io_apic.c 92.7 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3
/*
 *	Intel IO-APIC support for multi-Pentium hosts.
 *
I
Ingo Molnar 已提交
4
 *	Copyright (C) 1997, 1998, 1999, 2000, 2009 Ingo Molnar, Hajnalka Szabo
L
Linus Torvalds 已提交
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
 *
 *	Many thanks to Stig Venaas for trying out countless experimental
 *	patches and reporting/debugging problems patiently!
 *
 *	(c) 1999, Multiple IO-APIC support, developed by
 *	Ken-ichi Yaku <yaku@css1.kbnes.nec.co.jp> and
 *      Hidemi Kishimoto <kisimoto@css1.kbnes.nec.co.jp>,
 *	further tested and cleaned up by Zach Brown <zab@redhat.com>
 *	and Ingo Molnar <mingo@redhat.com>
 *
 *	Fixes
 *	Maciej W. Rozycki	:	Bits for genuine 82489DX APICs;
 *					thanks to Eric Gilmore
 *					and Rolf G. Tews
 *					for testing these extensively
 *	Paul Diefenbaugh	:	Added full ACPI support
 */

#include <linux/mm.h>
#include <linux/interrupt.h>
#include <linux/init.h>
#include <linux/delay.h>
#include <linux/sched.h>
28
#include <linux/pci.h>
L
Linus Torvalds 已提交
29 30 31
#include <linux/mc146818rtc.h>
#include <linux/compiler.h>
#include <linux/acpi.h>
32
#include <linux/module.h>
33
#include <linux/syscore_ops.h>
34
#include <linux/msi.h>
35
#include <linux/htirq.h>
36
#include <linux/freezer.h>
37
#include <linux/kthread.h>
38
#include <linux/jiffies.h>	/* time_after() */
39
#include <linux/slab.h>
40 41
#include <linux/bootmem.h>
#include <linux/dmar.h>
42
#include <linux/hpet.h>
43

44
#include <asm/idle.h>
L
Linus Torvalds 已提交
45 46
#include <asm/io.h>
#include <asm/smp.h>
47
#include <asm/cpu.h>
L
Linus Torvalds 已提交
48
#include <asm/desc.h>
49 50 51
#include <asm/proto.h>
#include <asm/acpi.h>
#include <asm/dma.h>
L
Linus Torvalds 已提交
52
#include <asm/timer.h>
53
#include <asm/i8259.h>
54
#include <asm/msidef.h>
55
#include <asm/hypertransport.h>
56
#include <asm/setup.h>
57
#include <asm/irq_remapping.h>
58
#include <asm/hpet.h>
59
#include <asm/hw_irq.h>
L
Linus Torvalds 已提交
60

I
Ingo Molnar 已提交
61
#include <asm/apic.h>
L
Linus Torvalds 已提交
62

63
#define __apicdebuginit(type) static type __init
64

65 66
#define for_each_irq_pin(entry, head) \
	for (entry = head; entry; entry = entry->next)
67

L
Linus Torvalds 已提交
68
/*
69 70
 *      Is the SiS APIC rmw bug present ?
 *      -1 = don't know, 0 = no, 1 = yes
L
Linus Torvalds 已提交
71 72 73
 */
int sis_apic_bug = -1;

74 75
static DEFINE_RAW_SPINLOCK(ioapic_lock);
static DEFINE_RAW_SPINLOCK(vector_lock);
Y
Yinghai Lu 已提交
76

S
Suresh Siddha 已提交
77 78 79 80 81
static struct ioapic {
	/*
	 * # of IRQ routing registers
	 */
	int nr_registers;
82 83 84 85
	/*
	 * Saved state during suspend/resume, or while enabling intr-remap.
	 */
	struct IO_APIC_route_entry *saved_registers;
86 87
	/* I/O APIC config */
	struct mpc_ioapic mp_config;
88 89
	/* IO APIC gsi routing info */
	struct mp_ioapic_gsi  gsi_config;
90
	DECLARE_BITMAP(pin_programmed, MP_MAX_IOAPIC_PIN + 1);
S
Suresh Siddha 已提交
91
} ioapics[MAX_IO_APICS];
L
Linus Torvalds 已提交
92

93
#define mpc_ioapic_ver(ioapic_idx)	ioapics[ioapic_idx].mp_config.apicver
94

95
int mpc_ioapic_id(int ioapic_idx)
96
{
97
	return ioapics[ioapic_idx].mp_config.apicid;
98 99
}

100
unsigned int mpc_ioapic_addr(int ioapic_idx)
101
{
102
	return ioapics[ioapic_idx].mp_config.apicaddr;
103 104
}

105
struct mp_ioapic_gsi *mp_ioapic_gsi_routing(int ioapic_idx)
106
{
107
	return &ioapics[ioapic_idx].gsi_config;
108
}
109

110
int nr_ioapics;
111

112 113
/* The one past the highest gsi number used */
u32 gsi_top;
114

115
/* MP IRQ source entries */
116
struct mpc_intsrc mp_irqs[MAX_IRQ_SOURCES];
117 118 119 120

/* # of MP IRQ source entries */
int mp_irq_entries;

121 122 123
/* GSI interrupts */
static int nr_irqs_gsi = NR_IRQS_LEGACY;

124
#ifdef CONFIG_EISA
125 126 127 128 129
int mp_bus_id_to_type[MAX_MP_BUSSES];
#endif

DECLARE_BITMAP(mp_bus_not_pci, MAX_MP_BUSSES);

Y
Yinghai Lu 已提交
130 131
int skip_ioapic_setup;

132 133 134 135
/**
 * disable_ioapic_support() - disables ioapic support at runtime
 */
void disable_ioapic_support(void)
136 137 138 139 140 141 142 143
{
#ifdef CONFIG_PCI
	noioapicquirk = 1;
	noioapicreroute = -1;
#endif
	skip_ioapic_setup = 1;
}

144
static int __init parse_noapic(char *str)
Y
Yinghai Lu 已提交
145 146
{
	/* disable IO-APIC */
147
	disable_ioapic_support();
Y
Yinghai Lu 已提交
148 149 150
	return 0;
}
early_param("noapic", parse_noapic);
151

152 153
static int io_apic_setup_irq_pin(unsigned int irq, int node,
				 struct io_apic_irq_attr *attr);
154

155 156 157 158 159 160 161 162 163 164 165
/* Will be called in mpparse/acpi/sfi codes for saving IRQ info */
void mp_save_irq(struct mpc_intsrc *m)
{
	int i;

	apic_printk(APIC_VERBOSE, "Int: type %d, pol %d, trig %d, bus %02x,"
		" IRQ %02x, APIC ID %x, APIC INT %02x\n",
		m->irqtype, m->irqflag & 3, (m->irqflag >> 2) & 3, m->srcbus,
		m->srcbusirq, m->dstapic, m->dstirq);

	for (i = 0; i < mp_irq_entries; i++) {
166
		if (!memcmp(&mp_irqs[i], m, sizeof(*m)))
167 168 169
			return;
	}

170
	memcpy(&mp_irqs[mp_irq_entries], m, sizeof(*m));
171 172 173 174
	if (++mp_irq_entries == MAX_IRQ_SOURCES)
		panic("Max # of irq sources exceeded!!\n");
}

175 176 177 178 179
struct irq_pin_list {
	int apic, pin;
	struct irq_pin_list *next;
};

T
Thomas Gleixner 已提交
180
static struct irq_pin_list *alloc_irq_pin_list(int node)
181
{
182
	return kzalloc_node(sizeof(struct irq_pin_list), GFP_KERNEL, node);
183 184
}

185

Y
Yinghai Lu 已提交
186
/* irq_cfg is indexed by the sum of all RTEs in all I/O APICs. */
187
static struct irq_cfg irq_cfgx[NR_IRQS_LEGACY];
Y
Yinghai Lu 已提交
188

189
int __init arch_early_irq_init(void)
190
{
191
	struct irq_cfg *cfg;
192
	int count, node, i;
T
Thomas Gleixner 已提交
193

194
	if (!legacy_pic->nr_legacy_irqs)
195 196
		io_apic_irqs = ~0UL;

197
	for (i = 0; i < nr_ioapics; i++) {
198
		ioapics[i].saved_registers =
199
			kzalloc(sizeof(struct IO_APIC_route_entry) *
S
Suresh Siddha 已提交
200
				ioapics[i].nr_registers, GFP_KERNEL);
201
		if (!ioapics[i].saved_registers)
202 203 204
			pr_err("IOAPIC %d: suspend/resume impossible!\n", i);
	}

205 206
	cfg = irq_cfgx;
	count = ARRAY_SIZE(irq_cfgx);
207
	node = cpu_to_node(0);
208

209
	for (i = 0; i < count; i++) {
210
		irq_set_chip_data(i, &cfg[i]);
211 212
		zalloc_cpumask_var_node(&cfg[i].domain, GFP_KERNEL, node);
		zalloc_cpumask_var_node(&cfg[i].old_domain, GFP_KERNEL, node);
213 214
		/*
		 * For legacy IRQ's, start with assigning irq0 to irq15 to
215
		 * IRQ0_VECTOR to IRQ15_VECTOR for all cpu's.
216
		 */
217
		if (i < legacy_pic->nr_legacy_irqs) {
218
			cfg[i].vector = IRQ0_VECTOR + i;
219
			cpumask_setall(cfg[i].domain);
220
		}
221
	}
222 223

	return 0;
224
}
225

226
static struct irq_cfg *irq_cfg(unsigned int irq)
227
{
228
	return irq_get_chip_data(irq);
229
}
T
Thomas Gleixner 已提交
230

231
static struct irq_cfg *alloc_irq_cfg(unsigned int irq, int node)
232
{
233
	struct irq_cfg *cfg;
234

235
	cfg = kzalloc_node(sizeof(*cfg), GFP_KERNEL, node);
236 237
	if (!cfg)
		return NULL;
238
	if (!zalloc_cpumask_var_node(&cfg->domain, GFP_KERNEL, node))
239
		goto out_cfg;
240
	if (!zalloc_cpumask_var_node(&cfg->old_domain, GFP_KERNEL, node))
241
		goto out_domain;
242
	return cfg;
243 244 245 246 247
out_domain:
	free_cpumask_var(cfg->domain);
out_cfg:
	kfree(cfg);
	return NULL;
248 249
}

250
static void free_irq_cfg(unsigned int at, struct irq_cfg *cfg)
251
{
252 253
	if (!cfg)
		return;
254
	irq_set_chip_data(at, NULL);
255 256 257 258 259 260 261 262 263 264 265 266 267
	free_cpumask_var(cfg->domain);
	free_cpumask_var(cfg->old_domain);
	kfree(cfg);
}

static struct irq_cfg *alloc_irq_and_cfg_at(unsigned int at, int node)
{
	int res = irq_alloc_desc_at(at, node);
	struct irq_cfg *cfg;

	if (res < 0) {
		if (res != -EEXIST)
			return NULL;
268
		cfg = irq_get_chip_data(at);
269 270 271 272
		if (cfg)
			return cfg;
	}

273
	cfg = alloc_irq_cfg(at, node);
274
	if (cfg)
275
		irq_set_chip_data(at, cfg);
276 277 278 279 280
	else
		irq_free_desc(at);
	return cfg;
}

L
Linus Torvalds 已提交
281 282 283 284
struct io_apic {
	unsigned int index;
	unsigned int unused[3];
	unsigned int data;
285 286
	unsigned int unused2[11];
	unsigned int eoi;
L
Linus Torvalds 已提交
287 288 289 290 291
};

static __attribute_const__ struct io_apic __iomem *io_apic_base(int idx)
{
	return (void __iomem *) __fix_to_virt(FIX_IO_APIC_BASE_0 + idx)
292
		+ (mpc_ioapic_addr(idx) & ~PAGE_MASK);
L
Linus Torvalds 已提交
293 294
}

295
void io_apic_eoi(unsigned int apic, unsigned int vector)
296 297 298 299 300
{
	struct io_apic __iomem *io_apic = io_apic_base(apic);
	writel(vector, &io_apic->eoi);
}

301
unsigned int native_io_apic_read(unsigned int apic, unsigned int reg)
L
Linus Torvalds 已提交
302 303 304 305 306 307
{
	struct io_apic __iomem *io_apic = io_apic_base(apic);
	writel(reg, &io_apic->index);
	return readl(&io_apic->data);
}

308
void native_io_apic_write(unsigned int apic, unsigned int reg, unsigned int value)
L
Linus Torvalds 已提交
309 310
{
	struct io_apic __iomem *io_apic = io_apic_base(apic);
311

L
Linus Torvalds 已提交
312 313 314 315 316 317 318 319 320 321
	writel(reg, &io_apic->index);
	writel(value, &io_apic->data);
}

/*
 * Re-write a value: to be used for read-modify-write
 * cycles where the read already set up the index register.
 *
 * Older SiS APIC requires we rewrite the index register
 */
322
void native_io_apic_modify(unsigned int apic, unsigned int reg, unsigned int value)
L
Linus Torvalds 已提交
323
{
324
	struct io_apic __iomem *io_apic = io_apic_base(apic);
T
Thomas Gleixner 已提交
325 326 327

	if (sis_apic_bug)
		writel(reg, &io_apic->index);
L
Linus Torvalds 已提交
328 329 330
	writel(value, &io_apic->data);
}

331 332 333 334 335
union entry_union {
	struct { u32 w1, w2; };
	struct IO_APIC_route_entry entry;
};

336 337 338 339 340 341
static struct IO_APIC_route_entry __ioapic_read_entry(int apic, int pin)
{
	union entry_union eu;

	eu.w1 = io_apic_read(apic, 0x10 + 2 * pin);
	eu.w2 = io_apic_read(apic, 0x11 + 2 * pin);
342

343 344 345
	return eu.entry;
}

346 347 348 349
static struct IO_APIC_route_entry ioapic_read_entry(int apic, int pin)
{
	union entry_union eu;
	unsigned long flags;
350

351
	raw_spin_lock_irqsave(&ioapic_lock, flags);
352
	eu.entry = __ioapic_read_entry(apic, pin);
353
	raw_spin_unlock_irqrestore(&ioapic_lock, flags);
354

355 356 357
	return eu.entry;
}

358 359 360 361 362 363
/*
 * When we write a new IO APIC routing entry, we need to write the high
 * word first! If the mask bit in the low word is clear, we will enable
 * the interrupt, and we need to make sure the entry is fully populated
 * before that happens.
 */
364
static void __ioapic_write_entry(int apic, int pin, struct IO_APIC_route_entry e)
365
{
366 367
	union entry_union eu = {{0, 0}};

368
	eu.entry = e;
369 370
	io_apic_write(apic, 0x11 + 2*pin, eu.w2);
	io_apic_write(apic, 0x10 + 2*pin, eu.w1);
371 372
}

373
static void ioapic_write_entry(int apic, int pin, struct IO_APIC_route_entry e)
374 375
{
	unsigned long flags;
376

377
	raw_spin_lock_irqsave(&ioapic_lock, flags);
378
	__ioapic_write_entry(apic, pin, e);
379
	raw_spin_unlock_irqrestore(&ioapic_lock, flags);
380 381 382 383 384 385 386 387 388 389 390 391
}

/*
 * When we mask an IO APIC routing entry, we need to write the low
 * word first, in order to set the mask bit before we change the
 * high bits!
 */
static void ioapic_mask_entry(int apic, int pin)
{
	unsigned long flags;
	union entry_union eu = { .entry.mask = 1 };

392
	raw_spin_lock_irqsave(&ioapic_lock, flags);
393 394
	io_apic_write(apic, 0x10 + 2*pin, eu.w1);
	io_apic_write(apic, 0x11 + 2*pin, eu.w2);
395
	raw_spin_unlock_irqrestore(&ioapic_lock, flags);
396 397
}

L
Linus Torvalds 已提交
398 399 400 401 402
/*
 * The common case is 1:1 IRQ<->pin mappings. Sometimes there are
 * shared ISA-space IRQs, so we have to support them. We are super
 * fast in the common case, and fast for shared ISA-space IRQs.
 */
403
static int __add_pin_to_irq_node(struct irq_cfg *cfg, int node, int apic, int pin)
L
Linus Torvalds 已提交
404
{
405
	struct irq_pin_list **last, *entry;
406

407 408 409
	/* don't allow duplicates */
	last = &cfg->irq_2_pin;
	for_each_irq_pin(entry, cfg->irq_2_pin) {
410
		if (entry->apic == apic && entry->pin == pin)
411
			return 0;
412
		last = &entry->next;
L
Linus Torvalds 已提交
413
	}
414

T
Thomas Gleixner 已提交
415
	entry = alloc_irq_pin_list(node);
416
	if (!entry) {
417 418
		pr_err("can not alloc irq_pin_list (%d,%d,%d)\n",
		       node, apic, pin);
419
		return -ENOMEM;
420
	}
L
Linus Torvalds 已提交
421 422
	entry->apic = apic;
	entry->pin = pin;
423

424
	*last = entry;
425 426 427 428 429
	return 0;
}

static void add_pin_to_irq_node(struct irq_cfg *cfg, int node, int apic, int pin)
{
T
Thomas Gleixner 已提交
430
	if (__add_pin_to_irq_node(cfg, node, apic, pin))
431
		panic("IO-APIC: failed to add irq-pin. Can not proceed\n");
L
Linus Torvalds 已提交
432 433 434 435 436
}

/*
 * Reroute an IRQ to a different pin.
 */
437
static void __init replace_pin_at_irq_node(struct irq_cfg *cfg, int node,
438 439
					   int oldapic, int oldpin,
					   int newapic, int newpin)
L
Linus Torvalds 已提交
440
{
441
	struct irq_pin_list *entry;
L
Linus Torvalds 已提交
442

443
	for_each_irq_pin(entry, cfg->irq_2_pin) {
L
Linus Torvalds 已提交
444 445 446
		if (entry->apic == oldapic && entry->pin == oldpin) {
			entry->apic = newapic;
			entry->pin = newpin;
447
			/* every one is different, right? */
448
			return;
449
		}
L
Linus Torvalds 已提交
450
	}
451

452 453
	/* old apic/pin didn't exist, so just add new ones */
	add_pin_to_irq_node(cfg, node, newapic, newpin);
L
Linus Torvalds 已提交
454 455
}

456 457 458 459 460 461 462 463 464 465 466 467 468 469 470
static void __io_apic_modify_irq(struct irq_pin_list *entry,
				 int mask_and, int mask_or,
				 void (*final)(struct irq_pin_list *entry))
{
	unsigned int reg, pin;

	pin = entry->pin;
	reg = io_apic_read(entry->apic, 0x10 + pin * 2);
	reg &= mask_and;
	reg |= mask_or;
	io_apic_modify(entry->apic, 0x10 + pin * 2, reg);
	if (final)
		final(entry);
}

471 472 473
static void io_apic_modify_irq(struct irq_cfg *cfg,
			       int mask_and, int mask_or,
			       void (*final)(struct irq_pin_list *entry))
474 475
{
	struct irq_pin_list *entry;
476

477 478 479 480
	for_each_irq_pin(entry, cfg->irq_2_pin)
		__io_apic_modify_irq(entry, mask_and, mask_or, final);
}

481
static void io_apic_sync(struct irq_pin_list *entry)
L
Linus Torvalds 已提交
482
{
483 484 485 486 487
	/*
	 * Synchronize the IO-APIC and the CPU by doing
	 * a dummy read from the IO-APIC
	 */
	struct io_apic __iomem *io_apic;
488

489
	io_apic = io_apic_base(entry->apic);
Y
Yinghai Lu 已提交
490
	readl(&io_apic->data);
L
Linus Torvalds 已提交
491 492
}

T
Thomas Gleixner 已提交
493
static void mask_ioapic(struct irq_cfg *cfg)
494
{
T
Thomas Gleixner 已提交
495 496 497
	unsigned long flags;

	raw_spin_lock_irqsave(&ioapic_lock, flags);
Y
Yinghai Lu 已提交
498
	io_apic_modify_irq(cfg, ~0, IO_APIC_REDIR_MASKED, &io_apic_sync);
T
Thomas Gleixner 已提交
499
	raw_spin_unlock_irqrestore(&ioapic_lock, flags);
500
}
L
Linus Torvalds 已提交
501

502
static void mask_ioapic_irq(struct irq_data *data)
L
Linus Torvalds 已提交
503
{
504
	mask_ioapic(data->chip_data);
T
Thomas Gleixner 已提交
505
}
Y
Yinghai Lu 已提交
506

T
Thomas Gleixner 已提交
507 508 509
static void __unmask_ioapic(struct irq_cfg *cfg)
{
	io_apic_modify_irq(cfg, ~IO_APIC_REDIR_MASKED, 0, NULL);
L
Linus Torvalds 已提交
510 511
}

T
Thomas Gleixner 已提交
512
static void unmask_ioapic(struct irq_cfg *cfg)
L
Linus Torvalds 已提交
513 514 515
{
	unsigned long flags;

516
	raw_spin_lock_irqsave(&ioapic_lock, flags);
T
Thomas Gleixner 已提交
517
	__unmask_ioapic(cfg);
518
	raw_spin_unlock_irqrestore(&ioapic_lock, flags);
L
Linus Torvalds 已提交
519 520
}

521
static void unmask_ioapic_irq(struct irq_data *data)
Y
Yinghai Lu 已提交
522
{
523
	unmask_ioapic(data->chip_data);
Y
Yinghai Lu 已提交
524 525
}

526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541
/*
 * IO-APIC versions below 0x20 don't support EOI register.
 * For the record, here is the information about various versions:
 *     0Xh     82489DX
 *     1Xh     I/OAPIC or I/O(x)APIC which are not PCI 2.2 Compliant
 *     2Xh     I/O(x)APIC which is PCI 2.2 Compliant
 *     30h-FFh Reserved
 *
 * Some of the Intel ICH Specs (ICH2 to ICH5) documents the io-apic
 * version as 0x2. This is an error with documentation and these ICH chips
 * use io-apic's of version 0x20.
 *
 * For IO-APIC's with EOI register, we use that to do an explicit EOI.
 * Otherwise, we simulate the EOI message manually by changing the trigger
 * mode to edge and then back to level, with RTE being masked during this.
 */
542
void native_eoi_ioapic_pin(int apic, int pin, int vector)
543 544
{
	if (mpc_ioapic_ver(apic) >= 0x20) {
545
		io_apic_eoi(apic, vector);
546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565
	} else {
		struct IO_APIC_route_entry entry, entry1;

		entry = entry1 = __ioapic_read_entry(apic, pin);

		/*
		 * Mask the entry and change the trigger mode to edge.
		 */
		entry1.mask = 1;
		entry1.trigger = IOAPIC_EDGE;

		__ioapic_write_entry(apic, pin, entry1);

		/*
		 * Restore the previous level triggered entry.
		 */
		__ioapic_write_entry(apic, pin, entry);
	}
}

566
void eoi_ioapic_irq(unsigned int irq, struct irq_cfg *cfg)
567 568 569 570 571 572
{
	struct irq_pin_list *entry;
	unsigned long flags;

	raw_spin_lock_irqsave(&ioapic_lock, flags);
	for_each_irq_pin(entry, cfg->irq_2_pin)
573 574
		x86_io_apic_ops.eoi_ioapic_pin(entry->apic, entry->pin,
					       cfg->vector);
575 576 577
	raw_spin_unlock_irqrestore(&ioapic_lock, flags);
}

L
Linus Torvalds 已提交
578 579 580
static void clear_IO_APIC_pin(unsigned int apic, unsigned int pin)
{
	struct IO_APIC_route_entry entry;
581

L
Linus Torvalds 已提交
582
	/* Check delivery_mode to be sure we're not clearing an SMI pin */
583
	entry = ioapic_read_entry(apic, pin);
L
Linus Torvalds 已提交
584 585
	if (entry.delivery_mode == dest_SMI)
		return;
586

L
Linus Torvalds 已提交
587
	/*
588 589 590 591 592 593 594 595 596 597
	 * Make sure the entry is masked and re-read the contents to check
	 * if it is a level triggered pin and if the remote-IRR is set.
	 */
	if (!entry.mask) {
		entry.mask = 1;
		ioapic_write_entry(apic, pin, entry);
		entry = ioapic_read_entry(apic, pin);
	}

	if (entry.irr) {
598 599
		unsigned long flags;

600 601 602 603 604 605 606 607 608 609
		/*
		 * Make sure the trigger mode is set to level. Explicit EOI
		 * doesn't clear the remote-IRR if the trigger mode is not
		 * set to level.
		 */
		if (!entry.trigger) {
			entry.trigger = IOAPIC_LEVEL;
			ioapic_write_entry(apic, pin, entry);
		}

610
		raw_spin_lock_irqsave(&ioapic_lock, flags);
611
		x86_io_apic_ops.eoi_ioapic_pin(apic, pin, entry.vector);
612
		raw_spin_unlock_irqrestore(&ioapic_lock, flags);
613 614 615 616 617
	}

	/*
	 * Clear the rest of the bits in the IO-APIC RTE except for the mask
	 * bit.
L
Linus Torvalds 已提交
618
	 */
619
	ioapic_mask_entry(apic, pin);
620 621
	entry = ioapic_read_entry(apic, pin);
	if (entry.irr)
622
		pr_err("Unable to reset IRR for apic: %d, pin :%d\n",
623
		       mpc_ioapic_id(apic), pin);
L
Linus Torvalds 已提交
624 625
}

626
static void clear_IO_APIC (void)
L
Linus Torvalds 已提交
627 628 629 630
{
	int apic, pin;

	for (apic = 0; apic < nr_ioapics; apic++)
S
Suresh Siddha 已提交
631
		for (pin = 0; pin < ioapics[apic].nr_registers; pin++)
L
Linus Torvalds 已提交
632 633 634
			clear_IO_APIC_pin(apic, pin);
}

635
#ifdef CONFIG_X86_32
L
Linus Torvalds 已提交
636 637 638 639 640 641
/*
 * support for broken MP BIOSs, enables hand-redirection of PIRQ0-7 to
 * specific CPU-side IRQs.
 */

#define MAX_PIRQS 8
Y
Yinghai Lu 已提交
642 643 644
static int pirq_entries[MAX_PIRQS] = {
	[0 ... MAX_PIRQS - 1] = -1
};
L
Linus Torvalds 已提交
645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670

static int __init ioapic_pirq_setup(char *str)
{
	int i, max;
	int ints[MAX_PIRQS+1];

	get_options(str, ARRAY_SIZE(ints), ints);

	apic_printk(APIC_VERBOSE, KERN_INFO
			"PIRQ redirection, working around broken MP-BIOS.\n");
	max = MAX_PIRQS;
	if (ints[0] < MAX_PIRQS)
		max = ints[0];

	for (i = 0; i < max; i++) {
		apic_printk(APIC_VERBOSE, KERN_DEBUG
				"... PIRQ%d -> IRQ %d\n", i, ints[i+1]);
		/*
		 * PIRQs are mapped upside down, usually.
		 */
		pirq_entries[MAX_PIRQS-i-1] = ints[i+1];
	}
	return 1;
}

__setup("pirq=", ioapic_pirq_setup);
671 672 673
#endif /* CONFIG_X86_32 */

/*
674
 * Saves all the IO-APIC RTE's
675
 */
676
int save_ioapic_entries(void)
677 678
{
	int apic, pin;
679
	int err = 0;
680 681

	for (apic = 0; apic < nr_ioapics; apic++) {
682
		if (!ioapics[apic].saved_registers) {
683 684 685
			err = -ENOMEM;
			continue;
		}
686

S
Suresh Siddha 已提交
687
		for (pin = 0; pin < ioapics[apic].nr_registers; pin++)
688
			ioapics[apic].saved_registers[pin] =
689
				ioapic_read_entry(apic, pin);
690
	}
691

692
	return err;
693 694
}

695 696 697
/*
 * Mask all IO APIC entries.
 */
698
void mask_ioapic_entries(void)
699 700 701 702
{
	int apic, pin;

	for (apic = 0; apic < nr_ioapics; apic++) {
703
		if (!ioapics[apic].saved_registers)
704
			continue;
705

S
Suresh Siddha 已提交
706
		for (pin = 0; pin < ioapics[apic].nr_registers; pin++) {
707 708
			struct IO_APIC_route_entry entry;

709
			entry = ioapics[apic].saved_registers[pin];
710 711 712 713 714 715 716 717
			if (!entry.mask) {
				entry.mask = 1;
				ioapic_write_entry(apic, pin, entry);
			}
		}
	}
}

718
/*
719
 * Restore IO APIC entries which was saved in the ioapic structure.
720
 */
721
int restore_ioapic_entries(void)
722 723 724
{
	int apic, pin;

725
	for (apic = 0; apic < nr_ioapics; apic++) {
726
		if (!ioapics[apic].saved_registers)
727
			continue;
728

S
Suresh Siddha 已提交
729
		for (pin = 0; pin < ioapics[apic].nr_registers; pin++)
730
			ioapic_write_entry(apic, pin,
731
					   ioapics[apic].saved_registers[pin]);
732
	}
733
	return 0;
734 735
}

L
Linus Torvalds 已提交
736 737 738
/*
 * Find the IRQ entry number of a certain pin.
 */
739
static int find_irq_entry(int ioapic_idx, int pin, int type)
L
Linus Torvalds 已提交
740 741 742 743
{
	int i;

	for (i = 0; i < mp_irq_entries; i++)
744
		if (mp_irqs[i].irqtype == type &&
745
		    (mp_irqs[i].dstapic == mpc_ioapic_id(ioapic_idx) ||
746 747
		     mp_irqs[i].dstapic == MP_APIC_ALL) &&
		    mp_irqs[i].dstirq == pin)
L
Linus Torvalds 已提交
748 749 750 751 752 753 754 755
			return i;

	return -1;
}

/*
 * Find the pin to which IRQ[irq] (ISA) is connected
 */
756
static int __init find_isa_irq_pin(int irq, int type)
L
Linus Torvalds 已提交
757 758 759 760
{
	int i;

	for (i = 0; i < mp_irq_entries; i++) {
761
		int lbus = mp_irqs[i].srcbus;
L
Linus Torvalds 已提交
762

A
Alexey Starikovskiy 已提交
763
		if (test_bit(lbus, mp_bus_not_pci) &&
764 765
		    (mp_irqs[i].irqtype == type) &&
		    (mp_irqs[i].srcbusirq == irq))
L
Linus Torvalds 已提交
766

767
			return mp_irqs[i].dstirq;
L
Linus Torvalds 已提交
768 769 770 771
	}
	return -1;
}

772 773 774 775 776
static int __init find_isa_irq_apic(int irq, int type)
{
	int i;

	for (i = 0; i < mp_irq_entries; i++) {
777
		int lbus = mp_irqs[i].srcbus;
778

A
Alexey Starikovskiy 已提交
779
		if (test_bit(lbus, mp_bus_not_pci) &&
780 781
		    (mp_irqs[i].irqtype == type) &&
		    (mp_irqs[i].srcbusirq == irq))
782 783
			break;
	}
784

785
	if (i < mp_irq_entries) {
786 787 788 789 790
		int ioapic_idx;

		for (ioapic_idx = 0; ioapic_idx < nr_ioapics; ioapic_idx++)
			if (mpc_ioapic_id(ioapic_idx) == mp_irqs[i].dstapic)
				return ioapic_idx;
791 792 793 794 795
	}

	return -1;
}

796
#ifdef CONFIG_EISA
L
Linus Torvalds 已提交
797 798 799 800 801
/*
 * EISA Edge/Level control register, ELCR
 */
static int EISA_ELCR(unsigned int irq)
{
802
	if (irq < legacy_pic->nr_legacy_irqs) {
L
Linus Torvalds 已提交
803 804 805 806 807 808 809
		unsigned int port = 0x4d0 + (irq >> 3);
		return (inb(port) >> (irq & 7)) & 1;
	}
	apic_printk(APIC_VERBOSE, KERN_INFO
			"Broken MPtable reports ISA irq %d\n", irq);
	return 0;
}
810

811
#endif
L
Linus Torvalds 已提交
812

A
Alexey Starikovskiy 已提交
813 814 815 816 817 818
/* ISA interrupts are always polarity zero edge triggered,
 * when listed as conforming in the MP table. */

#define default_ISA_trigger(idx)	(0)
#define default_ISA_polarity(idx)	(0)

L
Linus Torvalds 已提交
819 820 821 822 823
/* EISA interrupts are always polarity zero and can be edge or level
 * trigger depending on the ELCR value.  If an interrupt is listed as
 * EISA conforming in the MP table, that means its trigger type must
 * be read in from the ELCR */

824
#define default_EISA_trigger(idx)	(EISA_ELCR(mp_irqs[idx].srcbusirq))
A
Alexey Starikovskiy 已提交
825
#define default_EISA_polarity(idx)	default_ISA_polarity(idx)
L
Linus Torvalds 已提交
826 827 828 829 830 831 832

/* PCI interrupts are always polarity one level triggered,
 * when listed as conforming in the MP table. */

#define default_PCI_trigger(idx)	(1)
#define default_PCI_polarity(idx)	(1)

833
static int irq_polarity(int idx)
L
Linus Torvalds 已提交
834
{
835
	int bus = mp_irqs[idx].srcbus;
L
Linus Torvalds 已提交
836 837 838 839 840
	int polarity;

	/*
	 * Determine IRQ line polarity (high active or low active):
	 */
841
	switch (mp_irqs[idx].irqflag & 3)
842
	{
843 844 845 846 847 848 849 850 851 852 853 854 855
		case 0: /* conforms, ie. bus-type dependent polarity */
			if (test_bit(bus, mp_bus_not_pci))
				polarity = default_ISA_polarity(idx);
			else
				polarity = default_PCI_polarity(idx);
			break;
		case 1: /* high active */
		{
			polarity = 0;
			break;
		}
		case 2: /* reserved */
		{
856
			pr_warn("broken BIOS!!\n");
857 858 859 860 861 862 863 864 865 866
			polarity = 1;
			break;
		}
		case 3: /* low active */
		{
			polarity = 1;
			break;
		}
		default: /* invalid */
		{
867
			pr_warn("broken BIOS!!\n");
868 869 870
			polarity = 1;
			break;
		}
L
Linus Torvalds 已提交
871 872 873 874
	}
	return polarity;
}

875
static int irq_trigger(int idx)
L
Linus Torvalds 已提交
876
{
877
	int bus = mp_irqs[idx].srcbus;
L
Linus Torvalds 已提交
878 879 880 881 882
	int trigger;

	/*
	 * Determine IRQ trigger mode (edge or level sensitive):
	 */
883
	switch ((mp_irqs[idx].irqflag>>2) & 3)
L
Linus Torvalds 已提交
884
	{
885 886 887 888 889
		case 0: /* conforms, ie. bus-type dependent */
			if (test_bit(bus, mp_bus_not_pci))
				trigger = default_ISA_trigger(idx);
			else
				trigger = default_PCI_trigger(idx);
890
#ifdef CONFIG_EISA
891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908
			switch (mp_bus_id_to_type[bus]) {
				case MP_BUS_ISA: /* ISA pin */
				{
					/* set before the switch */
					break;
				}
				case MP_BUS_EISA: /* EISA pin */
				{
					trigger = default_EISA_trigger(idx);
					break;
				}
				case MP_BUS_PCI: /* PCI pin */
				{
					/* set before the switch */
					break;
				}
				default:
				{
909
					pr_warn("broken BIOS!!\n");
910 911 912 913 914
					trigger = 1;
					break;
				}
			}
#endif
L
Linus Torvalds 已提交
915
			break;
916
		case 1: /* edge */
L
Linus Torvalds 已提交
917
		{
918
			trigger = 0;
L
Linus Torvalds 已提交
919 920
			break;
		}
921
		case 2: /* reserved */
L
Linus Torvalds 已提交
922
		{
923
			pr_warn("broken BIOS!!\n");
924
			trigger = 1;
L
Linus Torvalds 已提交
925 926
			break;
		}
927
		case 3: /* level */
L
Linus Torvalds 已提交
928
		{
929
			trigger = 1;
L
Linus Torvalds 已提交
930 931
			break;
		}
932
		default: /* invalid */
L
Linus Torvalds 已提交
933
		{
934
			pr_warn("broken BIOS!!\n");
935
			trigger = 0;
L
Linus Torvalds 已提交
936 937 938 939 940 941 942 943
			break;
		}
	}
	return trigger;
}

static int pin_2_irq(int idx, int apic, int pin)
{
944
	int irq;
945
	int bus = mp_irqs[idx].srcbus;
946
	struct mp_ioapic_gsi *gsi_cfg = mp_ioapic_gsi_routing(apic);
L
Linus Torvalds 已提交
947 948 949 950

	/*
	 * Debugging check, we are in big trouble if this message pops up!
	 */
951
	if (mp_irqs[idx].dstirq != pin)
952
		pr_err("broken BIOS or MPTABLE parser, ayiee!!\n");
L
Linus Torvalds 已提交
953

954
	if (test_bit(bus, mp_bus_not_pci)) {
955
		irq = mp_irqs[idx].srcbusirq;
956
	} else {
957
		u32 gsi = gsi_cfg->gsi_base + pin;
958 959 960 961

		if (gsi >= NR_IRQS_LEGACY)
			irq = gsi;
		else
962
			irq = gsi_top + gsi;
L
Linus Torvalds 已提交
963 964
	}

965
#ifdef CONFIG_X86_32
L
Linus Torvalds 已提交
966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981
	/*
	 * PCI IRQ command line redirection. Yes, limits are hardcoded.
	 */
	if ((pin >= 16) && (pin <= 23)) {
		if (pirq_entries[pin-16] != -1) {
			if (!pirq_entries[pin-16]) {
				apic_printk(APIC_VERBOSE, KERN_DEBUG
						"disabling PIRQ%d\n", pin-16);
			} else {
				irq = pirq_entries[pin-16];
				apic_printk(APIC_VERBOSE, KERN_DEBUG
						"using PIRQ%d -> IRQ %d\n",
						pin-16, irq);
			}
		}
	}
982 983
#endif

L
Linus Torvalds 已提交
984 985 986
	return irq;
}

987 988 989 990 991
/*
 * Find a specific PCI IRQ entry.
 * Not an __init, possibly needed by modules
 */
int IO_APIC_get_PCI_irq_vector(int bus, int slot, int pin,
992
				struct io_apic_irq_attr *irq_attr)
993
{
994
	int ioapic_idx, i, best_guess = -1;
995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006

	apic_printk(APIC_DEBUG,
		    "querying PCI -> IRQ mapping bus:%d, slot:%d, pin:%d.\n",
		    bus, slot, pin);
	if (test_bit(bus, mp_bus_not_pci)) {
		apic_printk(APIC_VERBOSE,
			    "PCI BIOS passed nonexistent PCI bus %d!\n", bus);
		return -1;
	}
	for (i = 0; i < mp_irq_entries; i++) {
		int lbus = mp_irqs[i].srcbus;

1007 1008
		for (ioapic_idx = 0; ioapic_idx < nr_ioapics; ioapic_idx++)
			if (mpc_ioapic_id(ioapic_idx) == mp_irqs[i].dstapic ||
1009 1010 1011 1012 1013 1014 1015
			    mp_irqs[i].dstapic == MP_APIC_ALL)
				break;

		if (!test_bit(lbus, mp_bus_not_pci) &&
		    !mp_irqs[i].irqtype &&
		    (bus == lbus) &&
		    (slot == ((mp_irqs[i].srcbusirq >> 2) & 0x1f))) {
1016
			int irq = pin_2_irq(i, ioapic_idx, mp_irqs[i].dstirq);
1017

1018
			if (!(ioapic_idx || IO_APIC_IRQ(irq)))
1019 1020 1021
				continue;

			if (pin == (mp_irqs[i].srcbusirq & 3)) {
1022
				set_io_apic_irq_attr(irq_attr, ioapic_idx,
1023 1024 1025
						     mp_irqs[i].dstirq,
						     irq_trigger(i),
						     irq_polarity(i));
1026 1027 1028 1029 1030 1031 1032
				return irq;
			}
			/*
			 * Use the first all-but-pin matching entry as a
			 * best-guess fuzzy result for broken mptables.
			 */
			if (best_guess < 0) {
1033
				set_io_apic_irq_attr(irq_attr, ioapic_idx,
1034 1035 1036
						     mp_irqs[i].dstirq,
						     irq_trigger(i),
						     irq_polarity(i));
1037 1038 1039 1040 1041 1042 1043 1044
				best_guess = irq;
			}
		}
	}
	return best_guess;
}
EXPORT_SYMBOL(IO_APIC_get_PCI_irq_vector);

1045 1046 1047 1048 1049
void lock_vector_lock(void)
{
	/* Used to the online set of cpus does not change
	 * during assign_irq_vector.
	 */
1050
	raw_spin_lock(&vector_lock);
1051
}
L
Linus Torvalds 已提交
1052

1053
void unlock_vector_lock(void)
L
Linus Torvalds 已提交
1054
{
1055
	raw_spin_unlock(&vector_lock);
1056
}
L
Linus Torvalds 已提交
1057

1058 1059
static int
__assign_irq_vector(int irq, struct irq_cfg *cfg, const struct cpumask *mask)
1060
{
1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071
	/*
	 * NOTE! The local APIC isn't very good at handling
	 * multiple interrupts at the same interrupt level.
	 * As the interrupt level is determined by taking the
	 * vector number and shifting that right by 4, we
	 * want to spread these out a bit so that they don't
	 * all fall in the same interrupt level.
	 *
	 * Also, we've got to be careful not to trash gate
	 * 0x80, because int 0x80 is hm, kind of importantish. ;)
	 */
1072
	static int current_vector = FIRST_EXTERNAL_VECTOR + VECTOR_OFFSET_START;
1073
	static int current_offset = VECTOR_OFFSET_START % 16;
1074 1075
	int cpu, err;
	cpumask_var_t tmp_mask;
1076

1077
	if (cfg->move_in_progress)
1078
		return -EBUSY;
1079

1080 1081
	if (!alloc_cpumask_var(&tmp_mask, GFP_ATOMIC))
		return -ENOMEM;
1082

1083
	/* Only try and allocate irqs on cpus that are present */
1084
	err = -ENOSPC;
1085 1086 1087
	cpumask_clear(cfg->old_domain);
	cpu = cpumask_first_and(mask, cpu_online_mask);
	while (cpu < nr_cpu_ids) {
1088
		int new_cpu, vector, offset;
1089

1090
		apic->vector_allocation_domain(cpu, tmp_mask, mask);
1091

1092
		if (cpumask_subset(tmp_mask, cfg->domain)) {
1093 1094 1095 1096 1097 1098 1099 1100 1101
			err = 0;
			if (cpumask_equal(tmp_mask, cfg->domain))
				break;
			/*
			 * New cpumask using the vector is a proper subset of
			 * the current in use mask. So cleanup the vector
			 * allocation for the members that are not used anymore.
			 */
			cpumask_andnot(cfg->old_domain, cfg->domain, tmp_mask);
1102 1103
			cfg->move_in_progress =
			   cpumask_intersects(cfg->old_domain, cpu_online_mask);
1104 1105
			cpumask_and(cfg->domain, cfg->domain, tmp_mask);
			break;
1106
		}
1107

1108 1109
		vector = current_vector;
		offset = current_offset;
1110
next:
1111
		vector += 16;
1112
		if (vector >= first_system_vector) {
1113
			offset = (offset + 1) % 16;
1114
			vector = FIRST_EXTERNAL_VECTOR + offset;
1115
		}
1116 1117

		if (unlikely(current_vector == vector)) {
1118 1119 1120
			cpumask_or(cfg->old_domain, cfg->old_domain, tmp_mask);
			cpumask_andnot(tmp_mask, mask, cfg->old_domain);
			cpu = cpumask_first_and(tmp_mask, cpu_online_mask);
1121
			continue;
1122
		}
1123 1124

		if (test_bit(vector, used_vectors))
1125
			goto next;
1126

1127 1128
		for_each_cpu_and(new_cpu, tmp_mask, cpu_online_mask) {
			if (per_cpu(vector_irq, new_cpu)[vector] > VECTOR_UNDEFINED)
1129
				goto next;
1130
		}
1131 1132 1133
		/* Found one! */
		current_vector = vector;
		current_offset = offset;
1134
		if (cfg->vector) {
1135
			cpumask_copy(cfg->old_domain, cfg->domain);
1136 1137
			cfg->move_in_progress =
			   cpumask_intersects(cfg->old_domain, cpu_online_mask);
1138
		}
1139
		for_each_cpu_and(new_cpu, tmp_mask, cpu_online_mask)
1140 1141
			per_cpu(vector_irq, new_cpu)[vector] = irq;
		cfg->vector = vector;
1142 1143 1144
		cpumask_copy(cfg->domain, tmp_mask);
		err = 0;
		break;
1145
	}
1146 1147
	free_cpumask_var(tmp_mask);
	return err;
1148 1149
}

1150
int assign_irq_vector(int irq, struct irq_cfg *cfg, const struct cpumask *mask)
1151 1152
{
	int err;
1153 1154
	unsigned long flags;

1155
	raw_spin_lock_irqsave(&vector_lock, flags);
Y
Yinghai Lu 已提交
1156
	err = __assign_irq_vector(irq, cfg, mask);
1157
	raw_spin_unlock_irqrestore(&vector_lock, flags);
1158 1159 1160
	return err;
}

Y
Yinghai Lu 已提交
1161
static void __clear_irq_vector(int irq, struct irq_cfg *cfg)
1162 1163 1164 1165 1166 1167
{
	int cpu, vector;

	BUG_ON(!cfg->vector);

	vector = cfg->vector;
1168
	for_each_cpu_and(cpu, cfg->domain, cpu_online_mask)
1169
		per_cpu(vector_irq, cpu)[vector] = VECTOR_UNDEFINED;
1170 1171

	cfg->vector = 0;
1172
	cpumask_clear(cfg->domain);
1173 1174 1175

	if (likely(!cfg->move_in_progress))
		return;
1176
	for_each_cpu_and(cpu, cfg->old_domain, cpu_online_mask) {
1177
		for (vector = FIRST_EXTERNAL_VECTOR; vector < NR_VECTORS; vector++) {
1178 1179
			if (per_cpu(vector_irq, cpu)[vector] != irq)
				continue;
1180
			per_cpu(vector_irq, cpu)[vector] = VECTOR_UNDEFINED;
1181 1182 1183 1184
			break;
		}
	}
	cfg->move_in_progress = 0;
1185 1186 1187 1188 1189 1190 1191 1192
}

void __setup_vector_irq(int cpu)
{
	/* Initialize vector_irq on a new cpu */
	int irq, vector;
	struct irq_cfg *cfg;

1193 1194 1195 1196 1197
	/*
	 * vector_lock will make sure that we don't run into irq vector
	 * assignments that might be happening on another cpu in parallel,
	 * while we setup our initial vector to irq mappings.
	 */
1198
	raw_spin_lock(&vector_lock);
1199
	/* Mark the inuse vectors */
T
Thomas Gleixner 已提交
1200
	for_each_active_irq(irq) {
1201
		cfg = irq_get_chip_data(irq);
T
Thomas Gleixner 已提交
1202 1203
		if (!cfg)
			continue;
1204

1205
		if (!cpumask_test_cpu(cpu, cfg->domain))
1206 1207 1208 1209 1210 1211 1212
			continue;
		vector = cfg->vector;
		per_cpu(vector_irq, cpu)[vector] = irq;
	}
	/* Mark the free vectors */
	for (vector = 0; vector < NR_VECTORS; ++vector) {
		irq = per_cpu(vector_irq, cpu)[vector];
1213
		if (irq <= VECTOR_UNDEFINED)
1214 1215 1216
			continue;

		cfg = irq_cfg(irq);
1217
		if (!cpumask_test_cpu(cpu, cfg->domain))
1218
			per_cpu(vector_irq, cpu)[vector] = VECTOR_UNDEFINED;
1219
	}
1220
	raw_spin_unlock(&vector_lock);
L
Linus Torvalds 已提交
1221
}
1222

1223
static struct irq_chip ioapic_chip;
L
Linus Torvalds 已提交
1224

1225
#ifdef CONFIG_X86_32
1226 1227
static inline int IO_APIC_irq_trigger(int irq)
{
T
Thomas Gleixner 已提交
1228
	int apic, idx, pin;
1229

T
Thomas Gleixner 已提交
1230
	for (apic = 0; apic < nr_ioapics; apic++) {
S
Suresh Siddha 已提交
1231
		for (pin = 0; pin < ioapics[apic].nr_registers; pin++) {
T
Thomas Gleixner 已提交
1232 1233 1234 1235 1236 1237
			idx = find_irq_entry(apic, pin, mp_INT);
			if ((idx != -1) && (irq == pin_2_irq(idx, apic, pin)))
				return irq_trigger(idx);
		}
	}
	/*
1238 1239
         * nonexistent IRQs are edge default
         */
T
Thomas Gleixner 已提交
1240
	return 0;
1241
}
1242 1243 1244
#else
static inline int IO_APIC_irq_trigger(int irq)
{
1245
	return 1;
1246 1247
}
#endif
1248

1249 1250
static void ioapic_register_intr(unsigned int irq, struct irq_cfg *cfg,
				 unsigned long trigger)
L
Linus Torvalds 已提交
1251
{
1252 1253 1254
	struct irq_chip *chip = &ioapic_chip;
	irq_flow_handler_t hdl;
	bool fasteoi;
Y
Yinghai Lu 已提交
1255

1256
	if ((trigger == IOAPIC_AUTO && IO_APIC_irq_trigger(irq)) ||
1257
	    trigger == IOAPIC_LEVEL) {
1258
		irq_set_status_flags(irq, IRQ_LEVEL);
1259 1260
		fasteoi = true;
	} else {
1261
		irq_clear_status_flags(irq, IRQ_LEVEL);
1262 1263
		fasteoi = false;
	}
1264

1265
	if (setup_remapped_irq(irq, cfg, chip))
1266
		fasteoi = trigger != 0;
1267

1268 1269 1270
	hdl = fasteoi ? handle_fasteoi_irq : handle_edge_irq;
	irq_set_chip_and_handler_name(irq, chip, hdl,
				      fasteoi ? "fasteoi" : "edge");
L
Linus Torvalds 已提交
1271 1272
}

1273 1274 1275
int native_setup_ioapic_entry(int irq, struct IO_APIC_route_entry *entry,
			      unsigned int destination, int vector,
			      struct io_apic_irq_attr *attr)
1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288
{
	memset(entry, 0, sizeof(*entry));

	entry->delivery_mode = apic->irq_delivery_mode;
	entry->dest_mode     = apic->irq_dest_mode;
	entry->dest	     = destination;
	entry->vector	     = vector;
	entry->mask	     = 0;			/* enable IRQ */
	entry->trigger	     = attr->trigger;
	entry->polarity	     = attr->polarity;

	/*
	 * Mask level triggered irqs.
1289 1290
	 * Use IRQ_DELAYED_DISABLE for edge triggered irqs.
	 */
1291
	if (attr->trigger)
1292
		entry->mask = 1;
1293

1294 1295 1296
	return 0;
}

1297 1298
static void setup_ioapic_irq(unsigned int irq, struct irq_cfg *cfg,
				struct io_apic_irq_attr *attr)
1299
{
L
Linus Torvalds 已提交
1300
	struct IO_APIC_route_entry entry;
1301
	unsigned int dest;
1302 1303 1304

	if (!IO_APIC_IRQ(irq))
		return;
1305

1306
	if (assign_irq_vector(irq, cfg, apic->target_cpus()))
1307 1308
		return;

1309 1310 1311 1312 1313 1314 1315 1316
	if (apic->cpu_mask_to_apicid_and(cfg->domain, apic->target_cpus(),
					 &dest)) {
		pr_warn("Failed to obtain apicid for ioapic %d, pin %d\n",
			mpc_ioapic_id(attr->ioapic), attr->ioapic_pin);
		__clear_irq_vector(irq, cfg);

		return;
	}
1317 1318 1319

	apic_printk(APIC_VERBOSE,KERN_DEBUG
		    "IOAPIC[%d]: Set routing entry (%d-%d -> 0x%x -> "
1320
		    "IRQ %d Mode:%i Active:%i Dest:%d)\n",
1321 1322
		    attr->ioapic, mpc_ioapic_id(attr->ioapic), attr->ioapic_pin,
		    cfg->vector, irq, attr->trigger, attr->polarity, dest);
1323

1324 1325
	if (x86_io_apic_ops.setup_entry(irq, &entry, dest, cfg->vector, attr)) {
		pr_warn("Failed to setup ioapic entry for ioapic  %d, pin %d\n",
1326
			mpc_ioapic_id(attr->ioapic), attr->ioapic_pin);
Y
Yinghai Lu 已提交
1327
		__clear_irq_vector(irq, cfg);
1328

1329 1330 1331
		return;
	}

1332
	ioapic_register_intr(irq, cfg, attr->trigger);
1333
	if (irq < legacy_pic->nr_legacy_irqs)
1334
		legacy_pic->mask(irq);
1335

1336
	ioapic_write_entry(attr->ioapic, attr->ioapic_pin, entry);
1337 1338
}

1339
static bool __init io_apic_pin_not_connected(int idx, int ioapic_idx, int pin)
1340 1341 1342 1343 1344
{
	if (idx != -1)
		return false;

	apic_printk(APIC_VERBOSE, KERN_DEBUG " apic %d pin %d not connected\n",
1345
		    mpc_ioapic_id(ioapic_idx), pin);
1346 1347 1348
	return true;
}

1349
static void __init __io_apic_setup_irqs(unsigned int ioapic_idx)
1350
{
1351
	int idx, node = cpu_to_node(0);
1352
	struct io_apic_irq_attr attr;
1353
	unsigned int pin, irq;
L
Linus Torvalds 已提交
1354

1355 1356 1357
	for (pin = 0; pin < ioapics[ioapic_idx].nr_registers; pin++) {
		idx = find_irq_entry(ioapic_idx, pin, mp_INT);
		if (io_apic_pin_not_connected(idx, ioapic_idx, pin))
1358
			continue;
1359

1360
		irq = pin_2_irq(idx, ioapic_idx, pin);
1361

1362
		if ((ioapic_idx > 0) && (irq > 16))
E
Eric W. Biederman 已提交
1363 1364
			continue;

1365 1366 1367 1368 1369
		/*
		 * Skip the timer IRQ if there's a quirk handler
		 * installed and if it returns 1:
		 */
		if (apic->multi_timer_check &&
1370
		    apic->multi_timer_check(ioapic_idx, irq))
1371
			continue;
1372

1373
		set_io_apic_irq_attr(&attr, ioapic_idx, pin, irq_trigger(idx),
1374
				     irq_polarity(idx));
1375

1376
		io_apic_setup_irq_pin(irq, node, &attr);
L
Linus Torvalds 已提交
1377 1378 1379
	}
}

1380 1381
static void __init setup_IO_APIC_irqs(void)
{
1382
	unsigned int ioapic_idx;
1383 1384 1385

	apic_printk(APIC_VERBOSE, KERN_DEBUG "init IO_APIC IRQs\n");

1386 1387
	for (ioapic_idx = 0; ioapic_idx < nr_ioapics; ioapic_idx++)
		__io_apic_setup_irqs(ioapic_idx);
1388 1389
}

Y
Yinghai Lu 已提交
1390 1391 1392 1393 1394 1395 1396
/*
 * for the gsit that is not in first ioapic
 * but could not use acpi_register_gsi()
 * like some special sci in IBM x3330
 */
void setup_IO_APIC_irq_extra(u32 gsi)
{
1397
	int ioapic_idx = 0, pin, idx, irq, node = cpu_to_node(0);
1398
	struct io_apic_irq_attr attr;
Y
Yinghai Lu 已提交
1399 1400 1401 1402

	/*
	 * Convert 'gsi' to 'ioapic.pin'.
	 */
1403 1404
	ioapic_idx = mp_find_ioapic(gsi);
	if (ioapic_idx < 0)
Y
Yinghai Lu 已提交
1405 1406
		return;

1407 1408
	pin = mp_find_ioapic_pin(ioapic_idx, gsi);
	idx = find_irq_entry(ioapic_idx, pin, mp_INT);
Y
Yinghai Lu 已提交
1409 1410 1411
	if (idx == -1)
		return;

1412
	irq = pin_2_irq(idx, ioapic_idx, pin);
1413 1414

	/* Only handle the non legacy irqs on secondary ioapics */
1415
	if (ioapic_idx == 0 || irq < NR_IRQS_LEGACY)
Y
Yinghai Lu 已提交
1416
		return;
1417

1418
	set_io_apic_irq_attr(&attr, ioapic_idx, pin, irq_trigger(idx),
1419 1420
			     irq_polarity(idx));

1421
	io_apic_setup_irq_pin_once(irq, node, &attr);
Y
Yinghai Lu 已提交
1422 1423
}

L
Linus Torvalds 已提交
1424
/*
1425
 * Set up the timer pin, possibly with the 8259A-master behind.
L
Linus Torvalds 已提交
1426
 */
1427
static void __init setup_timer_IRQ0_pin(unsigned int ioapic_idx,
1428
					unsigned int pin, int vector)
L
Linus Torvalds 已提交
1429 1430
{
	struct IO_APIC_route_entry entry;
1431
	unsigned int dest;
L
Linus Torvalds 已提交
1432

1433
	memset(&entry, 0, sizeof(entry));
L
Linus Torvalds 已提交
1434 1435 1436 1437 1438

	/*
	 * We use logical delivery to get the timer IRQ
	 * to the first CPU.
	 */
1439 1440
	if (unlikely(apic->cpu_mask_to_apicid_and(apic->target_cpus(),
						  apic->target_cpus(), &dest)))
1441 1442
		dest = BAD_APICID;

1443
	entry.dest_mode = apic->irq_dest_mode;
Y
Yinghai Lu 已提交
1444
	entry.mask = 0;			/* don't mask IRQ for edge */
1445
	entry.dest = dest;
1446
	entry.delivery_mode = apic->irq_delivery_mode;
L
Linus Torvalds 已提交
1447 1448 1449 1450 1451 1452
	entry.polarity = 0;
	entry.trigger = 0;
	entry.vector = vector;

	/*
	 * The timer IRQ doesn't have to know that behind the
1453
	 * scene we may have a 8259A-master in AEOI mode ...
L
Linus Torvalds 已提交
1454
	 */
1455 1456
	irq_set_chip_and_handler_name(0, &ioapic_chip, handle_edge_irq,
				      "edge");
L
Linus Torvalds 已提交
1457 1458 1459 1460

	/*
	 * Add it to the IO-APIC irq-routing table:
	 */
1461
	ioapic_write_entry(ioapic_idx, pin, entry);
L
Linus Torvalds 已提交
1462 1463
}

1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490
void native_io_apic_print_entries(unsigned int apic, unsigned int nr_entries)
{
	int i;

	pr_debug(" NR Dst Mask Trig IRR Pol Stat Dmod Deli Vect:\n");

	for (i = 0; i <= nr_entries; i++) {
		struct IO_APIC_route_entry entry;

		entry = ioapic_read_entry(apic, i);

		pr_debug(" %02x %02X  ", i, entry.dest);
		pr_cont("%1d    %1d    %1d   %1d   %1d    "
			"%1d    %1d    %02X\n",
			entry.mask,
			entry.trigger,
			entry.irr,
			entry.polarity,
			entry.delivery_status,
			entry.dest_mode,
			entry.delivery_mode,
			entry.vector);
	}
}

void intel_ir_io_apic_print_entries(unsigned int apic,
				    unsigned int nr_entries)
L
Linus Torvalds 已提交
1491
{
1492
	int i;
1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518

	pr_debug(" NR Indx Fmt Mask Trig IRR Pol Stat Indx2 Zero Vect:\n");

	for (i = 0; i <= nr_entries; i++) {
		struct IR_IO_APIC_route_entry *ir_entry;
		struct IO_APIC_route_entry entry;

		entry = ioapic_read_entry(apic, i);

		ir_entry = (struct IR_IO_APIC_route_entry *)&entry;

		pr_debug(" %02x %04X ", i, ir_entry->index);
		pr_cont("%1d   %1d    %1d    %1d   %1d   "
			"%1d    %1d     %X    %02X\n",
			ir_entry->format,
			ir_entry->mask,
			ir_entry->trigger,
			ir_entry->irr,
			ir_entry->polarity,
			ir_entry->delivery_status,
			ir_entry->index2,
			ir_entry->zero,
			ir_entry->vector);
	}
}

1519 1520 1521 1522 1523
void ioapic_zap_locks(void)
{
	raw_spin_lock_init(&ioapic_lock);
}

1524 1525
__apicdebuginit(void) print_IO_APIC(int ioapic_idx)
{
L
Linus Torvalds 已提交
1526 1527 1528 1529 1530 1531
	union IO_APIC_reg_00 reg_00;
	union IO_APIC_reg_01 reg_01;
	union IO_APIC_reg_02 reg_02;
	union IO_APIC_reg_03 reg_03;
	unsigned long flags;

1532
	raw_spin_lock_irqsave(&ioapic_lock, flags);
1533 1534
	reg_00.raw = io_apic_read(ioapic_idx, 0);
	reg_01.raw = io_apic_read(ioapic_idx, 1);
L
Linus Torvalds 已提交
1535
	if (reg_01.bits.version >= 0x10)
1536
		reg_02.raw = io_apic_read(ioapic_idx, 2);
T
Thomas Gleixner 已提交
1537
	if (reg_01.bits.version >= 0x20)
1538
		reg_03.raw = io_apic_read(ioapic_idx, 3);
1539
	raw_spin_unlock_irqrestore(&ioapic_lock, flags);
L
Linus Torvalds 已提交
1540

1541
	printk(KERN_DEBUG "IO APIC #%d......\n", mpc_ioapic_id(ioapic_idx));
L
Linus Torvalds 已提交
1542 1543 1544 1545 1546
	printk(KERN_DEBUG ".... register #00: %08X\n", reg_00.raw);
	printk(KERN_DEBUG ".......    : physical APIC id: %02X\n", reg_00.bits.ID);
	printk(KERN_DEBUG ".......    : Delivery Type: %X\n", reg_00.bits.delivery_type);
	printk(KERN_DEBUG ".......    : LTS          : %X\n", reg_00.bits.LTS);

1547
	printk(KERN_DEBUG ".... register #01: %08X\n", *(int *)&reg_01);
1548 1549
	printk(KERN_DEBUG ".......     : max redirection entries: %02X\n",
		reg_01.bits.entries);
L
Linus Torvalds 已提交
1550 1551

	printk(KERN_DEBUG ".......     : PRQ implemented: %X\n", reg_01.bits.PRQ);
1552 1553
	printk(KERN_DEBUG ".......     : IO APIC version: %02X\n",
		reg_01.bits.version);
L
Linus Torvalds 已提交
1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577

	/*
	 * Some Intel chipsets with IO APIC VERSION of 0x1? don't have reg_02,
	 * but the value of reg_02 is read as the previous read register
	 * value, so ignore it if reg_02 == reg_01.
	 */
	if (reg_01.bits.version >= 0x10 && reg_02.raw != reg_01.raw) {
		printk(KERN_DEBUG ".... register #02: %08X\n", reg_02.raw);
		printk(KERN_DEBUG ".......     : arbitration: %02X\n", reg_02.bits.arbitration);
	}

	/*
	 * Some Intel chipsets with IO APIC VERSION of 0x2? don't have reg_02
	 * or reg_03, but the value of reg_0[23] is read as the previous read
	 * register value, so ignore it if reg_03 == reg_0[12].
	 */
	if (reg_01.bits.version >= 0x20 && reg_03.raw != reg_02.raw &&
	    reg_03.raw != reg_01.raw) {
		printk(KERN_DEBUG ".... register #03: %08X\n", reg_03.raw);
		printk(KERN_DEBUG ".......     : Boot DT    : %X\n", reg_03.bits.boot_DT);
	}

	printk(KERN_DEBUG ".... IRQ redirection table:\n");

1578
	x86_io_apic_ops.print_entries(ioapic_idx, reg_01.bits.entries);
1579 1580 1581 1582
}

__apicdebuginit(void) print_IO_APICs(void)
{
1583
	int ioapic_idx;
1584 1585
	struct irq_cfg *cfg;
	unsigned int irq;
1586
	struct irq_chip *chip;
1587 1588

	printk(KERN_DEBUG "number of MP IRQ sources: %d.\n", mp_irq_entries);
1589
	for (ioapic_idx = 0; ioapic_idx < nr_ioapics; ioapic_idx++)
1590
		printk(KERN_DEBUG "number of IO-APIC #%d registers: %d.\n",
1591 1592
		       mpc_ioapic_id(ioapic_idx),
		       ioapics[ioapic_idx].nr_registers);
1593 1594 1595 1596 1597 1598 1599

	/*
	 * We are a bit conservative about what we expect.  We have to
	 * know about every hardware change ASAP.
	 */
	printk(KERN_INFO "testing the IO APIC.......................\n");

1600 1601
	for (ioapic_idx = 0; ioapic_idx < nr_ioapics; ioapic_idx++)
		print_IO_APIC(ioapic_idx);
1602

L
Linus Torvalds 已提交
1603
	printk(KERN_DEBUG "IRQ to pin mappings:\n");
T
Thomas Gleixner 已提交
1604
	for_each_active_irq(irq) {
1605 1606
		struct irq_pin_list *entry;

1607 1608 1609 1610
		chip = irq_get_chip(irq);
		if (chip != &ioapic_chip)
			continue;

1611
		cfg = irq_get_chip_data(irq);
1612 1613
		if (!cfg)
			continue;
1614
		entry = cfg->irq_2_pin;
1615
		if (!entry)
L
Linus Torvalds 已提交
1616
			continue;
1617
		printk(KERN_DEBUG "IRQ%d ", irq);
1618
		for_each_irq_pin(entry, cfg->irq_2_pin)
1619 1620
			pr_cont("-> %d:%d", entry->apic, entry->pin);
		pr_cont("\n");
L
Linus Torvalds 已提交
1621 1622 1623 1624 1625
	}

	printk(KERN_INFO ".................................... done.\n");
}

1626
__apicdebuginit(void) print_APIC_field(int base)
L
Linus Torvalds 已提交
1627
{
1628
	int i;
L
Linus Torvalds 已提交
1629

1630 1631 1632
	printk(KERN_DEBUG);

	for (i = 0; i < 8; i++)
1633
		pr_cont("%08x", apic_read(base + i*0x10));
1634

1635
	pr_cont("\n");
L
Linus Torvalds 已提交
1636 1637
}

1638
__apicdebuginit(void) print_local_APIC(void *dummy)
L
Linus Torvalds 已提交
1639
{
1640
	unsigned int i, v, ver, maxlvt;
1641
	u64 icr;
L
Linus Torvalds 已提交
1642

1643
	printk(KERN_DEBUG "printing local APIC contents on CPU#%d/%d:\n",
L
Linus Torvalds 已提交
1644
		smp_processor_id(), hard_smp_processor_id());
1645
	v = apic_read(APIC_ID);
1646
	printk(KERN_INFO "... APIC ID:      %08x (%01x)\n", v, read_apic_id());
L
Linus Torvalds 已提交
1647 1648 1649
	v = apic_read(APIC_LVR);
	printk(KERN_INFO "... APIC VERSION: %08x\n", v);
	ver = GET_APIC_VERSION(v);
1650
	maxlvt = lapic_get_maxlvt();
L
Linus Torvalds 已提交
1651 1652 1653 1654

	v = apic_read(APIC_TASKPRI);
	printk(KERN_DEBUG "... APIC TASKPRI: %08x (%02x)\n", v, v & APIC_TPRI_MASK);

1655
	if (APIC_INTEGRATED(ver)) {                     /* !82489DX */
1656 1657 1658 1659 1660
		if (!APIC_XAPIC(ver)) {
			v = apic_read(APIC_ARBPRI);
			printk(KERN_DEBUG "... APIC ARBPRI: %08x (%02x)\n", v,
			       v & APIC_ARBPRI_MASK);
		}
L
Linus Torvalds 已提交
1661 1662 1663 1664
		v = apic_read(APIC_PROCPRI);
		printk(KERN_DEBUG "... APIC PROCPRI: %08x\n", v);
	}

1665 1666 1667 1668 1669 1670 1671 1672 1673
	/*
	 * Remote read supported only in the 82489DX and local APIC for
	 * Pentium processors.
	 */
	if (!APIC_INTEGRATED(ver) || maxlvt == 3) {
		v = apic_read(APIC_RRR);
		printk(KERN_DEBUG "... APIC RRR: %08x\n", v);
	}

L
Linus Torvalds 已提交
1674 1675
	v = apic_read(APIC_LDR);
	printk(KERN_DEBUG "... APIC LDR: %08x\n", v);
1676 1677 1678 1679
	if (!x2apic_enabled()) {
		v = apic_read(APIC_DFR);
		printk(KERN_DEBUG "... APIC DFR: %08x\n", v);
	}
L
Linus Torvalds 已提交
1680 1681 1682 1683
	v = apic_read(APIC_SPIV);
	printk(KERN_DEBUG "... APIC SPIV: %08x\n", v);

	printk(KERN_DEBUG "... APIC ISR field:\n");
1684
	print_APIC_field(APIC_ISR);
L
Linus Torvalds 已提交
1685
	printk(KERN_DEBUG "... APIC TMR field:\n");
1686
	print_APIC_field(APIC_TMR);
L
Linus Torvalds 已提交
1687
	printk(KERN_DEBUG "... APIC IRR field:\n");
1688
	print_APIC_field(APIC_IRR);
L
Linus Torvalds 已提交
1689

1690 1691
	if (APIC_INTEGRATED(ver)) {             /* !82489DX */
		if (maxlvt > 3)         /* Due to the Pentium erratum 3AP. */
L
Linus Torvalds 已提交
1692
			apic_write(APIC_ESR, 0);
1693

L
Linus Torvalds 已提交
1694 1695 1696 1697
		v = apic_read(APIC_ESR);
		printk(KERN_DEBUG "... APIC ESR: %08x\n", v);
	}

1698
	icr = apic_icr_read();
1699 1700
	printk(KERN_DEBUG "... APIC ICR: %08x\n", (u32)icr);
	printk(KERN_DEBUG "... APIC ICR2: %08x\n", (u32)(icr >> 32));
L
Linus Torvalds 已提交
1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724

	v = apic_read(APIC_LVTT);
	printk(KERN_DEBUG "... APIC LVTT: %08x\n", v);

	if (maxlvt > 3) {                       /* PC is LVT#4. */
		v = apic_read(APIC_LVTPC);
		printk(KERN_DEBUG "... APIC LVTPC: %08x\n", v);
	}
	v = apic_read(APIC_LVT0);
	printk(KERN_DEBUG "... APIC LVT0: %08x\n", v);
	v = apic_read(APIC_LVT1);
	printk(KERN_DEBUG "... APIC LVT1: %08x\n", v);

	if (maxlvt > 2) {			/* ERR is LVT#3. */
		v = apic_read(APIC_LVTERR);
		printk(KERN_DEBUG "... APIC LVTERR: %08x\n", v);
	}

	v = apic_read(APIC_TMICT);
	printk(KERN_DEBUG "... APIC TMICT: %08x\n", v);
	v = apic_read(APIC_TMCCT);
	printk(KERN_DEBUG "... APIC TMCCT: %08x\n", v);
	v = apic_read(APIC_TDCR);
	printk(KERN_DEBUG "... APIC TDCR: %08x\n", v);
1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736

	if (boot_cpu_has(X86_FEATURE_EXTAPIC)) {
		v = apic_read(APIC_EFEAT);
		maxlvt = (v >> 16) & 0xff;
		printk(KERN_DEBUG "... APIC EFEAT: %08x\n", v);
		v = apic_read(APIC_ECTRL);
		printk(KERN_DEBUG "... APIC ECTRL: %08x\n", v);
		for (i = 0; i < maxlvt; i++) {
			v = apic_read(APIC_EILVTn(i));
			printk(KERN_DEBUG "... APIC EILVT%d: %08x\n", i, v);
		}
	}
1737
	pr_cont("\n");
L
Linus Torvalds 已提交
1738 1739
}

1740
__apicdebuginit(void) print_local_APICs(int maxcpu)
L
Linus Torvalds 已提交
1741
{
1742 1743
	int cpu;

1744 1745 1746
	if (!maxcpu)
		return;

1747
	preempt_disable();
1748 1749 1750
	for_each_online_cpu(cpu) {
		if (cpu >= maxcpu)
			break;
1751
		smp_call_function_single(cpu, print_local_APIC, NULL, 1);
1752
	}
1753
	preempt_enable();
L
Linus Torvalds 已提交
1754 1755
}

1756
__apicdebuginit(void) print_PIC(void)
L
Linus Torvalds 已提交
1757 1758 1759 1760
{
	unsigned int v;
	unsigned long flags;

1761
	if (!legacy_pic->nr_legacy_irqs)
L
Linus Torvalds 已提交
1762 1763 1764 1765
		return;

	printk(KERN_DEBUG "\nprinting PIC contents\n");

1766
	raw_spin_lock_irqsave(&i8259A_lock, flags);
L
Linus Torvalds 已提交
1767 1768 1769 1770 1771 1772 1773

	v = inb(0xa1) << 8 | inb(0x21);
	printk(KERN_DEBUG "... PIC  IMR: %04x\n", v);

	v = inb(0xa0) << 8 | inb(0x20);
	printk(KERN_DEBUG "... PIC  IRR: %04x\n", v);

1774 1775
	outb(0x0b,0xa0);
	outb(0x0b,0x20);
L
Linus Torvalds 已提交
1776
	v = inb(0xa0) << 8 | inb(0x20);
1777 1778
	outb(0x0a,0xa0);
	outb(0x0a,0x20);
L
Linus Torvalds 已提交
1779

1780
	raw_spin_unlock_irqrestore(&i8259A_lock, flags);
L
Linus Torvalds 已提交
1781 1782 1783 1784 1785 1786 1787

	printk(KERN_DEBUG "... PIC  ISR: %04x\n", v);

	v = inb(0x4d1) << 8 | inb(0x4d0);
	printk(KERN_DEBUG "... PIC ELCR: %04x\n", v);
}

1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805
static int __initdata show_lapic = 1;
static __init int setup_show_lapic(char *arg)
{
	int num = -1;

	if (strcmp(arg, "all") == 0) {
		show_lapic = CONFIG_NR_CPUS;
	} else {
		get_option(&arg, &num);
		if (num >= 0)
			show_lapic = num;
	}

	return 1;
}
__setup("show_lapic=", setup_show_lapic);

__apicdebuginit(int) print_ICs(void)
1806
{
1807 1808 1809
	if (apic_verbosity == APIC_QUIET)
		return 0;

1810
	print_PIC();
1811 1812

	/* don't print out if apic is not there */
1813
	if (!cpu_has_apic && !apic_from_smp_config())
1814 1815
		return 0;

1816
	print_local_APICs(show_lapic);
1817
	print_IO_APICs();
1818 1819 1820 1821

	return 0;
}

1822
late_initcall(print_ICs);
1823

L
Linus Torvalds 已提交
1824

Y
Yinghai Lu 已提交
1825 1826 1827
/* Where if anywhere is the i8259 connect in external int mode */
static struct { int pin, apic; } ioapic_i8259 = { -1, -1 };

1828
void __init enable_IO_APIC(void)
L
Linus Torvalds 已提交
1829
{
1830
	int i8259_apic, i8259_pin;
1831
	int apic;
1832

1833
	if (!legacy_pic->nr_legacy_irqs)
1834 1835
		return;

1836
	for(apic = 0; apic < nr_ioapics; apic++) {
1837 1838
		int pin;
		/* See if any of the pins is in ExtINT mode */
S
Suresh Siddha 已提交
1839
		for (pin = 0; pin < ioapics[apic].nr_registers; pin++) {
1840
			struct IO_APIC_route_entry entry;
1841
			entry = ioapic_read_entry(apic, pin);
1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871

			/* If the interrupt line is enabled and in ExtInt mode
			 * I have found the pin where the i8259 is connected.
			 */
			if ((entry.mask == 0) && (entry.delivery_mode == dest_ExtINT)) {
				ioapic_i8259.apic = apic;
				ioapic_i8259.pin  = pin;
				goto found_i8259;
			}
		}
	}
 found_i8259:
	/* Look to see what if the MP table has reported the ExtINT */
	/* If we could not find the appropriate pin by looking at the ioapic
	 * the i8259 probably is not connected the ioapic but give the
	 * mptable a chance anyway.
	 */
	i8259_pin  = find_isa_irq_pin(0, mp_ExtINT);
	i8259_apic = find_isa_irq_apic(0, mp_ExtINT);
	/* Trust the MP table if nothing is setup in the hardware */
	if ((ioapic_i8259.pin == -1) && (i8259_pin >= 0)) {
		printk(KERN_WARNING "ExtINT not setup in hardware but reported by MP table\n");
		ioapic_i8259.pin  = i8259_pin;
		ioapic_i8259.apic = i8259_apic;
	}
	/* Complain if the MP table and the hardware disagree */
	if (((ioapic_i8259.apic != i8259_apic) || (ioapic_i8259.pin != i8259_pin)) &&
		(i8259_pin >= 0) && (ioapic_i8259.pin >= 0))
	{
		printk(KERN_WARNING "ExtINT in hardware and MP table differ\n");
L
Linus Torvalds 已提交
1872 1873 1874 1875 1876 1877 1878 1879
	}

	/*
	 * Do not trust the IO-APIC being empty at bootup
	 */
	clear_IO_APIC();
}

1880
void native_disable_io_apic(void)
L
Linus Torvalds 已提交
1881
{
1882
	/*
1883
	 * If the i8259 is routed through an IOAPIC
1884
	 * Put that IOAPIC in virtual wire mode
1885
	 * so legacy interrupts can be delivered.
1886
	 */
1887
	if (ioapic_i8259.pin != -1) {
1888 1889 1890 1891 1892 1893 1894 1895 1896
		struct IO_APIC_route_entry entry;

		memset(&entry, 0, sizeof(entry));
		entry.mask            = 0; /* Enabled */
		entry.trigger         = 0; /* Edge */
		entry.irr             = 0;
		entry.polarity        = 0; /* High */
		entry.delivery_status = 0;
		entry.dest_mode       = 0; /* Physical */
1897
		entry.delivery_mode   = dest_ExtINT; /* ExtInt */
1898
		entry.vector          = 0;
1899
		entry.dest            = read_apic_id();
1900 1901 1902 1903

		/*
		 * Add it to the IO-APIC irq-routing table:
		 */
1904
		ioapic_write_entry(ioapic_i8259.apic, ioapic_i8259.pin, entry);
1905
	}
1906

1907 1908 1909 1910 1911 1912 1913 1914 1915 1916
	if (cpu_has_apic || apic_from_smp_config())
		disconnect_bsp_APIC(ioapic_i8259.pin != -1);

}

/*
 * Not an __init, needed by the reboot code
 */
void disable_IO_APIC(void)
{
1917
	/*
1918
	 * Clear the IO-APIC before rebooting:
1919
	 */
1920 1921 1922 1923 1924 1925
	clear_IO_APIC();

	if (!legacy_pic->nr_legacy_irqs)
		return;

	x86_io_apic_ops.disable();
L
Linus Torvalds 已提交
1926 1927
}

1928
#ifdef CONFIG_X86_32
L
Linus Torvalds 已提交
1929 1930 1931 1932 1933 1934
/*
 * function to set the IO-APIC physical IDs based on the
 * values stored in the MPC table.
 *
 * by Matt Domsch <Matt_Domsch@dell.com>  Tue Dec 21 12:25:05 CST 1999
 */
1935
void __init setup_ioapic_ids_from_mpc_nocheck(void)
L
Linus Torvalds 已提交
1936 1937 1938
{
	union IO_APIC_reg_00 reg_00;
	physid_mask_t phys_id_present_map;
1939
	int ioapic_idx;
L
Linus Torvalds 已提交
1940 1941 1942 1943 1944 1945 1946 1947
	int i;
	unsigned char old_id;
	unsigned long flags;

	/*
	 * This is broken; anything with a real cpu count has to
	 * circumvent this idiocy regardless.
	 */
1948
	apic->ioapic_phys_id_map(&phys_cpu_present_map, &phys_id_present_map);
L
Linus Torvalds 已提交
1949 1950 1951 1952

	/*
	 * Set the IOAPIC ID to the value stored in the MPC table.
	 */
1953
	for (ioapic_idx = 0; ioapic_idx < nr_ioapics; ioapic_idx++) {
L
Linus Torvalds 已提交
1954
		/* Read the register 0 value */
1955
		raw_spin_lock_irqsave(&ioapic_lock, flags);
1956
		reg_00.raw = io_apic_read(ioapic_idx, 0);
1957
		raw_spin_unlock_irqrestore(&ioapic_lock, flags);
1958

1959
		old_id = mpc_ioapic_id(ioapic_idx);
L
Linus Torvalds 已提交
1960

1961
		if (mpc_ioapic_id(ioapic_idx) >= get_physical_broadcast()) {
L
Linus Torvalds 已提交
1962
			printk(KERN_ERR "BIOS bug, IO-APIC#%d ID is %d in the MPC table!...\n",
1963
				ioapic_idx, mpc_ioapic_id(ioapic_idx));
L
Linus Torvalds 已提交
1964 1965
			printk(KERN_ERR "... fixing up to %d. (tell your hw vendor)\n",
				reg_00.bits.ID);
1966
			ioapics[ioapic_idx].mp_config.apicid = reg_00.bits.ID;
L
Linus Torvalds 已提交
1967 1968 1969 1970 1971 1972 1973
		}

		/*
		 * Sanity check, is the ID really free? Every APIC in a
		 * system must have a unique ID or we get lots of nice
		 * 'stuck on smp_invalidate_needed IPI wait' messages.
		 */
1974
		if (apic->check_apicid_used(&phys_id_present_map,
1975
					    mpc_ioapic_id(ioapic_idx))) {
L
Linus Torvalds 已提交
1976
			printk(KERN_ERR "BIOS bug, IO-APIC#%d ID %d is already used!...\n",
1977
				ioapic_idx, mpc_ioapic_id(ioapic_idx));
L
Linus Torvalds 已提交
1978 1979 1980 1981 1982 1983 1984 1985
			for (i = 0; i < get_physical_broadcast(); i++)
				if (!physid_isset(i, phys_id_present_map))
					break;
			if (i >= get_physical_broadcast())
				panic("Max APIC ID exceeded!\n");
			printk(KERN_ERR "... fixing up to %d. (tell your hw vendor)\n",
				i);
			physid_set(i, phys_id_present_map);
1986
			ioapics[ioapic_idx].mp_config.apicid = i;
L
Linus Torvalds 已提交
1987 1988
		} else {
			physid_mask_t tmp;
1989
			apic->apicid_to_cpu_present(mpc_ioapic_id(ioapic_idx),
1990
						    &tmp);
L
Linus Torvalds 已提交
1991 1992
			apic_printk(APIC_VERBOSE, "Setting %d in the "
					"phys_id_present_map\n",
1993
					mpc_ioapic_id(ioapic_idx));
L
Linus Torvalds 已提交
1994 1995 1996 1997 1998 1999 2000
			physids_or(phys_id_present_map, phys_id_present_map, tmp);
		}

		/*
		 * We need to adjust the IRQ routing table
		 * if the ID changed.
		 */
2001
		if (old_id != mpc_ioapic_id(ioapic_idx))
L
Linus Torvalds 已提交
2002
			for (i = 0; i < mp_irq_entries; i++)
2003 2004
				if (mp_irqs[i].dstapic == old_id)
					mp_irqs[i].dstapic
2005
						= mpc_ioapic_id(ioapic_idx);
L
Linus Torvalds 已提交
2006 2007

		/*
2008 2009
		 * Update the ID register according to the right value
		 * from the MPC table if they are different.
2010
		 */
2011
		if (mpc_ioapic_id(ioapic_idx) == reg_00.bits.ID)
2012 2013
			continue;

L
Linus Torvalds 已提交
2014 2015
		apic_printk(APIC_VERBOSE, KERN_INFO
			"...changing IO-APIC physical APIC ID to %d ...",
2016
			mpc_ioapic_id(ioapic_idx));
L
Linus Torvalds 已提交
2017

2018
		reg_00.bits.ID = mpc_ioapic_id(ioapic_idx);
2019
		raw_spin_lock_irqsave(&ioapic_lock, flags);
2020
		io_apic_write(ioapic_idx, 0, reg_00.raw);
2021
		raw_spin_unlock_irqrestore(&ioapic_lock, flags);
L
Linus Torvalds 已提交
2022 2023 2024 2025

		/*
		 * Sanity check
		 */
2026
		raw_spin_lock_irqsave(&ioapic_lock, flags);
2027
		reg_00.raw = io_apic_read(ioapic_idx, 0);
2028
		raw_spin_unlock_irqrestore(&ioapic_lock, flags);
2029
		if (reg_00.bits.ID != mpc_ioapic_id(ioapic_idx))
2030
			pr_cont("could not set ID!\n");
L
Linus Torvalds 已提交
2031 2032 2033 2034
		else
			apic_printk(APIC_VERBOSE, " ok.\n");
	}
}
2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049

void __init setup_ioapic_ids_from_mpc(void)
{

	if (acpi_ioapic)
		return;
	/*
	 * Don't check I/O APIC IDs for xAPIC systems.  They have
	 * no meaning without the serial APIC bus.
	 */
	if (!(boot_cpu_data.x86_vendor == X86_VENDOR_INTEL)
		|| APIC_XAPIC(apic_version[boot_cpu_physical_apicid]))
		return;
	setup_ioapic_ids_from_mpc_nocheck();
}
2050
#endif
L
Linus Torvalds 已提交
2051

2052
int no_timer_check __initdata;
2053 2054 2055 2056 2057 2058 2059 2060

static int __init notimercheck(char *s)
{
	no_timer_check = 1;
	return 1;
}
__setup("no_timer_check", notimercheck);

L
Linus Torvalds 已提交
2061 2062 2063 2064 2065 2066 2067 2068
/*
 * There is a nasty bug in some older SMP boards, their mptable lies
 * about the timer IRQ. We do the following to work around the situation:
 *
 *	- timer IRQ defaults to IO-APIC IRQ
 *	- if this function detects that timer IRQs are defunct, then we fall
 *	  back to ISA timer IRQs
 */
2069
static int __init timer_irq_works(void)
L
Linus Torvalds 已提交
2070 2071
{
	unsigned long t1 = jiffies;
2072
	unsigned long flags;
L
Linus Torvalds 已提交
2073

2074 2075 2076
	if (no_timer_check)
		return 1;

2077
	local_save_flags(flags);
L
Linus Torvalds 已提交
2078 2079 2080
	local_irq_enable();
	/* Let ten ticks pass... */
	mdelay((10 * 1000) / HZ);
2081
	local_irq_restore(flags);
L
Linus Torvalds 已提交
2082 2083 2084 2085 2086 2087 2088 2089

	/*
	 * Expect a few ticks at least, to be sure some possible
	 * glue logic does not lock up after one or two first
	 * ticks in a non-ExtINT mode.  Also the local APIC
	 * might have cached one ExtINT interrupt.  Finally, at
	 * least one tick may be lost due to delays.
	 */
2090 2091

	/* jiffies wrap? */
2092
	if (time_after(jiffies, t1 + 4))
L
Linus Torvalds 已提交
2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118
		return 1;
	return 0;
}

/*
 * In the SMP+IOAPIC case it might happen that there are an unspecified
 * number of pending IRQ events unhandled. These cases are very rare,
 * so we 'resend' these IRQs via IPIs, to the same CPU. It's much
 * better to do it this way as thus we do not have to be aware of
 * 'pending' interrupts in the IRQ path, except at this point.
 */
/*
 * Edge triggered needs to resend any interrupt
 * that was delayed but this is now handled in the device
 * independent code.
 */

/*
 * Starting up a edge-triggered IO-APIC interrupt is
 * nasty - we need to make sure that we get the edge.
 * If it is already asserted for some reason, we need
 * return 1 to indicate that is was pending.
 *
 * This is not complete - we should be able to fake
 * an edge even if it isn't on the 8259A...
 */
2119

2120
static unsigned int startup_ioapic_irq(struct irq_data *data)
L
Linus Torvalds 已提交
2121
{
2122
	int was_pending = 0, irq = data->irq;
L
Linus Torvalds 已提交
2123 2124
	unsigned long flags;

2125
	raw_spin_lock_irqsave(&ioapic_lock, flags);
2126
	if (irq < legacy_pic->nr_legacy_irqs) {
2127
		legacy_pic->mask(irq);
2128
		if (legacy_pic->irq_pending(irq))
L
Linus Torvalds 已提交
2129 2130
			was_pending = 1;
	}
2131
	__unmask_ioapic(data->chip_data);
2132
	raw_spin_unlock_irqrestore(&ioapic_lock, flags);
L
Linus Torvalds 已提交
2133 2134 2135 2136

	return was_pending;
}

2137
static int ioapic_retrigger_irq(struct irq_data *data)
L
Linus Torvalds 已提交
2138
{
2139
	struct irq_cfg *cfg = data->chip_data;
2140
	unsigned long flags;
2141
	int cpu;
2142

2143
	raw_spin_lock_irqsave(&vector_lock, flags);
2144 2145
	cpu = cpumask_first_and(cfg->domain, cpu_online_mask);
	apic->send_IPI_mask(cpumask_of(cpu), cfg->vector);
2146
	raw_spin_unlock_irqrestore(&vector_lock, flags);
2147 2148 2149

	return 1;
}
2150

2151 2152 2153 2154 2155 2156 2157 2158
/*
 * Level and edge triggered IO-APIC interrupts need different handling,
 * so we use two separate IRQ descriptors. Edge triggered IRQs can be
 * handled with the level-triggered descriptor, but that one has slightly
 * more overhead. Level-triggered interrupts cannot be handled with the
 * edge-triggered handler, without risking IRQ storms and other ugly
 * races.
 */
2159

2160
#ifdef CONFIG_SMP
2161
void send_cleanup_vector(struct irq_cfg *cfg)
2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176
{
	cpumask_var_t cleanup_mask;

	if (unlikely(!alloc_cpumask_var(&cleanup_mask, GFP_ATOMIC))) {
		unsigned int i;
		for_each_cpu_and(i, cfg->old_domain, cpu_online_mask)
			apic->send_IPI_mask(cpumask_of(i), IRQ_MOVE_CLEANUP_VECTOR);
	} else {
		cpumask_and(cleanup_mask, cfg->old_domain, cpu_online_mask);
		apic->send_IPI_mask(cleanup_mask, IRQ_MOVE_CLEANUP_VECTOR);
		free_cpumask_var(cleanup_mask);
	}
	cfg->move_in_progress = 0;
}

2177
asmlinkage __visible void smp_irq_move_cleanup_interrupt(void)
2178 2179
{
	unsigned vector, me;
2180

2181 2182
	ack_APIC_irq();
	irq_enter();
2183
	exit_idle();
2184 2185 2186

	me = smp_processor_id();
	for (vector = FIRST_EXTERNAL_VECTOR; vector < NR_VECTORS; vector++) {
2187
		int irq;
2188
		unsigned int irr;
2189 2190
		struct irq_desc *desc;
		struct irq_cfg *cfg;
T
Tejun Heo 已提交
2191
		irq = __this_cpu_read(vector_irq[vector]);
2192

2193
		if (irq <= VECTOR_UNDEFINED)
2194 2195
			continue;

2196 2197 2198 2199 2200
		desc = irq_to_desc(irq);
		if (!desc)
			continue;

		cfg = irq_cfg(irq);
2201 2202 2203
		if (!cfg)
			continue;

2204
		raw_spin_lock(&desc->lock);
2205

2206 2207 2208 2209 2210 2211 2212
		/*
		 * Check if the irq migration is in progress. If so, we
		 * haven't received the cleanup request yet for this irq.
		 */
		if (cfg->move_in_progress)
			goto unlock;

2213
		if (vector == cfg->vector && cpumask_test_cpu(me, cfg->domain))
2214 2215
			goto unlock;

2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227
		irr = apic_read(APIC_IRR + (vector / 32 * 0x10));
		/*
		 * Check if the vector that needs to be cleanedup is
		 * registered at the cpu's IRR. If so, then this is not
		 * the best time to clean it up. Lets clean it up in the
		 * next attempt by sending another IRQ_MOVE_CLEANUP_VECTOR
		 * to myself.
		 */
		if (irr  & (1 << (vector % 32))) {
			apic->send_IPI_self(IRQ_MOVE_CLEANUP_VECTOR);
			goto unlock;
		}
T
Tejun Heo 已提交
2228
		__this_cpu_write(vector_irq[vector], -1);
2229
unlock:
2230
		raw_spin_unlock(&desc->lock);
2231 2232 2233 2234 2235
	}

	irq_exit();
}

T
Thomas Gleixner 已提交
2236
static void __irq_complete_move(struct irq_cfg *cfg, unsigned vector)
2237
{
2238
	unsigned me;
2239

2240
	if (likely(!cfg->move_in_progress))
2241 2242 2243
		return;

	me = smp_processor_id();
2244

2245
	if (vector == cfg->vector && cpumask_test_cpu(me, cfg->domain))
2246
		send_cleanup_vector(cfg);
2247
}
2248

T
Thomas Gleixner 已提交
2249
static void irq_complete_move(struct irq_cfg *cfg)
2250
{
T
Thomas Gleixner 已提交
2251
	__irq_complete_move(cfg, ~get_irq_regs()->orig_ax);
2252 2253 2254 2255
}

void irq_force_complete_move(int irq)
{
2256
	struct irq_cfg *cfg = irq_get_chip_data(irq);
2257

2258 2259 2260
	if (!cfg)
		return;

T
Thomas Gleixner 已提交
2261
	__irq_complete_move(cfg, cfg->vector);
2262
}
2263
#else
T
Thomas Gleixner 已提交
2264
static inline void irq_complete_move(struct irq_cfg *cfg) { }
2265
#endif
Y
Yinghai Lu 已提交
2266

2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277
static void __target_IO_APIC_irq(unsigned int irq, unsigned int dest, struct irq_cfg *cfg)
{
	int apic, pin;
	struct irq_pin_list *entry;
	u8 vector = cfg->vector;

	for_each_irq_pin(entry, cfg->irq_2_pin) {
		unsigned int reg;

		apic = entry->apic;
		pin = entry->pin;
2278 2279

		io_apic_write(apic, 0x11 + pin*2, dest);
2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299
		reg = io_apic_read(apic, 0x10 + pin*2);
		reg &= ~IO_APIC_REDIR_VECTOR_MASK;
		reg |= vector;
		io_apic_modify(apic, 0x10 + pin*2, reg);
	}
}

/*
 * Either sets data->affinity to a valid value, and returns
 * ->cpu_mask_to_apicid of that in dest_id, or returns -1 and
 * leaves data->affinity untouched.
 */
int __ioapic_set_affinity(struct irq_data *data, const struct cpumask *mask,
			  unsigned int *dest_id)
{
	struct irq_cfg *cfg = data->chip_data;
	unsigned int irq = data->irq;
	int err;

	if (!config_enabled(CONFIG_SMP))
2300
		return -EPERM;
2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320

	if (!cpumask_intersects(mask, cpu_online_mask))
		return -EINVAL;

	err = assign_irq_vector(irq, cfg, mask);
	if (err)
		return err;

	err = apic->cpu_mask_to_apicid_and(mask, cfg->domain, dest_id);
	if (err) {
		if (assign_irq_vector(irq, cfg, data->affinity))
			pr_err("Failed to recover vector for irq %d\n", irq);
		return err;
	}

	cpumask_copy(data->affinity, mask);

	return 0;
}

2321 2322 2323 2324

int native_ioapic_set_affinity(struct irq_data *data,
			       const struct cpumask *mask,
			       bool force)
2325 2326 2327 2328 2329 2330
{
	unsigned int dest, irq = data->irq;
	unsigned long flags;
	int ret;

	if (!config_enabled(CONFIG_SMP))
2331
		return -EPERM;
2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344

	raw_spin_lock_irqsave(&ioapic_lock, flags);
	ret = __ioapic_set_affinity(data, mask, &dest);
	if (!ret) {
		/* Only the high 8 bits are valid. */
		dest = SET_APIC_LOGICAL_ID(dest);
		__target_IO_APIC_irq(irq, dest, data->chip_data);
		ret = IRQ_SET_MASK_OK_NOCOPY;
	}
	raw_spin_unlock_irqrestore(&ioapic_lock, flags);
	return ret;
}

2345
static void ack_apic_edge(struct irq_data *data)
2346
{
2347
	irq_complete_move(data->chip_data);
2348
	irq_move_irq(data);
2349 2350 2351
	ack_APIC_irq();
}

Y
Yinghai Lu 已提交
2352 2353
atomic_t irq_mis_count;

2354
#ifdef CONFIG_GENERIC_PENDING_IRQ
2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377
static bool io_apic_level_ack_pending(struct irq_cfg *cfg)
{
	struct irq_pin_list *entry;
	unsigned long flags;

	raw_spin_lock_irqsave(&ioapic_lock, flags);
	for_each_irq_pin(entry, cfg->irq_2_pin) {
		unsigned int reg;
		int pin;

		pin = entry->pin;
		reg = io_apic_read(entry->apic, 0x10 + pin*2);
		/* Is the remote IRR bit set? */
		if (reg & IO_APIC_REDIR_REMOTE_IRR) {
			raw_spin_unlock_irqrestore(&ioapic_lock, flags);
			return true;
		}
	}
	raw_spin_unlock_irqrestore(&ioapic_lock, flags);

	return false;
}

2378 2379
static inline bool ioapic_irqd_mask(struct irq_data *data, struct irq_cfg *cfg)
{
2380
	/* If we are moving the irq we need to mask it */
2381
	if (unlikely(irqd_is_setaffinity_pending(data))) {
T
Thomas Gleixner 已提交
2382
		mask_ioapic(cfg);
2383
		return true;
2384
	}
2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431
	return false;
}

static inline void ioapic_irqd_unmask(struct irq_data *data,
				      struct irq_cfg *cfg, bool masked)
{
	if (unlikely(masked)) {
		/* Only migrate the irq if the ack has been received.
		 *
		 * On rare occasions the broadcast level triggered ack gets
		 * delayed going to ioapics, and if we reprogram the
		 * vector while Remote IRR is still set the irq will never
		 * fire again.
		 *
		 * To prevent this scenario we read the Remote IRR bit
		 * of the ioapic.  This has two effects.
		 * - On any sane system the read of the ioapic will
		 *   flush writes (and acks) going to the ioapic from
		 *   this cpu.
		 * - We get to see if the ACK has actually been delivered.
		 *
		 * Based on failed experiments of reprogramming the
		 * ioapic entry from outside of irq context starting
		 * with masking the ioapic entry and then polling until
		 * Remote IRR was clear before reprogramming the
		 * ioapic I don't trust the Remote IRR bit to be
		 * completey accurate.
		 *
		 * However there appears to be no other way to plug
		 * this race, so if the Remote IRR bit is not
		 * accurate and is causing problems then it is a hardware bug
		 * and you can go talk to the chipset vendor about it.
		 */
		if (!io_apic_level_ack_pending(cfg))
			irq_move_masked_irq(data);
		unmask_ioapic(cfg);
	}
}
#else
static inline bool ioapic_irqd_mask(struct irq_data *data, struct irq_cfg *cfg)
{
	return false;
}
static inline void ioapic_irqd_unmask(struct irq_data *data,
				      struct irq_cfg *cfg, bool masked)
{
}
2432 2433
#endif

2434 2435 2436 2437 2438 2439 2440 2441 2442 2443
static void ack_apic_level(struct irq_data *data)
{
	struct irq_cfg *cfg = data->chip_data;
	int i, irq = data->irq;
	unsigned long v;
	bool masked;

	irq_complete_move(cfg);
	masked = ioapic_irqd_mask(data, cfg);

Y
Yinghai Lu 已提交
2444
	/*
2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461
	 * It appears there is an erratum which affects at least version 0x11
	 * of I/O APIC (that's the 82093AA and cores integrated into various
	 * chipsets).  Under certain conditions a level-triggered interrupt is
	 * erroneously delivered as edge-triggered one but the respective IRR
	 * bit gets set nevertheless.  As a result the I/O unit expects an EOI
	 * message but it will never arrive and further interrupts are blocked
	 * from the source.  The exact reason is so far unknown, but the
	 * phenomenon was observed when two consecutive interrupt requests
	 * from a given source get delivered to the same CPU and the source is
	 * temporarily disabled in between.
	 *
	 * A workaround is to simulate an EOI message manually.  We achieve it
	 * by setting the trigger mode to edge and then to level when the edge
	 * trigger mode gets detected in the TMR of a local APIC for a
	 * level-triggered interrupt.  We mask the source for the time of the
	 * operation to prevent an edge-triggered interrupt escaping meanwhile.
	 * The idea is from Manfred Spraul.  --macro
2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474
	 *
	 * Also in the case when cpu goes offline, fixup_irqs() will forward
	 * any unhandled interrupt on the offlined cpu to the new cpu
	 * destination that is handling the corresponding interrupt. This
	 * interrupt forwarding is done via IPI's. Hence, in this case also
	 * level-triggered io-apic interrupt will be seen as an edge
	 * interrupt in the IRR. And we can't rely on the cpu's EOI
	 * to be broadcasted to the IO-APIC's which will clear the remoteIRR
	 * corresponding to the level-triggered interrupt. Hence on IO-APIC's
	 * supporting EOI register, we do an explicit EOI to clear the
	 * remote IRR and on IO-APIC's which don't have an EOI register,
	 * we use the above logic (mask+edge followed by unmask+level) from
	 * Manfred Spraul to clear the remote IRR.
2475
	 */
Y
Yinghai Lu 已提交
2476
	i = cfg->vector;
Y
Yinghai Lu 已提交
2477 2478
	v = apic_read(APIC_TMR + ((i & ~0x1f) >> 1));

2479 2480 2481 2482 2483 2484
	/*
	 * We must acknowledge the irq before we move it or the acknowledge will
	 * not propagate properly.
	 */
	ack_APIC_irq();

2485 2486 2487 2488 2489 2490 2491
	/*
	 * Tail end of clearing remote IRR bit (either by delivering the EOI
	 * message via io-apic EOI register write or simulating it using
	 * mask+edge followed by unnask+level logic) manually when the
	 * level triggered interrupt is seen as the edge triggered interrupt
	 * at the cpu.
	 */
2492 2493 2494
	if (!(v & (1 << (i & 0x1f)))) {
		atomic_inc(&irq_mis_count);

T
Thomas Gleixner 已提交
2495
		eoi_ioapic_irq(irq, cfg);
2496 2497
	}

2498
	ioapic_irqd_unmask(data, cfg, masked);
Y
Yinghai Lu 已提交
2499
}
2500

2501
static struct irq_chip ioapic_chip __read_mostly = {
2502 2503 2504 2505 2506 2507
	.name			= "IO-APIC",
	.irq_startup		= startup_ioapic_irq,
	.irq_mask		= mask_ioapic_irq,
	.irq_unmask		= unmask_ioapic_irq,
	.irq_ack		= ack_apic_edge,
	.irq_eoi		= ack_apic_level,
2508
	.irq_set_affinity	= native_ioapic_set_affinity,
2509
	.irq_retrigger		= ioapic_retrigger_irq,
L
Linus Torvalds 已提交
2510 2511 2512 2513
};

static inline void init_IO_APIC_traps(void)
{
2514
	struct irq_cfg *cfg;
T
Thomas Gleixner 已提交
2515
	unsigned int irq;
L
Linus Torvalds 已提交
2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527

	/*
	 * NOTE! The local APIC isn't very good at handling
	 * multiple interrupts at the same interrupt level.
	 * As the interrupt level is determined by taking the
	 * vector number and shifting that right by 4, we
	 * want to spread these out a bit so that they don't
	 * all fall in the same interrupt level.
	 *
	 * Also, we've got to be careful not to trash gate
	 * 0x80, because int 0x80 is hm, kind of importantish. ;)
	 */
T
Thomas Gleixner 已提交
2528
	for_each_active_irq(irq) {
2529
		cfg = irq_get_chip_data(irq);
2530
		if (IO_APIC_IRQ(irq) && cfg && !cfg->vector) {
L
Linus Torvalds 已提交
2531 2532 2533 2534 2535
			/*
			 * Hmm.. We don't have an entry for this,
			 * so default to an old-fashioned 8259
			 * interrupt if we can..
			 */
2536 2537
			if (irq < legacy_pic->nr_legacy_irqs)
				legacy_pic->make_irq(irq);
2538
			else
L
Linus Torvalds 已提交
2539
				/* Strange. Oh, well.. */
2540
				irq_set_chip(irq, &no_irq_chip);
L
Linus Torvalds 已提交
2541 2542 2543 2544
		}
	}
}

2545 2546 2547
/*
 * The local APIC irq-chip implementation:
 */
L
Linus Torvalds 已提交
2548

2549
static void mask_lapic_irq(struct irq_data *data)
L
Linus Torvalds 已提交
2550 2551 2552 2553
{
	unsigned long v;

	v = apic_read(APIC_LVT0);
2554
	apic_write(APIC_LVT0, v | APIC_LVT_MASKED);
L
Linus Torvalds 已提交
2555 2556
}

2557
static void unmask_lapic_irq(struct irq_data *data)
L
Linus Torvalds 已提交
2558
{
2559
	unsigned long v;
L
Linus Torvalds 已提交
2560

2561
	v = apic_read(APIC_LVT0);
2562
	apic_write(APIC_LVT0, v & ~APIC_LVT_MASKED);
2563
}
L
Linus Torvalds 已提交
2564

2565
static void ack_lapic_irq(struct irq_data *data)
2566 2567 2568 2569
{
	ack_APIC_irq();
}

2570
static struct irq_chip lapic_chip __read_mostly = {
2571
	.name		= "local-APIC",
2572 2573 2574
	.irq_mask	= mask_lapic_irq,
	.irq_unmask	= unmask_lapic_irq,
	.irq_ack	= ack_lapic_irq,
L
Linus Torvalds 已提交
2575 2576
};

2577
static void lapic_register_intr(int irq)
2578
{
2579
	irq_clear_status_flags(irq, IRQ_LEVEL);
2580
	irq_set_chip_and_handler_name(irq, &lapic_chip, handle_edge_irq,
2581 2582 2583
				      "edge");
}

L
Linus Torvalds 已提交
2584 2585 2586 2587 2588 2589 2590
/*
 * This looks a bit hackish but it's about the only one way of sending
 * a few INTA cycles to 8259As and any associated glue logic.  ICR does
 * not support the ExtINT mode, unfortunately.  We need to send these
 * cycles as some i82489DX-based boards have glue logic that keeps the
 * 8259A interrupt line asserted until INTA.  --macro
 */
2591
static inline void __init unlock_ExtINT_logic(void)
L
Linus Torvalds 已提交
2592
{
2593
	int apic, pin, i;
L
Linus Torvalds 已提交
2594 2595 2596
	struct IO_APIC_route_entry entry0, entry1;
	unsigned char save_control, save_freq_select;

2597
	pin  = find_isa_irq_pin(8, mp_INT);
2598 2599 2600 2601
	if (pin == -1) {
		WARN_ON_ONCE(1);
		return;
	}
2602
	apic = find_isa_irq_apic(8, mp_INT);
2603 2604
	if (apic == -1) {
		WARN_ON_ONCE(1);
L
Linus Torvalds 已提交
2605
		return;
2606
	}
L
Linus Torvalds 已提交
2607

2608
	entry0 = ioapic_read_entry(apic, pin);
2609
	clear_IO_APIC_pin(apic, pin);
L
Linus Torvalds 已提交
2610 2611 2612 2613 2614

	memset(&entry1, 0, sizeof(entry1));

	entry1.dest_mode = 0;			/* physical delivery */
	entry1.mask = 0;			/* unmask IRQ now */
2615
	entry1.dest = hard_smp_processor_id();
L
Linus Torvalds 已提交
2616 2617 2618 2619 2620
	entry1.delivery_mode = dest_ExtINT;
	entry1.polarity = entry0.polarity;
	entry1.trigger = 0;
	entry1.vector = 0;

2621
	ioapic_write_entry(apic, pin, entry1);
L
Linus Torvalds 已提交
2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637

	save_control = CMOS_READ(RTC_CONTROL);
	save_freq_select = CMOS_READ(RTC_FREQ_SELECT);
	CMOS_WRITE((save_freq_select & ~RTC_RATE_SELECT) | 0x6,
		   RTC_FREQ_SELECT);
	CMOS_WRITE(save_control | RTC_PIE, RTC_CONTROL);

	i = 100;
	while (i-- > 0) {
		mdelay(10);
		if ((CMOS_READ(RTC_INTR_FLAGS) & RTC_PF) == RTC_PF)
			i -= 10;
	}

	CMOS_WRITE(save_control, RTC_CONTROL);
	CMOS_WRITE(save_freq_select, RTC_FREQ_SELECT);
2638
	clear_IO_APIC_pin(apic, pin);
L
Linus Torvalds 已提交
2639

2640
	ioapic_write_entry(apic, pin, entry0);
L
Linus Torvalds 已提交
2641 2642
}

Y
Yinghai Lu 已提交
2643
static int disable_timer_pin_1 __initdata;
2644
/* Actually the next is obsolete, but keep it for paranoid reasons -AK */
2645
static int __init disable_timer_pin_setup(char *arg)
Y
Yinghai Lu 已提交
2646 2647 2648 2649
{
	disable_timer_pin_1 = 1;
	return 0;
}
2650
early_param("disable_timer_pin_1", disable_timer_pin_setup);
Y
Yinghai Lu 已提交
2651 2652 2653

int timer_through_8259 __initdata;

L
Linus Torvalds 已提交
2654 2655 2656 2657 2658
/*
 * This code may look a bit paranoid, but it's supposed to cooperate with
 * a wide range of boards and BIOS bugs.  Fortunately only the timer IRQ
 * is so screwy.  Thanks to Brian Perkins for testing/hacking this beast
 * fanatically on his truly buggy board.
2659 2660
 *
 * FIXME: really need to revamp this for all platforms.
L
Linus Torvalds 已提交
2661
 */
2662
static inline void __init check_timer(void)
L
Linus Torvalds 已提交
2663
{
2664
	struct irq_cfg *cfg = irq_get_chip_data(0);
2665
	int node = cpu_to_node(0);
2666
	int apic1, pin1, apic2, pin2;
2667
	unsigned long flags;
2668
	int no_pin1 = 0;
2669 2670

	local_irq_save(flags);
2671

L
Linus Torvalds 已提交
2672 2673 2674
	/*
	 * get/set the timer IRQ vector:
	 */
2675
	legacy_pic->mask(0);
2676
	assign_irq_vector(0, cfg, apic->target_cpus());
L
Linus Torvalds 已提交
2677 2678

	/*
2679 2680 2681 2682 2683 2684 2685
	 * As IRQ0 is to be enabled in the 8259A, the virtual
	 * wire has to be disabled in the local APIC.  Also
	 * timer interrupts need to be acknowledged manually in
	 * the 8259A for the i82489DX when using the NMI
	 * watchdog as that APIC treats NMIs as level-triggered.
	 * The AEOI mode will finish them in the 8259A
	 * automatically.
L
Linus Torvalds 已提交
2686
	 */
2687
	apic_write(APIC_LVT0, APIC_LVT_MASKED | APIC_DM_EXTINT);
2688
	legacy_pic->init(1);
L
Linus Torvalds 已提交
2689

2690 2691 2692 2693
	pin1  = find_isa_irq_pin(0, mp_INT);
	apic1 = find_isa_irq_apic(0, mp_INT);
	pin2  = ioapic_i8259.pin;
	apic2 = ioapic_i8259.apic;
L
Linus Torvalds 已提交
2694

2695 2696
	apic_printk(APIC_QUIET, KERN_INFO "..TIMER: vector=0x%02X "
		    "apic1=%d pin1=%d apic2=%d pin2=%d\n",
2697
		    cfg->vector, apic1, pin1, apic2, pin2);
L
Linus Torvalds 已提交
2698

2699 2700 2701 2702 2703 2704 2705 2706
	/*
	 * Some BIOS writers are clueless and report the ExtINTA
	 * I/O APIC input from the cascaded 8259A as the timer
	 * interrupt input.  So just in case, if only one pin
	 * was found above, try it both directly and through the
	 * 8259A.
	 */
	if (pin1 == -1) {
2707
		panic_if_irq_remap("BIOS bug: timer not connected to IO-APIC");
2708 2709 2710 2711 2712 2713 2714 2715
		pin1 = pin2;
		apic1 = apic2;
		no_pin1 = 1;
	} else if (pin2 == -1) {
		pin2 = pin1;
		apic2 = apic1;
	}

L
Linus Torvalds 已提交
2716 2717 2718 2719
	if (pin1 != -1) {
		/*
		 * Ok, does IRQ0 through the IOAPIC work?
		 */
2720
		if (no_pin1) {
2721
			add_pin_to_irq_node(cfg, node, apic1, pin1);
2722
			setup_timer_IRQ0_pin(apic1, pin1, cfg->vector);
Y
Yinghai Lu 已提交
2723
		} else {
2724
			/* for edge trigger, setup_ioapic_irq already
Y
Yinghai Lu 已提交
2725 2726 2727 2728 2729 2730 2731
			 * leave it unmasked.
			 * so only need to unmask if it is level-trigger
			 * do we really have level trigger timer?
			 */
			int idx;
			idx = find_irq_entry(apic1, pin1, mp_INT);
			if (idx != -1 && irq_trigger(idx))
T
Thomas Gleixner 已提交
2732
				unmask_ioapic(cfg);
2733
		}
L
Linus Torvalds 已提交
2734
		if (timer_irq_works()) {
2735 2736
			if (disable_timer_pin_1 > 0)
				clear_IO_APIC_pin(0, pin1);
2737
			goto out;
L
Linus Torvalds 已提交
2738
		}
2739
		panic_if_irq_remap("timer doesn't work through Interrupt-remapped IO-APIC");
Y
Yinghai Lu 已提交
2740
		local_irq_disable();
2741
		clear_IO_APIC_pin(apic1, pin1);
2742
		if (!no_pin1)
2743 2744
			apic_printk(APIC_QUIET, KERN_ERR "..MP-BIOS bug: "
				    "8254 timer not connected to IO-APIC\n");
L
Linus Torvalds 已提交
2745

2746 2747 2748 2749
		apic_printk(APIC_QUIET, KERN_INFO "...trying to set up timer "
			    "(IRQ0) through the 8259A ...\n");
		apic_printk(APIC_QUIET, KERN_INFO
			    "..... (found apic %d pin %d) ...\n", apic2, pin2);
L
Linus Torvalds 已提交
2750 2751 2752
		/*
		 * legacy devices should be connected to IO APIC #0
		 */
2753
		replace_pin_at_irq_node(cfg, node, apic1, pin1, apic2, pin2);
2754
		setup_timer_IRQ0_pin(apic2, pin2, cfg->vector);
2755
		legacy_pic->unmask(0);
L
Linus Torvalds 已提交
2756
		if (timer_irq_works()) {
2757
			apic_printk(APIC_QUIET, KERN_INFO "....... works.\n");
2758
			timer_through_8259 = 1;
2759
			goto out;
L
Linus Torvalds 已提交
2760 2761 2762 2763
		}
		/*
		 * Cleanup, just in case ...
		 */
Y
Yinghai Lu 已提交
2764
		local_irq_disable();
2765
		legacy_pic->mask(0);
2766
		clear_IO_APIC_pin(apic2, pin2);
2767
		apic_printk(APIC_QUIET, KERN_INFO "....... failed.\n");
L
Linus Torvalds 已提交
2768 2769
	}

2770 2771
	apic_printk(APIC_QUIET, KERN_INFO
		    "...trying to set up timer as Virtual Wire IRQ...\n");
L
Linus Torvalds 已提交
2772

2773
	lapic_register_intr(0);
2774
	apic_write(APIC_LVT0, APIC_DM_FIXED | cfg->vector);	/* Fixed mode */
2775
	legacy_pic->unmask(0);
L
Linus Torvalds 已提交
2776 2777

	if (timer_irq_works()) {
2778
		apic_printk(APIC_QUIET, KERN_INFO "..... works.\n");
2779
		goto out;
L
Linus Torvalds 已提交
2780
	}
Y
Yinghai Lu 已提交
2781
	local_irq_disable();
2782
	legacy_pic->mask(0);
2783
	apic_write(APIC_LVT0, APIC_LVT_MASKED | APIC_DM_FIXED | cfg->vector);
2784
	apic_printk(APIC_QUIET, KERN_INFO "..... failed.\n");
L
Linus Torvalds 已提交
2785

2786 2787
	apic_printk(APIC_QUIET, KERN_INFO
		    "...trying to set up timer as ExtINT IRQ...\n");
L
Linus Torvalds 已提交
2788

2789 2790
	legacy_pic->init(0);
	legacy_pic->make_irq(0);
2791
	apic_write(APIC_LVT0, APIC_DM_EXTINT);
L
Linus Torvalds 已提交
2792 2793 2794 2795

	unlock_ExtINT_logic();

	if (timer_irq_works()) {
2796
		apic_printk(APIC_QUIET, KERN_INFO "..... works.\n");
2797
		goto out;
L
Linus Torvalds 已提交
2798
	}
Y
Yinghai Lu 已提交
2799
	local_irq_disable();
2800
	apic_printk(APIC_QUIET, KERN_INFO "..... failed :(.\n");
2801 2802 2803 2804
	if (x2apic_preenabled)
		apic_printk(APIC_QUIET, KERN_INFO
			    "Perhaps problem with the pre-enabled x2apic mode\n"
			    "Try booting with x2apic and interrupt-remapping disabled in the bios.\n");
L
Linus Torvalds 已提交
2805
	panic("IO-APIC + timer doesn't work!  Boot with apic=debug and send a "
2806
		"report.  Then try booting with the 'noapic' option.\n");
2807 2808
out:
	local_irq_restore(flags);
L
Linus Torvalds 已提交
2809 2810 2811
}

/*
2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826
 * Traditionally ISA IRQ2 is the cascade IRQ, and is not available
 * to devices.  However there may be an I/O APIC pin available for
 * this interrupt regardless.  The pin may be left unconnected, but
 * typically it will be reused as an ExtINT cascade interrupt for
 * the master 8259A.  In the MPS case such a pin will normally be
 * reported as an ExtINT interrupt in the MP table.  With ACPI
 * there is no provision for ExtINT interrupts, and in the absence
 * of an override it would be treated as an ordinary ISA I/O APIC
 * interrupt, that is edge-triggered and unmasked by default.  We
 * used to do this, but it caused problems on some systems because
 * of the NMI watchdog and sometimes IRQ0 of the 8254 timer using
 * the same ExtINT cascade interrupt to drive the local APIC of the
 * bootstrap processor.  Therefore we refrain from routing IRQ2 to
 * the I/O APIC in all cases now.  No actual device should request
 * it anyway.  --macro
L
Linus Torvalds 已提交
2827
 */
2828
#define PIC_IRQS	(1UL << PIC_CASCADE_IR)
L
Linus Torvalds 已提交
2829 2830 2831

void __init setup_IO_APIC(void)
{
2832 2833 2834 2835

	/*
	 * calling enable_IO_APIC() is moved to setup_local_APIC for BP
	 */
2836
	io_apic_irqs = legacy_pic->nr_legacy_irqs ? ~PIC_IRQS : ~0UL;
L
Linus Torvalds 已提交
2837

2838
	apic_printk(APIC_VERBOSE, "ENABLING IO-APIC IRQs\n");
T
Thomas Gleixner 已提交
2839
	/*
2840 2841
         * Set up IO-APIC IRQ routing.
         */
2842 2843
	x86_init.mpparse.setup_ioapic_ids();

L
Linus Torvalds 已提交
2844 2845 2846
	sync_Arb_IDs();
	setup_IO_APIC_irqs();
	init_IO_APIC_traps();
2847
	if (legacy_pic->nr_legacy_irqs)
2848
		check_timer();
L
Linus Torvalds 已提交
2849 2850 2851
}

/*
L
Lucas De Marchi 已提交
2852
 *      Called after all the initialization is done. If we didn't find any
2853
 *      APIC bugs then we can allow the modify fast path
L
Linus Torvalds 已提交
2854
 */
2855

L
Linus Torvalds 已提交
2856 2857
static int __init io_apic_bug_finalize(void)
{
T
Thomas Gleixner 已提交
2858 2859 2860
	if (sis_apic_bug == -1)
		sis_apic_bug = 0;
	return 0;
L
Linus Torvalds 已提交
2861 2862 2863 2864
}

late_initcall(io_apic_bug_finalize);

2865
static void resume_ioapic_id(int ioapic_idx)
L
Linus Torvalds 已提交
2866 2867 2868
{
	unsigned long flags;
	union IO_APIC_reg_00 reg_00;
2869

2870
	raw_spin_lock_irqsave(&ioapic_lock, flags);
2871 2872 2873 2874
	reg_00.raw = io_apic_read(ioapic_idx, 0);
	if (reg_00.bits.ID != mpc_ioapic_id(ioapic_idx)) {
		reg_00.bits.ID = mpc_ioapic_id(ioapic_idx);
		io_apic_write(ioapic_idx, 0, reg_00.raw);
L
Linus Torvalds 已提交
2875
	}
2876
	raw_spin_unlock_irqrestore(&ioapic_lock, flags);
2877
}
L
Linus Torvalds 已提交
2878

2879 2880
static void ioapic_resume(void)
{
2881
	int ioapic_idx;
2882

2883 2884
	for (ioapic_idx = nr_ioapics - 1; ioapic_idx >= 0; ioapic_idx--)
		resume_ioapic_id(ioapic_idx);
2885 2886

	restore_ioapic_entries();
L
Linus Torvalds 已提交
2887 2888
}

2889
static struct syscore_ops ioapic_syscore_ops = {
2890
	.suspend = save_ioapic_entries,
L
Linus Torvalds 已提交
2891 2892 2893
	.resume = ioapic_resume,
};

2894
static int __init ioapic_init_ops(void)
L
Linus Torvalds 已提交
2895
{
2896 2897
	register_syscore_ops(&ioapic_syscore_ops);

L
Linus Torvalds 已提交
2898 2899 2900
	return 0;
}

2901
device_initcall(ioapic_init_ops);
L
Linus Torvalds 已提交
2902

2903
/*
2904
 * Dynamic irq allocate and deallocation. Should be replaced by irq domains!
2905
 */
2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938
int arch_setup_hwirq(unsigned int irq, int node)
{
	struct irq_cfg *cfg;
	unsigned long flags;
	int ret;

	cfg = alloc_irq_cfg(irq, node);
	if (!cfg)
		return -ENOMEM;

	raw_spin_lock_irqsave(&vector_lock, flags);
	ret = __assign_irq_vector(irq, cfg, apic->target_cpus());
	raw_spin_unlock_irqrestore(&vector_lock, flags);

	if (!ret)
		irq_set_chip_data(irq, cfg);
	else
		free_irq_cfg(irq, cfg);
	return ret;
}

void arch_teardown_hwirq(unsigned int irq)
{
	struct irq_cfg *cfg = irq_get_chip_data(irq);
	unsigned long flags;

	free_remapped_irq(irq);
	raw_spin_lock_irqsave(&vector_lock, flags);
	__clear_irq_vector(irq, cfg);
	raw_spin_unlock_irqrestore(&vector_lock, flags);
	free_irq_cfg(irq, cfg);
}

2939
/*
S
Simon Arlott 已提交
2940
 * MSI message composition
2941
 */
2942 2943 2944
void native_compose_msi_msg(struct pci_dev *pdev,
			    unsigned int irq, unsigned int dest,
			    struct msi_msg *msg, u8 hpet_id)
2945
{
2946
	struct irq_cfg *cfg = irq_cfg(irq);
2947

2948
	msg->address_hi = MSI_ADDR_BASE_HI;
2949

2950
	if (x2apic_enabled())
2951
		msg->address_hi |= MSI_ADDR_EXT_DEST_ID(dest);
2952

2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969
	msg->address_lo =
		MSI_ADDR_BASE_LO |
		((apic->irq_dest_mode == 0) ?
			MSI_ADDR_DEST_MODE_PHYSICAL:
			MSI_ADDR_DEST_MODE_LOGICAL) |
		((apic->irq_delivery_mode != dest_LowestPrio) ?
			MSI_ADDR_REDIRECTION_CPU:
			MSI_ADDR_REDIRECTION_LOWPRI) |
		MSI_ADDR_DEST_ID(dest);

	msg->data =
		MSI_DATA_TRIGGER_EDGE |
		MSI_DATA_LEVEL_ASSERT |
		((apic->irq_delivery_mode != dest_LowestPrio) ?
			MSI_DATA_DELIVERY_FIXED:
			MSI_DATA_DELIVERY_LOWPRI) |
		MSI_DATA_VECTOR(cfg->vector);
2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993
}

#ifdef CONFIG_PCI_MSI
static int msi_compose_msg(struct pci_dev *pdev, unsigned int irq,
			   struct msi_msg *msg, u8 hpet_id)
{
	struct irq_cfg *cfg;
	int err;
	unsigned dest;

	if (disable_apic)
		return -ENXIO;

	cfg = irq_cfg(irq);
	err = assign_irq_vector(irq, cfg, apic->target_cpus());
	if (err)
		return err;

	err = apic->cpu_mask_to_apicid_and(cfg->domain,
					   apic->target_cpus(), &dest);
	if (err)
		return err;

	x86_msi.compose_msi_msg(pdev, irq, dest, msg, hpet_id);
2994

2995
	return 0;
2996 2997
}

2998 2999
static int
msi_set_affinity(struct irq_data *data, const struct cpumask *mask, bool force)
3000
{
3001
	struct irq_cfg *cfg = data->chip_data;
3002 3003
	struct msi_msg msg;
	unsigned int dest;
3004
	int ret;
3005

3006 3007 3008
	ret = __ioapic_set_affinity(data, mask, &dest);
	if (ret)
		return ret;
3009

3010
	__get_cached_msi_msg(data->msi_desc, &msg);
3011 3012

	msg.data &= ~MSI_DATA_VECTOR_MASK;
3013
	msg.data |= MSI_DATA_VECTOR(cfg->vector);
3014 3015 3016
	msg.address_lo &= ~MSI_ADDR_DEST_ID_MASK;
	msg.address_lo |= MSI_ADDR_DEST_ID(dest);

3017
	__write_msi_msg(data->msi_desc, &msg);
3018

3019
	return IRQ_SET_MASK_OK_NOCOPY;
3020 3021
}

3022 3023 3024 3025 3026
/*
 * IRQ Chip for MSI PCI/PCI-X/PCI-Express Devices,
 * which implement the MSI or MSI-X Capability Structure.
 */
static struct irq_chip msi_chip = {
3027 3028 3029 3030 3031 3032
	.name			= "PCI-MSI",
	.irq_unmask		= unmask_msi_irq,
	.irq_mask		= mask_msi_irq,
	.irq_ack		= ack_apic_edge,
	.irq_set_affinity	= msi_set_affinity,
	.irq_retrigger		= ioapic_retrigger_irq,
3033 3034
};

3035 3036
int setup_msi_irq(struct pci_dev *dev, struct msi_desc *msidesc,
		  unsigned int irq_base, unsigned int irq_offset)
3037
{
3038
	struct irq_chip *chip = &msi_chip;
3039
	struct msi_msg msg;
3040
	unsigned int irq = irq_base + irq_offset;
3041
	int ret;
3042

3043
	ret = msi_compose_msg(dev, irq, &msg, -1);
3044 3045 3046
	if (ret < 0)
		return ret;

3047 3048 3049 3050 3051 3052 3053 3054
	irq_set_msi_desc_off(irq_base, irq_offset, msidesc);

	/*
	 * MSI-X message is written per-IRQ, the offset is always 0.
	 * MSI message denotes a contiguous group of IRQs, written for 0th IRQ.
	 */
	if (!irq_offset)
		write_msi_msg(irq, &msg);
3055

3056
	setup_remapped_irq(irq, irq_get_chip_data(irq), chip);
3057 3058

	irq_set_chip_and_handler_name(irq, chip, handle_edge_irq, "edge");
3059

Y
Yinghai Lu 已提交
3060 3061
	dev_printk(KERN_DEBUG, &dev->dev, "irq %d for MSI/MSI-X\n", irq);

3062 3063 3064
	return 0;
}

3065
int native_setup_msi_irqs(struct pci_dev *dev, int nvec, int type)
3066
{
3067
	struct msi_desc *msidesc;
3068
	unsigned int irq;
3069 3070 3071 3072 3073
	int node, ret;

	/* Multiple MSI vectors only supported with interrupt remapping */
	if (type == PCI_CAP_ID_MSI && nvec > 1)
		return 1;
3074

3075
	node = dev_to_node(&dev->dev);
3076

3077
	list_for_each_entry(msidesc, &dev->msi_list, list) {
3078 3079
		irq = irq_alloc_hwirq(node);
		if (!irq)
3080
			return -ENOSPC;
3081

3082
		ret = setup_msi_irq(dev, msidesc, irq, 0);
3083 3084 3085 3086 3087
		if (ret < 0) {
			irq_free_hwirq(irq);
			return ret;
		}

3088 3089
	}
	return 0;
3090 3091
}

S
Stefano Stabellini 已提交
3092
void native_teardown_msi_irq(unsigned int irq)
3093
{
3094
	irq_free_hwirq(irq);
3095 3096
}

3097
#ifdef CONFIG_DMAR_TABLE
3098 3099 3100
static int
dmar_msi_set_affinity(struct irq_data *data, const struct cpumask *mask,
		      bool force)
3101
{
3102 3103
	struct irq_cfg *cfg = data->chip_data;
	unsigned int dest, irq = data->irq;
3104
	struct msi_msg msg;
3105
	int ret;
3106

3107 3108 3109
	ret = __ioapic_set_affinity(data, mask, &dest);
	if (ret)
		return ret;
3110 3111 3112 3113 3114 3115 3116

	dmar_msi_read(irq, &msg);

	msg.data &= ~MSI_DATA_VECTOR_MASK;
	msg.data |= MSI_DATA_VECTOR(cfg->vector);
	msg.address_lo &= ~MSI_ADDR_DEST_ID_MASK;
	msg.address_lo |= MSI_ADDR_DEST_ID(dest);
3117
	msg.address_hi = MSI_ADDR_BASE_HI | MSI_ADDR_EXT_DEST_ID(dest);
3118 3119

	dmar_msi_write(irq, &msg);
3120

3121
	return IRQ_SET_MASK_OK_NOCOPY;
3122
}
Y
Yinghai Lu 已提交
3123

3124
static struct irq_chip dmar_msi_type = {
3125 3126 3127 3128 3129 3130
	.name			= "DMAR_MSI",
	.irq_unmask		= dmar_msi_unmask,
	.irq_mask		= dmar_msi_mask,
	.irq_ack		= ack_apic_edge,
	.irq_set_affinity	= dmar_msi_set_affinity,
	.irq_retrigger		= ioapic_retrigger_irq,
3131 3132 3133 3134 3135 3136
};

int arch_setup_dmar_msi(unsigned int irq)
{
	int ret;
	struct msi_msg msg;
3137

3138
	ret = msi_compose_msg(NULL, irq, &msg, -1);
3139 3140 3141
	if (ret < 0)
		return ret;
	dmar_msi_write(irq, &msg);
3142 3143
	irq_set_chip_and_handler_name(irq, &dmar_msi_type, handle_edge_irq,
				      "edge");
3144 3145 3146 3147
	return 0;
}
#endif

3148 3149
#ifdef CONFIG_HPET_TIMER

3150 3151
static int hpet_msi_set_affinity(struct irq_data *data,
				 const struct cpumask *mask, bool force)
3152
{
3153
	struct irq_cfg *cfg = data->chip_data;
3154 3155
	struct msi_msg msg;
	unsigned int dest;
3156
	int ret;
3157

3158 3159 3160
	ret = __ioapic_set_affinity(data, mask, &dest);
	if (ret)
		return ret;
3161

3162
	hpet_msi_read(data->handler_data, &msg);
3163 3164 3165 3166 3167 3168

	msg.data &= ~MSI_DATA_VECTOR_MASK;
	msg.data |= MSI_DATA_VECTOR(cfg->vector);
	msg.address_lo &= ~MSI_ADDR_DEST_ID_MASK;
	msg.address_lo |= MSI_ADDR_DEST_ID(dest);

3169
	hpet_msi_write(data->handler_data, &msg);
3170

3171
	return IRQ_SET_MASK_OK_NOCOPY;
3172
}
Y
Yinghai Lu 已提交
3173

3174
static struct irq_chip hpet_msi_type = {
3175
	.name = "HPET_MSI",
3176 3177
	.irq_unmask = hpet_msi_unmask,
	.irq_mask = hpet_msi_mask,
3178
	.irq_ack = ack_apic_edge,
3179
	.irq_set_affinity = hpet_msi_set_affinity,
3180
	.irq_retrigger = ioapic_retrigger_irq,
3181 3182
};

3183
int default_setup_hpet_msi(unsigned int irq, unsigned int id)
3184
{
3185
	struct irq_chip *chip = &hpet_msi_type;
3186
	struct msi_msg msg;
3187
	int ret;
3188

3189
	ret = msi_compose_msg(NULL, irq, &msg, id);
3190 3191 3192
	if (ret < 0)
		return ret;

3193
	hpet_msi_write(irq_get_handler_data(irq), &msg);
3194
	irq_set_status_flags(irq, IRQ_MOVE_PCNTXT);
3195
	setup_remapped_irq(irq, irq_get_chip_data(irq), chip);
Y
Yinghai Lu 已提交
3196

3197
	irq_set_chip_and_handler_name(irq, chip, handle_edge_irq, "edge");
3198 3199 3200 3201
	return 0;
}
#endif

3202
#endif /* CONFIG_PCI_MSI */
3203 3204 3205 3206 3207
/*
 * Hypertransport interrupt support
 */
#ifdef CONFIG_HT_IRQ

3208
static void target_ht_irq(unsigned int irq, unsigned int dest, u8 vector)
3209
{
3210 3211
	struct ht_irq_msg msg;
	fetch_ht_irq_msg(irq, &msg);
3212

3213
	msg.address_lo &= ~(HT_IRQ_LOW_VECTOR_MASK | HT_IRQ_LOW_DEST_ID_MASK);
3214
	msg.address_hi &= ~(HT_IRQ_HIGH_DEST_ID_MASK);
3215

3216
	msg.address_lo |= HT_IRQ_LOW_VECTOR(vector) | HT_IRQ_LOW_DEST_ID(dest);
3217
	msg.address_hi |= HT_IRQ_HIGH_DEST_ID(dest);
3218

3219
	write_ht_irq_msg(irq, &msg);
3220 3221
}

3222 3223
static int
ht_set_affinity(struct irq_data *data, const struct cpumask *mask, bool force)
3224
{
3225
	struct irq_cfg *cfg = data->chip_data;
3226
	unsigned int dest;
3227
	int ret;
3228

3229 3230 3231
	ret = __ioapic_set_affinity(data, mask, &dest);
	if (ret)
		return ret;
3232

3233
	target_ht_irq(data->irq, dest, cfg->vector);
3234
	return IRQ_SET_MASK_OK_NOCOPY;
3235
}
Y
Yinghai Lu 已提交
3236

3237
static struct irq_chip ht_irq_chip = {
3238 3239 3240 3241 3242 3243
	.name			= "PCI-HT",
	.irq_mask		= mask_ht_irq,
	.irq_unmask		= unmask_ht_irq,
	.irq_ack		= ack_apic_edge,
	.irq_set_affinity	= ht_set_affinity,
	.irq_retrigger		= ioapic_retrigger_irq,
3244 3245 3246 3247
};

int arch_setup_ht_irq(unsigned int irq, struct pci_dev *dev)
{
3248
	struct irq_cfg *cfg;
3249 3250
	struct ht_irq_msg msg;
	unsigned dest;
3251
	int err;
3252

J
Jan Beulich 已提交
3253 3254 3255
	if (disable_apic)
		return -ENXIO;

Y
Yinghai Lu 已提交
3256
	cfg = irq_cfg(irq);
3257
	err = assign_irq_vector(irq, cfg, apic->target_cpus());
3258 3259
	if (err)
		return err;
3260

3261 3262 3263 3264
	err = apic->cpu_mask_to_apicid_and(cfg->domain,
					   apic->target_cpus(), &dest);
	if (err)
		return err;
3265

3266
	msg.address_hi = HT_IRQ_HIGH_DEST_ID(dest);
3267

3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279
	msg.address_lo =
		HT_IRQ_LOW_BASE |
		HT_IRQ_LOW_DEST_ID(dest) |
		HT_IRQ_LOW_VECTOR(cfg->vector) |
		((apic->irq_dest_mode == 0) ?
			HT_IRQ_LOW_DM_PHYSICAL :
			HT_IRQ_LOW_DM_LOGICAL) |
		HT_IRQ_LOW_RQEOI_EDGE |
		((apic->irq_delivery_mode != dest_LowestPrio) ?
			HT_IRQ_LOW_MT_FIXED :
			HT_IRQ_LOW_MT_ARBITRATED) |
		HT_IRQ_LOW_IRQ_MASKED;
3280

3281
	write_ht_irq_msg(irq, &msg);
3282

3283 3284
	irq_set_chip_and_handler_name(irq, &ht_irq_chip,
				      handle_edge_irq, "edge");
3285

3286
	dev_printk(KERN_DEBUG, &dev->dev, "irq %d for HT\n", irq);
Y
Yinghai Lu 已提交
3287

3288
	return 0;
3289 3290 3291
}
#endif /* CONFIG_HT_IRQ */

3292
static int
3293 3294 3295 3296 3297 3298 3299 3300 3301
io_apic_setup_irq_pin(unsigned int irq, int node, struct io_apic_irq_attr *attr)
{
	struct irq_cfg *cfg = alloc_irq_and_cfg_at(irq, node);
	int ret;

	if (!cfg)
		return -EINVAL;
	ret = __add_pin_to_irq_node(cfg, node, attr->ioapic, attr->ioapic_pin);
	if (!ret)
3302
		setup_ioapic_irq(irq, cfg, attr);
3303 3304 3305
	return ret;
}

3306 3307
int io_apic_setup_irq_pin_once(unsigned int irq, int node,
			       struct io_apic_irq_attr *attr)
3308
{
3309
	unsigned int ioapic_idx = attr->ioapic, pin = attr->ioapic_pin;
3310
	int ret;
3311
	struct IO_APIC_route_entry orig_entry;
3312 3313

	/* Avoid redundant programming */
3314
	if (test_bit(pin, ioapics[ioapic_idx].pin_programmed)) {
3315 3316 3317 3318 3319
		pr_debug("Pin %d-%d already programmed\n", mpc_ioapic_id(ioapic_idx), pin);
		orig_entry = ioapic_read_entry(attr->ioapic, pin);
		if (attr->trigger == orig_entry.trigger && attr->polarity == orig_entry.polarity)
			return 0;
		return -EBUSY;
3320 3321 3322
	}
	ret = io_apic_setup_irq_pin(irq, node, attr);
	if (!ret)
3323
		set_bit(pin, ioapics[ioapic_idx].pin_programmed);
3324 3325 3326
	return ret;
}

3327
static int __init io_apic_get_redir_entries(int ioapic)
3328 3329 3330 3331
{
	union IO_APIC_reg_01	reg_01;
	unsigned long flags;

3332
	raw_spin_lock_irqsave(&ioapic_lock, flags);
3333
	reg_01.raw = io_apic_read(ioapic, 1);
3334
	raw_spin_unlock_irqrestore(&ioapic_lock, flags);
3335

3336 3337 3338 3339 3340
	/* The register returns the maximum index redir index
	 * supported, which is one less than the total number of redir
	 * entries.
	 */
	return reg_01.bits.entries + 1;
3341 3342
}

3343
static void __init probe_nr_irqs_gsi(void)
3344
{
3345
	int nr;
3346

3347
	nr = gsi_top + NR_IRQS_LEGACY;
3348
	if (nr > nr_irqs_gsi)
3349
		nr_irqs_gsi = nr;
3350 3351

	printk(KERN_DEBUG "nr_irqs_gsi: %d\n", nr_irqs_gsi);
3352 3353
}

3354 3355 3356 3357 3358
unsigned int arch_dynirq_lower_bound(unsigned int from)
{
	return from < nr_irqs_gsi ? nr_irqs_gsi : from;
}

Y
Yinghai Lu 已提交
3359 3360 3361 3362
int __init arch_probe_nr_irqs(void)
{
	int nr;

Y
Yinghai Lu 已提交
3363 3364
	if (nr_irqs > (NR_VECTORS * nr_cpu_ids))
		nr_irqs = NR_VECTORS * nr_cpu_ids;
Y
Yinghai Lu 已提交
3365

Y
Yinghai Lu 已提交
3366 3367 3368 3369 3370 3371 3372 3373
	nr = nr_irqs_gsi + 8 * nr_cpu_ids;
#if defined(CONFIG_PCI_MSI) || defined(CONFIG_HT_IRQ)
	/*
	 * for MSI and HT dyn irq
	 */
	nr += nr_irqs_gsi * 16;
#endif
	if (nr < nr_irqs)
Y
Yinghai Lu 已提交
3374 3375
		nr_irqs = nr;

3376
	return NR_IRQS_LEGACY;
Y
Yinghai Lu 已提交
3377 3378
}

3379 3380
int io_apic_set_pci_routing(struct device *dev, int irq,
			    struct io_apic_irq_attr *irq_attr)
3381 3382 3383 3384 3385
{
	int node;

	if (!IO_APIC_IRQ(irq)) {
		apic_printk(APIC_QUIET,KERN_ERR "IOAPIC[%d]: Invalid reference to IRQ 0\n",
3386
			    irq_attr->ioapic);
3387 3388 3389
		return -EINVAL;
	}

3390
	node = dev ? dev_to_node(dev) : cpu_to_node(0);
3391

3392
	return io_apic_setup_irq_pin_once(irq, node, irq_attr);
3393 3394
}

3395
#ifdef CONFIG_X86_32
3396
static int __init io_apic_get_unique_id(int ioapic, int apic_id)
L
Linus Torvalds 已提交
3397 3398 3399 3400 3401 3402 3403 3404
{
	union IO_APIC_reg_00 reg_00;
	static physid_mask_t apic_id_map = PHYSID_MASK_NONE;
	physid_mask_t tmp;
	unsigned long flags;
	int i = 0;

	/*
3405 3406
	 * The P4 platform supports up to 256 APIC IDs on two separate APIC
	 * buses (one for LAPICs, one for IOAPICs), where predecessors only
L
Linus Torvalds 已提交
3407
	 * supports up to 16 on one shared APIC bus.
3408
	 *
L
Linus Torvalds 已提交
3409 3410 3411 3412 3413
	 * TBD: Expand LAPIC/IOAPIC support on P4-class systems to take full
	 *      advantage of new APIC bus architecture.
	 */

	if (physids_empty(apic_id_map))
3414
		apic->ioapic_phys_id_map(&phys_cpu_present_map, &apic_id_map);
L
Linus Torvalds 已提交
3415

3416
	raw_spin_lock_irqsave(&ioapic_lock, flags);
L
Linus Torvalds 已提交
3417
	reg_00.raw = io_apic_read(ioapic, 0);
3418
	raw_spin_unlock_irqrestore(&ioapic_lock, flags);
L
Linus Torvalds 已提交
3419 3420 3421 3422 3423 3424 3425 3426

	if (apic_id >= get_physical_broadcast()) {
		printk(KERN_WARNING "IOAPIC[%d]: Invalid apic_id %d, trying "
			"%d\n", ioapic, apic_id, reg_00.bits.ID);
		apic_id = reg_00.bits.ID;
	}

	/*
3427
	 * Every APIC in a system must have a unique ID or we get lots of nice
L
Linus Torvalds 已提交
3428 3429
	 * 'stuck on smp_invalidate_needed IPI wait' messages.
	 */
3430
	if (apic->check_apicid_used(&apic_id_map, apic_id)) {
L
Linus Torvalds 已提交
3431 3432

		for (i = 0; i < get_physical_broadcast(); i++) {
3433
			if (!apic->check_apicid_used(&apic_id_map, i))
L
Linus Torvalds 已提交
3434 3435 3436 3437 3438 3439 3440 3441 3442 3443
				break;
		}

		if (i == get_physical_broadcast())
			panic("Max apic_id exceeded!\n");

		printk(KERN_WARNING "IOAPIC[%d]: apic_id %d already used, "
			"trying %d\n", ioapic, apic_id, i);

		apic_id = i;
3444
	}
L
Linus Torvalds 已提交
3445

3446
	apic->apicid_to_cpu_present(apic_id, &tmp);
L
Linus Torvalds 已提交
3447 3448 3449 3450 3451
	physids_or(apic_id_map, apic_id_map, tmp);

	if (reg_00.bits.ID != apic_id) {
		reg_00.bits.ID = apic_id;

3452
		raw_spin_lock_irqsave(&ioapic_lock, flags);
L
Linus Torvalds 已提交
3453 3454
		io_apic_write(ioapic, 0, reg_00.raw);
		reg_00.raw = io_apic_read(ioapic, 0);
3455
		raw_spin_unlock_irqrestore(&ioapic_lock, flags);
L
Linus Torvalds 已提交
3456 3457

		/* Sanity check */
3458
		if (reg_00.bits.ID != apic_id) {
3459 3460
			pr_err("IOAPIC[%d]: Unable to change apic_id!\n",
			       ioapic);
3461 3462
			return -1;
		}
L
Linus Torvalds 已提交
3463 3464 3465 3466 3467 3468 3469
	}

	apic_printk(APIC_VERBOSE, KERN_INFO
			"IOAPIC[%d]: Assigned apic_id %d\n", ioapic, apic_id);

	return apic_id;
}
3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486

static u8 __init io_apic_unique_id(u8 id)
{
	if ((boot_cpu_data.x86_vendor == X86_VENDOR_INTEL) &&
	    !APIC_XAPIC(apic_version[boot_cpu_physical_apicid]))
		return io_apic_get_unique_id(nr_ioapics, id);
	else
		return id;
}
#else
static u8 __init io_apic_unique_id(u8 id)
{
	int i;
	DECLARE_BITMAP(used, 256);

	bitmap_zero(used, 256);
	for (i = 0; i < nr_ioapics; i++) {
3487
		__set_bit(mpc_ioapic_id(i), used);
3488 3489 3490 3491 3492
	}
	if (!test_bit(id, used))
		return id;
	return find_first_zero_bit(used, 256);
}
3493
#endif
L
Linus Torvalds 已提交
3494

3495
static int __init io_apic_get_version(int ioapic)
L
Linus Torvalds 已提交
3496 3497 3498 3499
{
	union IO_APIC_reg_01	reg_01;
	unsigned long flags;

3500
	raw_spin_lock_irqsave(&ioapic_lock, flags);
L
Linus Torvalds 已提交
3501
	reg_01.raw = io_apic_read(ioapic, 1);
3502
	raw_spin_unlock_irqrestore(&ioapic_lock, flags);
L
Linus Torvalds 已提交
3503 3504 3505 3506

	return reg_01.bits.version;
}

3507
int acpi_get_override_irq(u32 gsi, int *trigger, int *polarity)
3508
{
3509
	int ioapic, pin, idx;
3510 3511 3512 3513

	if (skip_ioapic_setup)
		return -1;

3514 3515
	ioapic = mp_find_ioapic(gsi);
	if (ioapic < 0)
3516 3517
		return -1;

3518 3519 3520 3521 3522 3523
	pin = mp_find_ioapic_pin(ioapic, gsi);
	if (pin < 0)
		return -1;

	idx = find_irq_entry(ioapic, pin, mp_INT);
	if (idx < 0)
3524 3525
		return -1;

3526 3527
	*trigger = irq_trigger(idx);
	*polarity = irq_polarity(idx);
3528 3529 3530
	return 0;
}

3531 3532 3533
/*
 * This function currently is only a helper for the i386 smp boot process where
 * we need to reprogram the ioredtbls to cater for the cpus which have come online
3534
 * so mask in all cases should simply be apic->target_cpus()
3535 3536 3537 3538
 */
#ifdef CONFIG_SMP
void __init setup_ioapic_dest(void)
{
E
Eric W. Biederman 已提交
3539
	int pin, ioapic, irq, irq_entry;
3540
	const struct cpumask *mask;
3541
	struct irq_data *idata;
3542 3543 3544 3545

	if (skip_ioapic_setup == 1)
		return;

E
Eric W. Biederman 已提交
3546
	for (ioapic = 0; ioapic < nr_ioapics; ioapic++)
S
Suresh Siddha 已提交
3547
	for (pin = 0; pin < ioapics[ioapic].nr_registers; pin++) {
3548 3549 3550 3551
		irq_entry = find_irq_entry(ioapic, pin, mp_INT);
		if (irq_entry == -1)
			continue;
		irq = pin_2_irq(irq_entry, ioapic, pin);
3552

E
Eric W. Biederman 已提交
3553 3554 3555
		if ((ioapic > 0) && (irq > 16))
			continue;

3556
		idata = irq_get_irq_data(irq);
3557

3558 3559 3560
		/*
		 * Honour affinities which have been set in early boot
		 */
3561 3562
		if (!irqd_can_balance(idata) || irqd_affinity_was_set(idata))
			mask = idata->affinity;
3563 3564
		else
			mask = apic->target_cpus();
3565

3566
		x86_io_apic_ops.set_affinity(idata, mask, false);
3567
	}
3568

3569 3570 3571
}
#endif

3572 3573 3574 3575
#define IOAPIC_RESOURCE_NAME_SIZE 11

static struct resource *ioapic_resources;

3576
static struct resource * __init ioapic_setup_resources(int nr_ioapics)
3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591
{
	unsigned long n;
	struct resource *res;
	char *mem;
	int i;

	if (nr_ioapics <= 0)
		return NULL;

	n = IOAPIC_RESOURCE_NAME_SIZE + sizeof(struct resource);
	n *= nr_ioapics;

	mem = alloc_bootmem(n);
	res = (void *)mem;

3592
	mem += sizeof(struct resource) * nr_ioapics;
3593

3594 3595 3596
	for (i = 0; i < nr_ioapics; i++) {
		res[i].name = mem;
		res[i].flags = IORESOURCE_MEM | IORESOURCE_BUSY;
3597
		snprintf(mem, IOAPIC_RESOURCE_NAME_SIZE, "IOAPIC %u", i);
3598
		mem += IOAPIC_RESOURCE_NAME_SIZE;
3599 3600 3601 3602 3603 3604 3605
	}

	ioapic_resources = res;

	return res;
}

3606
void __init native_io_apic_init_mappings(void)
3607 3608
{
	unsigned long ioapic_phys, idx = FIX_IO_APIC_BASE_0;
3609
	struct resource *ioapic_res;
T
Thomas Gleixner 已提交
3610
	int i;
3611

3612
	ioapic_res = ioapic_setup_resources(nr_ioapics);
3613 3614
	for (i = 0; i < nr_ioapics; i++) {
		if (smp_found_config) {
3615
			ioapic_phys = mpc_ioapic_addr(i);
3616
#ifdef CONFIG_X86_32
T
Thomas Gleixner 已提交
3617 3618 3619 3620 3621 3622 3623 3624 3625
			if (!ioapic_phys) {
				printk(KERN_ERR
				       "WARNING: bogus zero IO-APIC "
				       "address found in MPTABLE, "
				       "disabling IO/APIC support!\n");
				smp_found_config = 0;
				skip_ioapic_setup = 1;
				goto fake_ioapic_page;
			}
3626
#endif
3627
		} else {
3628
#ifdef CONFIG_X86_32
3629
fake_ioapic_page:
3630
#endif
3631
			ioapic_phys = (unsigned long)alloc_bootmem_pages(PAGE_SIZE);
3632 3633 3634
			ioapic_phys = __pa(ioapic_phys);
		}
		set_fixmap_nocache(idx, ioapic_phys);
3635 3636 3637
		apic_printk(APIC_VERBOSE, "mapped IOAPIC to %08lx (%08lx)\n",
			__fix_to_virt(idx) + (ioapic_phys & ~PAGE_MASK),
			ioapic_phys);
3638
		idx++;
3639

3640
		ioapic_res->start = ioapic_phys;
3641
		ioapic_res->end = ioapic_phys + IO_APIC_SLOT_SIZE - 1;
3642
		ioapic_res++;
3643
	}
3644 3645

	probe_nr_irqs_gsi();
3646 3647
}

3648
void __init ioapic_insert_resources(void)
3649 3650 3651 3652 3653
{
	int i;
	struct resource *r = ioapic_resources;

	if (!r) {
3654
		if (nr_ioapics > 0)
3655 3656
			printk(KERN_ERR
				"IO APIC resources couldn't be allocated.\n");
3657
		return;
3658 3659 3660 3661 3662 3663 3664
	}

	for (i = 0; i < nr_ioapics; i++) {
		insert_resource(&iomem_resource, r);
		r++;
	}
}
3665

3666
int mp_find_ioapic(u32 gsi)
3667 3668 3669
{
	int i = 0;

3670 3671 3672
	if (nr_ioapics == 0)
		return -1;

3673 3674
	/* Find the IOAPIC that manages this GSI. */
	for (i = 0; i < nr_ioapics; i++) {
3675 3676 3677
		struct mp_ioapic_gsi *gsi_cfg = mp_ioapic_gsi_routing(i);
		if ((gsi >= gsi_cfg->gsi_base)
		    && (gsi <= gsi_cfg->gsi_end))
3678 3679
			return i;
	}
3680

3681 3682 3683 3684
	printk(KERN_ERR "ERROR: Unable to locate IOAPIC for GSI %d\n", gsi);
	return -1;
}

3685
int mp_find_ioapic_pin(int ioapic, u32 gsi)
3686
{
3687 3688
	struct mp_ioapic_gsi *gsi_cfg;

3689 3690
	if (WARN_ON(ioapic == -1))
		return -1;
3691 3692 3693

	gsi_cfg = mp_ioapic_gsi_routing(ioapic);
	if (WARN_ON(gsi > gsi_cfg->gsi_end))
3694 3695
		return -1;

3696
	return gsi - gsi_cfg->gsi_base;
3697 3698
}

3699
static __init int bad_ioapic(unsigned long address)
3700 3701
{
	if (nr_ioapics >= MAX_IO_APICS) {
3702 3703
		pr_warn("WARNING: Max # of I/O APICs (%d) exceeded (found %d), skipping\n",
			MAX_IO_APICS, nr_ioapics);
3704 3705 3706
		return 1;
	}
	if (!address) {
3707
		pr_warn("WARNING: Bogus (zero) I/O APIC address found in table, skipping!\n");
3708 3709
		return 1;
	}
3710 3711 3712
	return 0;
}

3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731
static __init int bad_ioapic_register(int idx)
{
	union IO_APIC_reg_00 reg_00;
	union IO_APIC_reg_01 reg_01;
	union IO_APIC_reg_02 reg_02;

	reg_00.raw = io_apic_read(idx, 0);
	reg_01.raw = io_apic_read(idx, 1);
	reg_02.raw = io_apic_read(idx, 2);

	if (reg_00.raw == -1 && reg_01.raw == -1 && reg_02.raw == -1) {
		pr_warn("I/O APIC 0x%x registers return all ones, skipping!\n",
			mpc_ioapic_addr(idx));
		return 1;
	}

	return 0;
}

3732 3733 3734
void __init mp_register_ioapic(int id, u32 address, u32 gsi_base)
{
	int idx = 0;
3735
	int entries;
3736
	struct mp_ioapic_gsi *gsi_cfg;
3737 3738 3739 3740 3741 3742

	if (bad_ioapic(address))
		return;

	idx = nr_ioapics;

3743 3744 3745
	ioapics[idx].mp_config.type = MP_IOAPIC;
	ioapics[idx].mp_config.flags = MPC_APIC_USABLE;
	ioapics[idx].mp_config.apicaddr = address;
3746 3747

	set_fixmap_nocache(FIX_IO_APIC_BASE_0 + idx, address);
3748 3749 3750 3751 3752 3753

	if (bad_ioapic_register(idx)) {
		clear_fixmap(FIX_IO_APIC_BASE_0 + idx);
		return;
	}

3754 3755
	ioapics[idx].mp_config.apicid = io_apic_unique_id(id);
	ioapics[idx].mp_config.apicver = io_apic_get_version(idx);
3756 3757 3758 3759 3760

	/*
	 * Build basic GSI lookup table to facilitate gsi->io_apic lookups
	 * and to prevent reprogramming of IOAPIC pins (PCI GSIs).
	 */
3761
	entries = io_apic_get_redir_entries(idx);
3762 3763 3764
	gsi_cfg = mp_ioapic_gsi_routing(idx);
	gsi_cfg->gsi_base = gsi_base;
	gsi_cfg->gsi_end = gsi_base + entries - 1;
3765 3766 3767 3768

	/*
	 * The number of IO-APIC IRQ registers (== #pins):
	 */
S
Suresh Siddha 已提交
3769
	ioapics[idx].nr_registers = entries;
3770

3771 3772
	if (gsi_cfg->gsi_end >= gsi_top)
		gsi_top = gsi_cfg->gsi_end + 1;
3773

3774 3775 3776 3777
	pr_info("IOAPIC[%d]: apic_id %d, version %d, address 0x%x, GSI %d-%d\n",
		idx, mpc_ioapic_id(idx),
		mpc_ioapic_ver(idx), mpc_ioapic_addr(idx),
		gsi_cfg->gsi_base, gsi_cfg->gsi_end);
3778 3779 3780

	nr_ioapics++;
}
3781 3782 3783 3784

/* Enable IOAPIC early just for system timer */
void __init pre_init_apic_IRQ0(void)
{
3785
	struct io_apic_irq_attr attr = { 0, 0, 0, 0 };
3786 3787 3788

	printk(KERN_INFO "Early APIC setup for system timer0\n");
#ifndef CONFIG_SMP
3789 3790
	physid_set_mask_of_physid(boot_cpu_physical_apicid,
					 &phys_cpu_present_map);
3791 3792 3793
#endif
	setup_local_APIC();

3794
	io_apic_setup_irq_pin(0, 0, &attr);
3795 3796
	irq_set_chip_and_handler_name(0, &ioapic_chip, handle_edge_irq,
				      "edge");
3797
}