000
 *
 *  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.
 */
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/smp.h>
#include <linux/interrupt.h>
#include <linux/kernel_stat.h>
#include <linux/delay.h>
#include <linux/init.h>
#include <linux/spinlock.h>
#include <linux/errno.h>
#include <linux/hardirq.h>
#include <linux/cpu.h>
35
#include <linux/compiler.h>
36 37 38

#include <asm/ptrace.h>
#include <asm/atomic.h>
39
#include <asm/code-patching.h>
40 41 42 43 44 45 46 47 48 49
#include <asm/irq.h>
#include <asm/page.h>
#include <asm/pgtable.h>
#include <asm/sections.h>
#include <asm/io.h>
#include <asm/prom.h>
#include <asm/smp.h>
#include <asm/machdep.h>
#include <asm/pmac_feature.h>
#include <asm/time.h>
50
#include <asm/mpic.h>
51 52
#include <asm/cacheflush.h>
#include <asm/keylargo.h>
53
#include <asm/pmac_low_i2c.h>
54
#include <asm/pmac_pfunc.h>
55

56 57
#include "pmac.h"

58
#undef DEBUG
59 60 61 62 63 64 65 66

#ifdef DEBUG
#define DBG(fmt...) udbg_printf(fmt)
#else
#define DBG(fmt...)
#endif

extern void __secondary_start_pmac_0(void);
67
extern int pmac_pfunc_base_install(void);
68

69 70 71
static void (*pmac_tb_freeze)(int freeze);
static u64 timebase;
static int tb_req;
72

73
#ifdef CONFIG_PPC32
74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158

/*
 * Powersurge (old powermac SMP) support.
 */

/* Addresses for powersurge registers */
#define HAMMERHEAD_BASE		0xf8000000
#define HHEAD_CONFIG		0x90
#define HHEAD_SEC_INTR		0xc0

/* register for interrupting the primary processor on the powersurge */
/* N.B. this is actually the ethernet ROM! */
#define PSURGE_PRI_INTR		0xf3019000

/* register for storing the start address for the secondary processor */
/* N.B. this is the PCI config space address register for the 1st bridge */
#define PSURGE_START		0xf2800000

/* Daystar/XLR8 4-CPU card */
#define PSURGE_QUAD_REG_ADDR	0xf8800000

#define PSURGE_QUAD_IRQ_SET	0
#define PSURGE_QUAD_IRQ_CLR	1
#define PSURGE_QUAD_IRQ_PRIMARY	2
#define PSURGE_QUAD_CKSTOP_CTL	3
#define PSURGE_QUAD_PRIMARY_ARB	4
#define PSURGE_QUAD_BOARD_ID	6
#define PSURGE_QUAD_WHICH_CPU	7
#define PSURGE_QUAD_CKSTOP_RDBK	8
#define PSURGE_QUAD_RESET_CTL	11

#define PSURGE_QUAD_OUT(r, v)	(out_8(quad_base + ((r) << 4) + 4, (v)))
#define PSURGE_QUAD_IN(r)	(in_8(quad_base + ((r) << 4) + 4) & 0x0f)
#define PSURGE_QUAD_BIS(r, v)	(PSURGE_QUAD_OUT((r), PSURGE_QUAD_IN(r) | (v)))
#define PSURGE_QUAD_BIC(r, v)	(PSURGE_QUAD_OUT((r), PSURGE_QUAD_IN(r) & ~(v)))

/* virtual addresses for the above */
static volatile u8 __iomem *hhead_base;
static volatile u8 __iomem *quad_base;
static volatile u32 __iomem *psurge_pri_intr;
static volatile u8 __iomem *psurge_sec_intr;
static volatile u32 __iomem *psurge_start;

/* values for psurge_type */
#define PSURGE_NONE		-1
#define PSURGE_DUAL		0
#define PSURGE_QUAD_OKEE	1
#define PSURGE_QUAD_COTTON	2
#define PSURGE_QUAD_ICEGRASS	3

/* what sort of powersurge board we have */
static int psurge_type = PSURGE_NONE;

/*
 * Set and clear IPIs for powersurge.
 */
static inline void psurge_set_ipi(int cpu)
{
	if (psurge_type == PSURGE_NONE)
		return;
	if (cpu == 0)
		in_be32(psurge_pri_intr);
	else if (psurge_type == PSURGE_DUAL)
		out_8(psurge_sec_intr, 0);
	else
		PSURGE_QUAD_OUT(PSURGE_QUAD_IRQ_SET, 1 << cpu);
}

static inline void psurge_clr_ipi(int cpu)
{
	if (cpu > 0) {
		switch(psurge_type) {
		case PSURGE_DUAL:
			out_8(psurge_sec_intr, ~0);
		case PSURGE_NONE:
			break;
		default:
			PSURGE_QUAD_OUT(PSURGE_QUAD_IRQ_CLR, 1 << cpu);
		}
	}
}

/*
 * On powersurge (old SMP powermac architecture) we don't have
 * separate IPIs for separate messages like openpic does.  Instead
159
 * use the generic demux helpers
160 161
 *  -- paulus.
 */
O
Olaf Hering 已提交
162
void psurge_smp_message_recv(void)
163
{
164 165
	psurge_clr_ipi(smp_processor_id());
	smp_ipi_demux();
166 167
}

O
Olaf Hering 已提交
168
irqreturn_t psurge_primary_intr(int irq, void *d)
169
{
O
Olaf Hering 已提交
170
	psurge_smp_message_recv();
171 172 173
	return IRQ_HANDLED;
}

174
static void smp_psurge_cause_ipi(int cpu, unsigned long data)
175
{
176
	psurge_set_ipi(cpu);
177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242
}

/*
 * Determine a quad card presence. We read the board ID register, we
 * force the data bus to change to something else, and we read it again.
 * It it's stable, then the register probably exist (ugh !)
 */
static int __init psurge_quad_probe(void)
{
	int type;
	unsigned int i;

	type = PSURGE_QUAD_IN(PSURGE_QUAD_BOARD_ID);
	if (type < PSURGE_QUAD_OKEE || type > PSURGE_QUAD_ICEGRASS
	    || type != PSURGE_QUAD_IN(PSURGE_QUAD_BOARD_ID))
		return PSURGE_DUAL;

	/* looks OK, try a slightly more rigorous test */
	/* bogus is not necessarily cacheline-aligned,
	   though I don't suppose that really matters.  -- paulus */
	for (i = 0; i < 100; i++) {
		volatile u32 bogus[8];
		bogus[(0+i)%8] = 0x00000000;
		bogus[(1+i)%8] = 0x55555555;
		bogus[(2+i)%8] = 0xFFFFFFFF;
		bogus[(3+i)%8] = 0xAAAAAAAA;
		bogus[(4+i)%8] = 0x33333333;
		bogus[(5+i)%8] = 0xCCCCCCCC;
		bogus[(6+i)%8] = 0xCCCCCCCC;
		bogus[(7+i)%8] = 0x33333333;
		wmb();
		asm volatile("dcbf 0,%0" : : "r" (bogus) : "memory");
		mb();
		if (type != PSURGE_QUAD_IN(PSURGE_QUAD_BOARD_ID))
			return PSURGE_DUAL;
	}
	return type;
}

static void __init psurge_quad_init(void)
{
	int procbits;

	if (ppc_md.progress) ppc_md.progress("psurge_quad_init", 0x351);
	procbits = ~PSURGE_QUAD_IN(PSURGE_QUAD_WHICH_CPU);
	if (psurge_type == PSURGE_QUAD_ICEGRASS)
		PSURGE_QUAD_BIS(PSURGE_QUAD_RESET_CTL, procbits);
	else
		PSURGE_QUAD_BIC(PSURGE_QUAD_CKSTOP_CTL, procbits);
	mdelay(33);
	out_8(psurge_sec_intr, ~0);
	PSURGE_QUAD_OUT(PSURGE_QUAD_IRQ_CLR, procbits);
	PSURGE_QUAD_BIS(PSURGE_QUAD_RESET_CTL, procbits);
	if (psurge_type != PSURGE_QUAD_ICEGRASS)
		PSURGE_QUAD_BIS(PSURGE_QUAD_CKSTOP_CTL, procbits);
	PSURGE_QUAD_BIC(PSURGE_QUAD_PRIMARY_ARB, procbits);
	mdelay(33);
	PSURGE_QUAD_BIC(PSURGE_QUAD_RESET_CTL, procbits);
	mdelay(33);
	PSURGE_QUAD_BIS(PSURGE_QUAD_PRIMARY_ARB, procbits);
	mdelay(33);
}

static int __init smp_psurge_probe(void)
{
	int i, ncpus;
243
	struct device_node *dn;
244 245 246 247 248 249 250 251 252 253 254 255 256 257 258

	/* We don't do SMP on the PPC601 -- paulus */
	if (PVR_VER(mfspr(SPRN_PVR)) == 1)
		return 1;

	/*
	 * The powersurge cpu board can be used in the generation
	 * of powermacs that have a socket for an upgradeable cpu card,
	 * including the 7500, 8500, 9500, 9600.
	 * The device tree doesn't tell you if you have 2 cpus because
	 * OF doesn't know anything about the 2nd processor.
	 * Instead we look for magic bits in magic registers,
	 * in the hammerhead memory controller in the case of the
	 * dual-cpu powersurge board.  -- paulus.
	 */
259 260
	dn = of_find_node_by_name(NULL, "hammerhead");
	if (dn == NULL)
261
		return 1;
262
	of_node_put(dn);
263 264 265 266 267 268 269 270 271 272

	hhead_base = ioremap(HAMMERHEAD_BASE, 0x800);
	quad_base = ioremap(PSURGE_QUAD_REG_ADDR, 1024);
	psurge_sec_intr = hhead_base + HHEAD_SEC_INTR;

	psurge_type = psurge_quad_probe();
	if (psurge_type != PSURGE_DUAL) {
		psurge_quad_init();
		/* All released cards using this HW design have 4 CPUs */
		ncpus = 4;
273 274 275
		/* No sure how timebase sync works on those, let's use SW */
		smp_ops->give_timebase = smp_generic_give_timebase;
		smp_ops->take_timebase = smp_generic_take_timebase;
276 277 278 279 280 281 282 283 284 285 286 287 288 289
	} else {
		iounmap(quad_base);
		if ((in_8(hhead_base + HHEAD_CONFIG) & 0x02) == 0) {
			/* not a dual-cpu card */
			iounmap(hhead_base);
			psurge_type = PSURGE_NONE;
			return 1;
		}
		ncpus = 2;
	}

	psurge_start = ioremap(PSURGE_START, 4);
	psurge_pri_intr = ioremap(PSURGE_PRI_INTR, 4);

290
	/* This is necessary because OF doesn't know about the
291 292
	 * secondary cpu(s), and thus there aren't nodes in the
	 * device tree for them, and smp_setup_cpu_maps hasn't
293
	 * set their bits in cpu_present_mask.
294 295 296
	 */
	if (ncpus > NR_CPUS)
		ncpus = NR_CPUS;
297
	for (i = 1; i < ncpus ; ++i)
298
		set_cpu_present(i, true);
299 300 301 302 303 304

	if (ppc_md.progress) ppc_md.progress("smp_psurge_probe - done", 0x352);

	return ncpus;
}

305
static int __init smp_psurge_kick_cpu(int nr)
306 307
{
	unsigned long start = __pa(__secondary_start_pmac_0) + nr * 8;
308 309 310 311 312 313 314 315
	unsigned long a, flags;
	int i, j;

	/* Defining this here is evil ... but I prefer hiding that
	 * crap to avoid giving people ideas that they can do the
	 * same.
	 */
	extern volatile unsigned int cpu_callin_map[NR_CPUS];
316 317 318 319 320 321 322 323

	/* may need to flush here if secondary bats aren't setup */
	for (a = KERNELBASE; a < KERNELBASE + 0x800000; a += 32)
		asm volatile("dcbf 0,%0" : : "r" (a) : "memory");
	asm volatile("sync");

	if (ppc_md.progress) ppc_md.progress("smp_psurge_kick_cpu", 0x353);

324 325 326
	/* This is going to freeze the timeebase, we disable interrupts */
	local_irq_save(flags);

327 328 329 330
	out_be32(psurge_start, start);
	mb();

	psurge_set_ipi(nr);
331

332 333 334 335
	/*
	 * We can't use udelay here because the timebase is now frozen.
	 */
	for (i = 0; i < 2000; ++i)
336
		asm volatile("nop" : : : "memory");
337 338
	psurge_clr_ipi(nr);

339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361
	/*
	 * Also, because the timebase is frozen, we must not return to the
	 * caller which will try to do udelay's etc... Instead, we wait -here-
	 * for the CPU to callin.
	 */
	for (i = 0; i < 100000 && !cpu_callin_map[nr]; ++i) {
		for (j = 1; j < 10000; j++)
			asm volatile("nop" : : : "memory");
		asm volatile("sync" : : : "memory");
	}
	if (!cpu_callin_map[nr])
		goto stuck;

	/* And we do the TB sync here too for standard dual CPU cards */
	if (psurge_type == PSURGE_DUAL) {
		while(!tb_req)
			barrier();
		tb_req = 0;
		mb();
		timebase = get_tb();
		mb();
		while (timebase)
			barrier();
362 363
		mb();
	}
364 365 366 367
 stuck:
	/* now interrupt the secondary, restarting both TBs */
	if (psurge_type == PSURGE_DUAL)
		psurge_set_ipi(1);
368

369
	if (ppc_md.progress) ppc_md.progress("smp_psurge_kick_cpu - done", 0x354);
370 371

	return 0;
372 373 374 375
}

static struct irqaction psurge_irqaction = {
	.handler = psurge_primary_intr,
376
	.flags = IRQF_DISABLED,
377 378 379 380 381
	.name = "primary IPI",
};

static void __init smp_psurge_setup_cpu(int cpu_nr)
{
382 383
	if (cpu_nr != 0)
		return;
384

385 386 387
	/* reset the entry point so if we get another intr we won't
	 * try to startup again */
	out_be32(psurge_start, 0x100);
388
	if (setup_irq(irq_create_mapping(NULL, 30), &psurge_irqaction))
389
		printk(KERN_ERR "Couldn't get primary IPI interrupt");
390 391 392 393
}

void __init smp_psurge_take_timebase(void)
{
394 395 396 397 398 399 400 401 402 403 404 405
	if (psurge_type != PSURGE_DUAL)
		return;

	tb_req = 1;
	mb();
	while (!timebase)
		barrier();
	mb();
	set_tb(timebase >> 32, timebase & 0xffffffff);
	timebase = 0;
	mb();
	set_dec(tb_ticks_per_jiffy/2);
406 407 408 409
}

void __init smp_psurge_give_timebase(void)
{
410
	/* Nothing to do here */
411 412
}

413 414
/* PowerSurge-style Macs */
struct smp_ops_t psurge_smp_ops = {
415 416
	.message_pass	= smp_muxed_ipi_message_pass,
	.cause_ipi	= smp_psurge_cause_ipi,
417 418 419 420 421 422 423
	.probe		= smp_psurge_probe,
	.kick_cpu	= smp_psurge_kick_cpu,
	.setup_cpu	= smp_psurge_setup_cpu,
	.give_timebase	= smp_psurge_give_timebase,
	.take_timebase	= smp_psurge_take_timebase,
};
#endif /* CONFIG_PPC32 - actually powersurge support */
424

425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470
/*
 * Core 99 and later support
 */


static void smp_core99_give_timebase(void)
{
	unsigned long flags;

	local_irq_save(flags);

	while(!tb_req)
		barrier();
	tb_req = 0;
	(*pmac_tb_freeze)(1);
	mb();
	timebase = get_tb();
	mb();
	while (timebase)
		barrier();
	mb();
	(*pmac_tb_freeze)(0);
	mb();

	local_irq_restore(flags);
}


static void __devinit smp_core99_take_timebase(void)
{
	unsigned long flags;

	local_irq_save(flags);

	tb_req = 1;
	mb();
	while (!timebase)
		barrier();
	mb();
	set_tb(timebase >> 32, timebase & 0xffffffff);
	timebase = 0;
	mb();

	local_irq_restore(flags);
}

471 472 473 474
#ifdef CONFIG_PPC64
/*
 * G5s enable/disable the timebase via an i2c-connected clock chip.
 */
475
static struct pmac_i2c_bus *pmac_tb_clock_chip_host;
476
static u8 pmac_tb_pulsar_addr;
477

478 479 480 481
static void smp_core99_cypress_tb_freeze(int freeze)
{
	u8 data;
	int rc;
482

483 484 485
	/* Strangely, the device-tree says address is 0xd2, but darwin
	 * accesses 0xd0 ...
	 */
486 487 488 489 490
	pmac_i2c_setmode(pmac_tb_clock_chip_host,
			 pmac_i2c_mode_combined);
	rc = pmac_i2c_xfer(pmac_tb_clock_chip_host,
			   0xd0 | pmac_i2c_read,
			   1, 0x81, &data, 1);
491 492 493 494 495
	if (rc != 0)
		goto bail;

	data = (data & 0xf3) | (freeze ? 0x00 : 0x0c);

496 497 498 499
       	pmac_i2c_setmode(pmac_tb_clock_chip_host, pmac_i2c_mode_stdsub);
	rc = pmac_i2c_xfer(pmac_tb_clock_chip_host,
			   0xd0 | pmac_i2c_write,
			   1, 0x81, &data, 1);
500 501 502 503 504 505

 bail:
	if (rc != 0) {
		printk("Cypress Timebase %s rc: %d\n",
		       freeze ? "freeze" : "unfreeze", rc);
		panic("Timebase freeze failed !\n");
506 507 508 509
	}
}


510 511 512 513 514
static void smp_core99_pulsar_tb_freeze(int freeze)
{
	u8 data;
	int rc;

515 516 517 518 519
	pmac_i2c_setmode(pmac_tb_clock_chip_host,
			 pmac_i2c_mode_combined);
	rc = pmac_i2c_xfer(pmac_tb_clock_chip_host,
			   pmac_tb_pulsar_addr | pmac_i2c_read,
			   1, 0x2e, &data, 1);
520 521 522 523 524
	if (rc != 0)
		goto bail;

	data = (data & 0x88) | (freeze ? 0x11 : 0x22);

525 526 527 528
	pmac_i2c_setmode(pmac_tb_clock_chip_host, pmac_i2c_mode_stdsub);
	rc = pmac_i2c_xfer(pmac_tb_clock_chip_host,
			   pmac_tb_pulsar_addr | pmac_i2c_write,
			   1, 0x2e, &data, 1);
529 530 531 532 533 534 535
 bail:
	if (rc != 0) {
		printk(KERN_ERR "Pulsar Timebase %s rc: %d\n",
		       freeze ? "freeze" : "unfreeze", rc);
		panic("Timebase freeze failed !\n");
	}
}
536

537
static void __init smp_core99_setup_i2c_hwsync(int ncpus)
538
{
539 540
	struct device_node *cc = NULL;	
	struct device_node *p;
541
	const char *name = NULL;
542
	const u32 *reg;
543 544 545 546 547
	int ok;

	/* Look for the clock chip */
	while ((cc = of_find_node_by_name(cc, "i2c-hwclock")) != NULL) {
		p = of_get_parent(cc);
548
		ok = p && of_device_is_compatible(p, "uni-n-i2c");
549 550 551
		of_node_put(p);
		if (!ok)
			continue;
552

553 554 555
		pmac_tb_clock_chip_host = pmac_i2c_find_bus(cc);
		if (pmac_tb_clock_chip_host == NULL)
			continue;
556
		reg = of_get_property(cc, "reg", NULL);
557 558 559 560
		if (reg == NULL)
			continue;
		switch (*reg) {
		case 0xd2:
561
			if (of_device_is_compatible(cc,"pulsar-legacy-slewing")) {
562 563
				pmac_tb_freeze = smp_core99_pulsar_tb_freeze;
				pmac_tb_pulsar_addr = 0xd2;
564
				name = "Pulsar";
565
			} else if (of_device_is_compatible(cc, "cy28508")) {
566
				pmac_tb_freeze = smp_core99_cypress_tb_freeze;
567
				name = "Cypress";
568 569 570 571 572
			}
			break;
		case 0xd4:
			pmac_tb_freeze = smp_core99_pulsar_tb_freeze;
			pmac_tb_pulsar_addr = 0xd4;
573
			name = "Pulsar";
574 575
			break;
		}
576
		if (pmac_tb_freeze != NULL)
577 578
			break;
	}
579 580
	if (pmac_tb_freeze != NULL) {
		/* Open i2c bus for synchronous access */
581 582 583
		if (pmac_i2c_open(pmac_tb_clock_chip_host, 1)) {
			printk(KERN_ERR "Failed top open i2c bus for clock"
			       " sync, fallback to software sync !\n");
584 585 586 587 588
			goto no_i2c_sync;
		}
		printk(KERN_INFO "Processor timebase sync using %s i2c clock\n",
		       name);
		return;
589
	}
590 591
 no_i2c_sync:
	pmac_tb_freeze = NULL;
592
	pmac_tb_clock_chip_host = NULL;
593 594 595 596
}



597
/*
598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617
 * Newer G5s uses a platform function
 */

static void smp_core99_pfunc_tb_freeze(int freeze)
{
	struct device_node *cpus;
	struct pmf_args args;

	cpus = of_find_node_by_path("/cpus");
	BUG_ON(cpus == NULL);
	args.count = 1;
	args.u[0].v = !freeze;
	pmf_call_function(cpus, "cpu-timebase", &args);
	of_node_put(cpus);
}

#else /* CONFIG_PPC64 */

/*
 * SMP G4 use a GPIO to enable/disable the timebase.
618
 */
619

620
static unsigned int core99_tb_gpio;	/* Timebase freeze GPIO */
621

622
static void smp_core99_gpio_tb_freeze(int freeze)
623
{
624 625 626 627
	if (freeze)
		pmac_call_feature(PMAC_FTR_WRITE_GPIO, NULL, core99_tb_gpio, 4);
	else
		pmac_call_feature(PMAC_FTR_WRITE_GPIO, NULL, core99_tb_gpio, 0);
628
	pmac_call_feature(PMAC_FTR_READ_GPIO, NULL, core99_tb_gpio, 0);
629 630
}

631 632 633

#endif /* !CONFIG_PPC64 */

634 635 636 637 638
/* L2 and L3 cache settings to pass from CPU0 to CPU1 on G4 cpus */
volatile static long int core99_l2_cache;
volatile static long int core99_l3_cache;

static void __devinit core99_init_caches(int cpu)
639
{
640
#ifndef CONFIG_PPC64
641
	if (!cpu_has_feature(CPU_FTR_L2CR))
642
		return;
643 644 645 646 647 648 649 650 651

	if (cpu == 0) {
		core99_l2_cache = _get_L2CR();
		printk("CPU0: L2CR is %lx\n", core99_l2_cache);
	} else {
		printk("CPU%d: L2CR was %lx\n", cpu, _get_L2CR());
		_set_L2CR(0);
		_set_L2CR(core99_l2_cache);
		printk("CPU%d: L2CR set to %lx\n", cpu, core99_l2_cache);
652
	}
653 654 655 656 657 658 659 660 661 662 663 664

	if (!cpu_has_feature(CPU_FTR_L3CR))
		return;

	if (cpu == 0){
		core99_l3_cache = _get_L3CR();
		printk("CPU0: L3CR is %lx\n", core99_l3_cache);
	} else {
		printk("CPU%d: L3CR was %lx\n", cpu, _get_L3CR());
		_set_L3CR(0);
		_set_L3CR(core99_l3_cache);
		printk("CPU%d: L3CR set to %lx\n", cpu, core99_l3_cache);
665
	}
666
#endif /* !CONFIG_PPC64 */
667 668
}

669 670
static void __init smp_core99_setup(int ncpus)
{
671
#ifdef CONFIG_PPC64
672

673
	/* i2c based HW sync on some G5s */
674 675 676
	if (of_machine_is_compatible("PowerMac7,2") ||
	    of_machine_is_compatible("PowerMac7,3") ||
	    of_machine_is_compatible("RackMac3,1"))
677 678
		smp_core99_setup_i2c_hwsync(ncpus);

679
	/* pfunc based HW sync on recent G5s */
680
	if (pmac_tb_freeze == NULL) {
681 682 683
		struct device_node *cpus =
			of_find_node_by_path("/cpus");
		if (cpus &&
684
		    of_get_property(cpus, "platform-cpu-timebase", NULL)) {
685
			pmac_tb_freeze = smp_core99_pfunc_tb_freeze;
686
			printk(KERN_INFO "Processor timebase sync using"
687
			       " platform function\n");
688
		}
689 690
	}

691 692 693
#else /* CONFIG_PPC64 */

	/* GPIO based HW sync on ppc32 Core99 */
694
	if (pmac_tb_freeze == NULL && !of_machine_is_compatible("MacRISC4")) {
695
		struct device_node *cpu;
696
		const u32 *tbprop = NULL;
697 698 699 700

		core99_tb_gpio = KL_GPIO_TB_ENABLE;	/* default value */
		cpu = of_find_node_by_type(NULL, "cpu");
		if (cpu != NULL) {
701
			tbprop = of_get_property(cpu, "timebase-enable", NULL);
702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725
			if (tbprop)
				core99_tb_gpio = *tbprop;
			of_node_put(cpu);
		}
		pmac_tb_freeze = smp_core99_gpio_tb_freeze;
		printk(KERN_INFO "Processor timebase sync using"
		       " GPIO 0x%02x\n", core99_tb_gpio);
	}

#endif /* CONFIG_PPC64 */

	/* No timebase sync, fallback to software */
	if (pmac_tb_freeze == NULL) {
		smp_ops->give_timebase = smp_generic_give_timebase;
		smp_ops->take_timebase = smp_generic_take_timebase;
		printk(KERN_INFO "Processor timebase sync using software\n");
	}

#ifndef CONFIG_PPC64
	{
		int i;

		/* XXX should get this from reg properties */
		for (i = 1; i < ncpus; ++i)
726
			set_hard_smp_processor_id(i, i);
727
	}
728 729
#endif

730
	/* 32 bits SMP can't NAP */
731
	if (!of_machine_is_compatible("MacRISC4"))
732 733 734
		powersave_nap = 0;
}

735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751
static int __init smp_core99_probe(void)
{
	struct device_node *cpus;
	int ncpus = 0;

	if (ppc_md.progress) ppc_md.progress("smp_core99_probe", 0x345);

	/* Count CPUs in the device-tree */
       	for (cpus = NULL; (cpus = of_find_node_by_type(cpus, "cpu")) != NULL;)
	       	++ncpus;

	printk(KERN_INFO "PowerMac SMP probe found %d cpus\n", ncpus);

	/* Nothing more to do if less than 2 of them */
	if (ncpus <= 1)
		return 1;

752 753 754
	/* We need to perform some early initialisations before we can start
	 * setting up SMP as we are running before initcalls
	 */
755
	pmac_pfunc_base_install();
756 757 758
	pmac_i2c_init();

	/* Setup various bits like timebase sync method, ability to nap, ... */
759
	smp_core99_setup(ncpus);
760 761

	/* Install IPIs */
762
	mpic_request_ipis();
763 764

	/* Collect l2cr and l3cr values from CPU 0 */
765 766 767 768 769
	core99_init_caches(0);

	return ncpus;
}

770
static int __devinit smp_core99_kick_cpu(int nr)
771 772
{
	unsigned int save_vector;
773
	unsigned long target, flags;
774
	unsigned int *vector = (unsigned int *)(PAGE_OFFSET+0x100);
775 776

	if (nr < 0 || nr > 3)
777
		return -ENOENT;
778 779 780

	if (ppc_md.progress)
		ppc_md.progress("smp_core99_kick_cpu", 0x346);
781 782 783 784 785 786

	local_irq_save(flags);

	/* Save reset vector */
	save_vector = *vector;

787
	/* Setup fake reset vector that does
788
	 *   b __secondary_start_pmac_0 + nr*8
789
	 */
790
	target = (unsigned long) __secondary_start_pmac_0 + nr * 8;
791
	patch_branch(vector, target, BRANCH_SET_LINK);
792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808

	/* Put some life in our friend */
	pmac_call_feature(PMAC_FTR_RESET_CPU, NULL, nr, 0);

	/* FIXME: We wait a bit for the CPU to take the exception, I should
	 * instead wait for the entry code to set something for me. Well,
	 * ideally, all that crap will be done in prom.c and the CPU left
	 * in a RAM-based wait loop like CHRP.
	 */
	mdelay(1);

	/* Restore our exception vector */
	*vector = save_vector;
	flush_icache_range((unsigned long) vector, (unsigned long) vector + 4);

	local_irq_restore(flags);
	if (ppc_md.progress) ppc_md.progress("smp_core99_kick_cpu done", 0x347);
809 810

	return 0;
811 812 813 814 815 816 817 818 819 820
}

static void __devinit smp_core99_setup_cpu(int cpu_nr)
{
	/* Setup L2/L3 */
	if (cpu_nr != 0)
		core99_init_caches(cpu_nr);

	/* Setup openpic */
	mpic_setup_this_cpu();
821
}
822

823
#ifdef CONFIG_PPC64
824 825 826 827 828
#ifdef CONFIG_HOTPLUG_CPU
static int smp_core99_cpu_notify(struct notifier_block *self,
				 unsigned long action, void *hcpu)
{
	int rc;
829

830 831 832 833
	switch(action) {
	case CPU_UP_PREPARE:
	case CPU_UP_PREPARE_FROZEN:
		/* Open i2c bus if it was used for tb sync */
834
		if (pmac_tb_clock_chip_host) {
835 836 837 838 839
			rc = pmac_i2c_open(pmac_tb_clock_chip_host, 1);
			if (rc) {
				pr_err("Failed to open i2c bus for time sync\n");
				return notifier_from_errno(rc);
			}
840
		}
841 842 843 844 845 846 847 848 849 850 851 852
		break;
	case CPU_ONLINE:
	case CPU_UP_CANCELED:
		/* Close i2c bus if it was used for tb sync */
		if (pmac_tb_clock_chip_host)
			pmac_i2c_close(pmac_tb_clock_chip_host);
		break;
	default:
		break;
	}
	return NOTIFY_OK;
}
853

854 855 856 857
static struct notifier_block __cpuinitdata smp_core99_cpu_nb = {
	.notifier_call	= smp_core99_cpu_notify,
};
#endif /* CONFIG_HOTPLUG_CPU */
858

859 860 861 862 863 864 865 866 867 868 869 870 871 872 873
static void __init smp_core99_bringup_done(void)
{
	extern void g5_phy_disable_cpu1(void);

	/* Close i2c bus if it was used for tb sync */
	if (pmac_tb_clock_chip_host)
		pmac_i2c_close(pmac_tb_clock_chip_host);

	/* If we didn't start the second CPU, we must take
	 * it off the bus.
	 */
	if (of_machine_is_compatible("MacRISC4") &&
	    num_online_cpus() < 2) {
		set_cpu_present(1, false);
		g5_phy_disable_cpu1();
874
	}
875 876 877
#ifdef CONFIG_HOTPLUG_CPU
	register_cpu_notifier(&smp_core99_cpu_nb);
#endif
878

879 880 881
	if (ppc_md.progress)
		ppc_md.progress("smp_core99_bringup_done", 0x349);
}
882
#endif /* CONFIG_PPC64 */
883

884
#ifdef CONFIG_HOTPLUG_CPU
885

886
static int smp_core99_cpu_disable(void)
887
{
888 889 890
	int rc = generic_cpu_disable();
	if (rc)
		return rc;
891

892
	mpic_cpu_set_priority(0xf);
893

894 895 896
	return 0;
}

897 898 899
#ifdef CONFIG_PPC32

static void pmac_cpu_die(void)
900
{
901 902
	int cpu = smp_processor_id();

903
	local_irq_disable();
904
	idle_task_exit();
905 906
	pr_debug("CPU%d offline\n", cpu);
	generic_set_cpu_dead(cpu);
907
	smp_wmb();
908 909 910 911
	mb();
	low_cpu_die();
}

912 913 914 915
#else /* CONFIG_PPC32 */

static void pmac_cpu_die(void)
{
916 917
	int cpu = smp_processor_id();

918 919 920 921 922 923 924 925 926
	local_irq_disable();
	idle_task_exit();

	/*
	 * turn off as much as possible, we'll be
	 * kicked out as this will only be invoked
	 * on core99 platforms for now ...
	 */

927 928
	printk(KERN_INFO "CPU#%d offline\n", cpu);
	generic_set_cpu_dead(cpu);
929 930 931
	smp_wmb();

	/*
932 933 934 935 936
	 * Re-enable interrupts. The NAP code needs to enable them
	 * anyways, do it now so we deal with the case where one already
	 * happened while soft-disabled.
	 * We shouldn't get any external interrupts, only decrementer, and the
	 * decrementer handler is safe for use on offline CPUs
937
	 */
938
	local_irq_enable();
939 940 941 942 943

	while (1) {
		/* let's not take timer interrupts too often ... */
		set_dec(0x7fffffff);

944 945
		/* Enter NAP mode */
		power4_idle();
946 947 948 949 950
	}
}

#endif /* else CONFIG_PPC32 */
#endif /* CONFIG_HOTPLUG_CPU */
951 952 953 954 955

/* Core99 Macs (dual G4s and G5s) */
struct smp_ops_t core99_smp_ops = {
	.message_pass	= smp_mpic_message_pass,
	.probe		= smp_core99_probe,
956
#ifdef CONFIG_PPC64
957
	.bringup_done	= smp_core99_bringup_done,
958
#endif
959 960 961 962
	.kick_cpu	= smp_core99_kick_cpu,
	.setup_cpu	= smp_core99_setup_cpu,
	.give_timebase	= smp_core99_give_timebase,
	.take_timebase	= smp_core99_take_timebase,
963
#if defined(CONFIG_HOTPLUG_CPU)
964
	.cpu_disable	= smp_core99_cpu_disable,
965
	.cpu_die	= generic_cpu_die,
966 967
#endif
};
968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984

void __init pmac_setup_smp(void)
{
	struct device_node *np;

	/* Check for Core99 */
	np = of_find_node_by_name(NULL, "uni-n");
	if (!np)
		np = of_find_node_by_name(NULL, "u3");
	if (!np)
		np = of_find_node_by_name(NULL, "u4");
	if (np) {
		of_node_put(np);
		smp_ops = &core99_smp_ops;
	}
#ifdef CONFIG_PPC32
	else {
985
		/* We have to set bits in cpu_possible_mask here since the
986 987 988 989 990 991 992
		 * secondary CPU(s) aren't in the device tree. Various
		 * things won't be initialized for CPUs not in the possible
		 * map, so we really need to fix it up here.
		 */
		int cpu;

		for (cpu = 1; cpu < 4 && cpu < NR_CPUS; ++cpu)
993
			set_cpu_possible(cpu, true);
994 995 996
		smp_ops = &psurge_smp_ops;
	}
#endif /* CONFIG_PPC32 */
997 998 999 1000

#ifdef CONFIG_HOTPLUG_CPU
	ppc_md.cpu_die = pmac_cpu_die;
#endif
1001 1002
}

1003