events_base.c 37.3 KB
Newer Older
J
Jeremy Fitzhardinge 已提交
1 2 3 4 5 6 7
/*
 * Xen event channels
 *
 * Xen models interrupts with abstract event channels.  Because each
 * domain gets 1024 event channels, but NR_IRQ is not that large, we
 * must dynamically map irqs<->event channels.  The event channels
 * interface with the rest of the kernel by defining a xen interrupt
L
Lucas De Marchi 已提交
8
 * chip.  When an event is received, it is mapped to an irq and sent
J
Jeremy Fitzhardinge 已提交
9 10 11 12 13 14 15 16 17 18
 * through the normal interrupt processing path.
 *
 * There are four kinds of events which can be mapped to an event
 * channel:
 *
 * 1. Inter-domain notifications.  This includes all the virtual
 *    device events, since they're driven by front-ends in another domain
 *    (typically dom0).
 * 2. VIRQs, typically used for timers.  These are per-cpu events.
 * 3. IPIs.
19
 * 4. PIRQs - Hardware interrupts.
J
Jeremy Fitzhardinge 已提交
20 21 22 23
 *
 * Jeremy Fitzhardinge <jeremy@xensource.com>, XenSource Inc, 2007
 */

J
Joe Perches 已提交
24 25
#define pr_fmt(fmt) "xen:" KBUILD_MODNAME ": " fmt

J
Jeremy Fitzhardinge 已提交
26 27 28 29 30
#include <linux/linkage.h>
#include <linux/interrupt.h>
#include <linux/irq.h>
#include <linux/module.h>
#include <linux/string.h>
31
#include <linux/bootmem.h>
32
#include <linux/slab.h>
33
#include <linux/irqnr.h>
34
#include <linux/pci.h>
J
Jeremy Fitzhardinge 已提交
35

36
#ifdef CONFIG_X86
37
#include <asm/desc.h>
J
Jeremy Fitzhardinge 已提交
38 39
#include <asm/ptrace.h>
#include <asm/irq.h>
40
#include <asm/idle.h>
41
#include <asm/io_apic.h>
S
Stefano Stabellini 已提交
42
#include <asm/xen/page.h>
43
#include <asm/xen/pci.h>
44 45
#endif
#include <asm/sync_bitops.h>
J
Jeremy Fitzhardinge 已提交
46
#include <asm/xen/hypercall.h>
47
#include <asm/xen/hypervisor.h>
J
Jeremy Fitzhardinge 已提交
48

49 50
#include <xen/xen.h>
#include <xen/hvm.h>
51
#include <xen/xen-ops.h>
J
Jeremy Fitzhardinge 已提交
52 53 54
#include <xen/events.h>
#include <xen/interface/xen.h>
#include <xen/interface/event_channel.h>
55 56
#include <xen/interface/hvm/hvm_op.h>
#include <xen/interface/hvm/params.h>
57 58
#include <xen/interface/physdev.h>
#include <xen/interface/sched.h>
59
#include <xen/interface/vcpu.h>
60
#include <asm/hw_irq.h>
J
Jeremy Fitzhardinge 已提交
61

62 63
#include "events_internal.h"

64 65
const struct evtchn_ops *evtchn_ops;

J
Jeremy Fitzhardinge 已提交
66 67 68 69
/*
 * This lock protects updates to the following mapping and reference-count
 * arrays. The lock does not need to be acquired to read the mapping tables.
 */
70
static DEFINE_MUTEX(irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
71

72 73
static LIST_HEAD(xen_irq_list_head);

J
Jeremy Fitzhardinge 已提交
74
/* IRQ <-> VIRQ mapping. */
75
static DEFINE_PER_CPU(int [NR_VIRQS], virq_to_irq) = {[0 ... NR_VIRQS-1] = -1};
J
Jeremy Fitzhardinge 已提交
76

J
Jeremy Fitzhardinge 已提交
77
/* IRQ <-> IPI mapping */
78
static DEFINE_PER_CPU(int [XEN_NR_IPIS], ipi_to_irq) = {[0 ... XEN_NR_IPIS-1] = -1};
J
Jeremy Fitzhardinge 已提交
79

80
int **evtchn_to_irq;
81
#ifdef CONFIG_X86
S
Stefano Stabellini 已提交
82
static unsigned long *pirq_eoi_map;
83
#endif
S
Stefano Stabellini 已提交
84
static bool (*pirq_needs_eoi)(unsigned irq);
85

86 87 88 89
#define EVTCHN_ROW(e)  (e / (PAGE_SIZE/sizeof(**evtchn_to_irq)))
#define EVTCHN_COL(e)  (e % (PAGE_SIZE/sizeof(**evtchn_to_irq)))
#define EVTCHN_PER_ROW (PAGE_SIZE / sizeof(**evtchn_to_irq))

J
Jeremy Fitzhardinge 已提交
90 91 92 93
/* Xen will never allocate port zero for any purpose. */
#define VALID_EVTCHN(chn)	((chn) != 0)

static struct irq_chip xen_dynamic_chip;
94
static struct irq_chip xen_percpu_chip;
95
static struct irq_chip xen_pirq_chip;
96 97
static void enable_dynirq(struct irq_data *data);
static void disable_dynirq(struct irq_data *data);
J
Jeremy Fitzhardinge 已提交
98

99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153
static void clear_evtchn_to_irq_row(unsigned row)
{
	unsigned col;

	for (col = 0; col < EVTCHN_PER_ROW; col++)
		evtchn_to_irq[row][col] = -1;
}

static void clear_evtchn_to_irq_all(void)
{
	unsigned row;

	for (row = 0; row < EVTCHN_ROW(xen_evtchn_max_channels()); row++) {
		if (evtchn_to_irq[row] == NULL)
			continue;
		clear_evtchn_to_irq_row(row);
	}
}

static int set_evtchn_to_irq(unsigned evtchn, unsigned irq)
{
	unsigned row;
	unsigned col;

	if (evtchn >= xen_evtchn_max_channels())
		return -EINVAL;

	row = EVTCHN_ROW(evtchn);
	col = EVTCHN_COL(evtchn);

	if (evtchn_to_irq[row] == NULL) {
		/* Unallocated irq entries return -1 anyway */
		if (irq == -1)
			return 0;

		evtchn_to_irq[row] = (int *)get_zeroed_page(GFP_KERNEL);
		if (evtchn_to_irq[row] == NULL)
			return -ENOMEM;

		clear_evtchn_to_irq_row(row);
	}

	evtchn_to_irq[EVTCHN_ROW(evtchn)][EVTCHN_COL(evtchn)] = irq;
	return 0;
}

int get_evtchn_to_irq(unsigned evtchn)
{
	if (evtchn >= xen_evtchn_max_channels())
		return -1;
	if (evtchn_to_irq[EVTCHN_ROW(evtchn)] == NULL)
		return -1;
	return evtchn_to_irq[EVTCHN_ROW(evtchn)][EVTCHN_COL(evtchn)];
}

154
/* Get info for IRQ */
155
struct irq_info *info_for_irq(unsigned irq)
156
{
T
Thomas Gleixner 已提交
157
	return irq_get_handler_data(irq);
158 159
}

160
/* Constructors for packed IRQ information. */
161
static int xen_irq_info_common_setup(struct irq_info *info,
162
				     unsigned irq,
163
				     enum xen_irq_type type,
164
				     unsigned evtchn,
165
				     unsigned short cpu)
166
{
167
	int ret;
168 169 170 171

	BUG_ON(info->type != IRQT_UNBOUND && info->type != type);

	info->type = type;
172
	info->irq = irq;
173 174
	info->evtchn = evtchn;
	info->cpu = cpu;
175

176 177 178
	ret = set_evtchn_to_irq(evtchn, irq);
	if (ret < 0)
		return ret;
179 180

	irq_clear_status_flags(irq, IRQ_NOREQUEST|IRQ_NOAUTOEN);
181

182
	return xen_evtchn_port_setup(info);
183 184
}

185
static int xen_irq_info_evtchn_setup(unsigned irq,
186
				     unsigned evtchn)
187
{
188 189
	struct irq_info *info = info_for_irq(irq);

190
	return xen_irq_info_common_setup(info, irq, IRQT_EVTCHN, evtchn, 0);
191 192
}

193
static int xen_irq_info_ipi_setup(unsigned cpu,
194
				  unsigned irq,
195
				  unsigned evtchn,
196
				  enum ipi_vector ipi)
J
Jeremy Fitzhardinge 已提交
197
{
198 199 200
	struct irq_info *info = info_for_irq(irq);

	info->u.ipi = ipi;
201 202

	per_cpu(ipi_to_irq, cpu)[ipi] = irq;
203 204

	return xen_irq_info_common_setup(info, irq, IRQT_IPI, evtchn, 0);
205 206
}

207
static int xen_irq_info_virq_setup(unsigned cpu,
208
				   unsigned irq,
209 210
				   unsigned evtchn,
				   unsigned virq)
211
{
212 213 214
	struct irq_info *info = info_for_irq(irq);

	info->u.virq = virq;
215 216

	per_cpu(virq_to_irq, cpu)[virq] = irq;
217 218

	return xen_irq_info_common_setup(info, irq, IRQT_VIRQ, evtchn, 0);
219 220
}

221
static int xen_irq_info_pirq_setup(unsigned irq,
222 223 224
				   unsigned evtchn,
				   unsigned pirq,
				   unsigned gsi,
225
				   uint16_t domid,
226
				   unsigned char flags)
227
{
228 229 230 231
	struct irq_info *info = info_for_irq(irq);

	info->u.pirq.pirq = pirq;
	info->u.pirq.gsi = gsi;
232
	info->u.pirq.domid = domid;
233
	info->u.pirq.flags = flags;
234 235

	return xen_irq_info_common_setup(info, irq, IRQT_PIRQ, evtchn, 0);
J
Jeremy Fitzhardinge 已提交
236 237
}

238 239 240 241 242 243
static void xen_irq_info_cleanup(struct irq_info *info)
{
	set_evtchn_to_irq(info->evtchn, -1);
	info->evtchn = 0;
}

J
Jeremy Fitzhardinge 已提交
244 245 246
/*
 * Accessors for packed IRQ information.
 */
247
unsigned int evtchn_from_irq(unsigned irq)
J
Jeremy Fitzhardinge 已提交
248
{
249 250 251
	if (unlikely(WARN(irq < 0 || irq >= nr_irqs, "Invalid irq %d!\n", irq)))
		return 0;

252
	return info_for_irq(irq)->evtchn;
J
Jeremy Fitzhardinge 已提交
253 254
}

I
Ian Campbell 已提交
255 256
unsigned irq_from_evtchn(unsigned int evtchn)
{
257
	return get_evtchn_to_irq(evtchn);
I
Ian Campbell 已提交
258 259 260
}
EXPORT_SYMBOL_GPL(irq_from_evtchn);

261 262 263 264 265
int irq_from_virq(unsigned int cpu, unsigned int virq)
{
	return per_cpu(virq_to_irq, cpu)[virq];
}

266
static enum ipi_vector ipi_from_irq(unsigned irq)
J
Jeremy Fitzhardinge 已提交
267
{
268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285
	struct irq_info *info = info_for_irq(irq);

	BUG_ON(info == NULL);
	BUG_ON(info->type != IRQT_IPI);

	return info->u.ipi;
}

static unsigned virq_from_irq(unsigned irq)
{
	struct irq_info *info = info_for_irq(irq);

	BUG_ON(info == NULL);
	BUG_ON(info->type != IRQT_VIRQ);

	return info->u.virq;
}

S
Stefano Stabellini 已提交
286 287 288 289 290 291 292 293 294 295
static unsigned pirq_from_irq(unsigned irq)
{
	struct irq_info *info = info_for_irq(irq);

	BUG_ON(info == NULL);
	BUG_ON(info->type != IRQT_PIRQ);

	return info->u.pirq.pirq;
}

296 297 298 299 300
static enum xen_irq_type type_from_irq(unsigned irq)
{
	return info_for_irq(irq)->type;
}

301
unsigned cpu_from_irq(unsigned irq)
302 303 304 305
{
	return info_for_irq(irq)->cpu;
}

306
unsigned int cpu_from_evtchn(unsigned int evtchn)
307
{
308
	int irq = get_evtchn_to_irq(evtchn);
309 310 311 312 313 314
	unsigned ret = 0;

	if (irq != -1)
		ret = cpu_from_irq(irq);

	return ret;
J
Jeremy Fitzhardinge 已提交
315 316
}

317
#ifdef CONFIG_X86
S
Stefano Stabellini 已提交
318
static bool pirq_check_eoi_map(unsigned irq)
319
{
320
	return test_bit(pirq_from_irq(irq), pirq_eoi_map);
S
Stefano Stabellini 已提交
321
}
322
#endif
323

S
Stefano Stabellini 已提交
324 325 326
static bool pirq_needs_eoi_flag(unsigned irq)
{
	struct irq_info *info = info_for_irq(irq);
327 328 329 330 331
	BUG_ON(info->type != IRQT_PIRQ);

	return info->u.pirq.flags & PIRQ_NEEDS_EOI;
}

J
Jeremy Fitzhardinge 已提交
332 333
static void bind_evtchn_to_cpu(unsigned int chn, unsigned int cpu)
{
334
	int irq = get_evtchn_to_irq(chn);
335
	struct irq_info *info = info_for_irq(irq);
J
Jeremy Fitzhardinge 已提交
336 337 338

	BUG_ON(irq == -1);
#ifdef CONFIG_SMP
339
	cpumask_copy(irq_get_irq_data(irq)->affinity, cpumask_of(cpu));
J
Jeremy Fitzhardinge 已提交
340
#endif
341
	xen_evtchn_port_bind_to_cpu(info, cpu);
342

343
	info->cpu = cpu;
344 345
}

346 347 348 349 350 351 352 353
static void xen_evtchn_mask_all(void)
{
	unsigned int evtchn;

	for (evtchn = 0; evtchn < xen_evtchn_nr_channels(); evtchn++)
		mask_evtchn(evtchn);
}

J
Jeremy Fitzhardinge 已提交
354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370
/**
 * notify_remote_via_irq - send event to remote end of event channel via irq
 * @irq: irq of event channel to send event to
 *
 * Unlike notify_remote_via_evtchn(), this is safe to use across
 * save/restore. Notifications on a broken connection are silently
 * dropped.
 */
void notify_remote_via_irq(int irq)
{
	int evtchn = evtchn_from_irq(irq);

	if (VALID_EVTCHN(evtchn))
		notify_remote_via_evtchn(evtchn);
}
EXPORT_SYMBOL_GPL(notify_remote_via_irq);

371 372 373
static void xen_irq_init(unsigned irq)
{
	struct irq_info *info;
374
#ifdef CONFIG_SMP
375
	/* By default all event channels notify CPU#0. */
376
	cpumask_copy(irq_get_irq_data(irq)->affinity, cpumask_of(0));
377
#endif
378

379 380 381
	info = kzalloc(sizeof(*info), GFP_KERNEL);
	if (info == NULL)
		panic("Unable to allocate metadata for IRQ%d\n", irq);
382 383

	info->type = IRQT_UNBOUND;
384
	info->refcnt = -1;
385

T
Thomas Gleixner 已提交
386
	irq_set_handler_data(irq, info);
387

388 389 390
	list_add_tail(&info->list, &xen_irq_list_head);
}

391
static int __must_check xen_allocate_irq_dynamic(void)
392
{
393 394
	int first = 0;
	int irq;
395 396

#ifdef CONFIG_X86_IO_APIC
397 398
	/*
	 * For an HVM guest or domain 0 which see "real" (emulated or
L
Lucas De Marchi 已提交
399
	 * actual respectively) GSIs we allocate dynamic IRQs
400 401 402 403 404 405
	 * e.g. those corresponding to event channels or MSIs
	 * etc. from the range above those "real" GSIs to avoid
	 * collisions.
	 */
	if (xen_initial_domain() || xen_hvm_domain())
		first = get_nr_irqs_gsi();
406 407
#endif

408
	irq = irq_alloc_desc_from(first, -1);
409

410 411
	if (irq >= 0)
		xen_irq_init(irq);
412

J
Jeremy Fitzhardinge 已提交
413
	return irq;
414 415
}

416
static int __must_check xen_allocate_irq_gsi(unsigned gsi)
417 418 419
{
	int irq;

420 421 422 423 424 425 426
	/*
	 * A PV guest has no concept of a GSI (since it has no ACPI
	 * nor access to/knowledge of the physical APICs). Therefore
	 * all IRQs are dynamically allocated from the entire IRQ
	 * space.
	 */
	if (xen_pv_domain() && !xen_initial_domain())
427 428 429 430
		return xen_allocate_irq_dynamic();

	/* Legacy IRQ descriptors are already allocated by the arch. */
	if (gsi < NR_IRQS_LEGACY)
431 432 433
		irq = gsi;
	else
		irq = irq_alloc_desc_at(gsi, -1);
434

435
	xen_irq_init(irq);
436 437 438 439 440 441

	return irq;
}

static void xen_free_irq(unsigned irq)
{
T
Thomas Gleixner 已提交
442
	struct irq_info *info = irq_get_handler_data(irq);
443

444 445 446
	if (WARN_ON(!info))
		return;

447
	list_del(&info->list);
448

T
Thomas Gleixner 已提交
449
	irq_set_handler_data(irq, NULL);
450

451 452
	WARN_ON(info->refcnt > 0);

453 454
	kfree(info);

455 456 457 458
	/* Legacy IRQ descriptors are managed by the arch. */
	if (irq < NR_IRQS_LEGACY)
		return;

459 460 461
	irq_free_desc(irq);
}

462 463 464 465 466 467 468 469 470 471 472 473
static void xen_evtchn_close(unsigned int port)
{
	struct evtchn_close close;

	close.port = port;
	if (HYPERVISOR_event_channel_op(EVTCHNOP_close, &close) != 0)
		BUG();

	/* Closed ports are implicitly re-bound to VCPU0. */
	bind_evtchn_to_cpu(port, 0);
}

474 475 476 477 478 479 480
static void pirq_query_unmask(int irq)
{
	struct physdev_irq_status_query irq_status;
	struct irq_info *info = info_for_irq(irq);

	BUG_ON(info->type != IRQT_PIRQ);

S
Stefano Stabellini 已提交
481
	irq_status.irq = pirq_from_irq(irq);
482 483 484 485 486 487 488 489
	if (HYPERVISOR_physdev_op(PHYSDEVOP_irq_status_query, &irq_status))
		irq_status.flags = 0;

	info->u.pirq.flags &= ~PIRQ_NEEDS_EOI;
	if (irq_status.flags & XENIRQSTAT_needs_eoi)
		info->u.pirq.flags |= PIRQ_NEEDS_EOI;
}

490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512
static void eoi_pirq(struct irq_data *data)
{
	int evtchn = evtchn_from_irq(data->irq);
	struct physdev_eoi eoi = { .irq = pirq_from_irq(data->irq) };
	int rc = 0;

	irq_move_irq(data);

	if (VALID_EVTCHN(evtchn))
		clear_evtchn(evtchn);

	if (pirq_needs_eoi(data->irq)) {
		rc = HYPERVISOR_physdev_op(PHYSDEVOP_eoi, &eoi);
		WARN_ON(rc);
	}
}

static void mask_ack_pirq(struct irq_data *data)
{
	disable_dynirq(data);
	eoi_pirq(data);
}

513
static unsigned int __startup_pirq(unsigned int irq)
514 515 516 517
{
	struct evtchn_bind_pirq bind_pirq;
	struct irq_info *info = info_for_irq(irq);
	int evtchn = evtchn_from_irq(irq);
518
	int rc;
519 520 521 522 523 524

	BUG_ON(info->type != IRQT_PIRQ);

	if (VALID_EVTCHN(evtchn))
		goto out;

S
Stefano Stabellini 已提交
525
	bind_pirq.pirq = pirq_from_irq(irq);
526
	/* NB. We are happy to share unless we are probing. */
527 528 529 530
	bind_pirq.flags = info->u.pirq.flags & PIRQ_SHAREABLE ?
					BIND_PIRQ__WILL_SHARE : 0;
	rc = HYPERVISOR_event_channel_op(EVTCHNOP_bind_pirq, &bind_pirq);
	if (rc != 0) {
531
		pr_warn("Failed to obtain physical IRQ %d\n", irq);
532 533 534 535 536 537
		return 0;
	}
	evtchn = bind_pirq.port;

	pirq_query_unmask(irq);

538 539 540 541 542 543 544
	rc = set_evtchn_to_irq(evtchn, irq);
	if (rc != 0) {
		pr_err("irq%d: Failed to set port to irq mapping (%d)\n",
		       irq, rc);
		xen_evtchn_close(evtchn);
		return 0;
	}
545 546 547 548 549
	bind_evtchn_to_cpu(evtchn, 0);
	info->evtchn = evtchn;

out:
	unmask_evtchn(evtchn);
550
	eoi_pirq(irq_get_irq_data(irq));
551 552 553 554

	return 0;
}

555 556 557 558 559 560
static unsigned int startup_pirq(struct irq_data *data)
{
	return __startup_pirq(data->irq);
}

static void shutdown_pirq(struct irq_data *data)
561
{
562
	unsigned int irq = data->irq;
563
	struct irq_info *info = info_for_irq(irq);
564
	unsigned evtchn = evtchn_from_irq(irq);
565 566 567 568 569 570 571

	BUG_ON(info->type != IRQT_PIRQ);

	if (!VALID_EVTCHN(evtchn))
		return;

	mask_evtchn(evtchn);
572 573
	xen_evtchn_close(evtchn);
	xen_irq_info_cleanup(info);
574 575
}

576
static void enable_pirq(struct irq_data *data)
577
{
578
	startup_pirq(data);
579 580
}

581
static void disable_pirq(struct irq_data *data)
582
{
583
	disable_dynirq(data);
584 585
}

586
int xen_irq_from_gsi(unsigned gsi)
587
{
588
	struct irq_info *info;
589

590 591
	list_for_each_entry(info, &xen_irq_list_head, list) {
		if (info->type != IRQT_PIRQ)
592 593
			continue;

594 595
		if (info->u.pirq.gsi == gsi)
			return info->irq;
596 597 598 599
	}

	return -1;
}
600
EXPORT_SYMBOL_GPL(xen_irq_from_gsi);
601

602 603 604 605 606 607 608 609 610 611 612 613
static void __unbind_from_irq(unsigned int irq)
{
	int evtchn = evtchn_from_irq(irq);
	struct irq_info *info = irq_get_handler_data(irq);

	if (info->refcnt > 0) {
		info->refcnt--;
		if (info->refcnt != 0)
			return;
	}

	if (VALID_EVTCHN(evtchn)) {
614 615 616
		unsigned int cpu = cpu_from_irq(irq);

		xen_evtchn_close(evtchn);
617 618 619

		switch (type_from_irq(irq)) {
		case IRQT_VIRQ:
620
			per_cpu(virq_to_irq, cpu)[virq_from_irq(irq)] = -1;
621 622
			break;
		case IRQT_IPI:
623
			per_cpu(ipi_to_irq, cpu)[ipi_from_irq(irq)] = -1;
624 625 626 627 628
			break;
		default:
			break;
		}

629
		xen_irq_info_cleanup(info);
630 631 632 633 634 635 636
	}

	BUG_ON(info_for_irq(irq)->type == IRQT_UNBOUND);

	xen_free_irq(irq);
}

I
Ian Campbell 已提交
637 638 639
/*
 * Do not make any assumptions regarding the relationship between the
 * IRQ number returned here and the Xen pirq argument.
S
Stefano Stabellini 已提交
640 641 642
 *
 * Note: We don't assign an event channel until the irq actually started
 * up.  Return an existing irq if we've already got one for the gsi.
643 644 645
 *
 * Shareable implies level triggered, not shareable implies edge
 * triggered here.
646
 */
647 648
int xen_bind_pirq_gsi_to_irq(unsigned gsi,
			     unsigned pirq, int shareable, char *name)
649
{
650
	int irq = -1;
651
	struct physdev_irq irq_op;
652
	int ret;
653

654
	mutex_lock(&irq_mapping_update_lock);
655

656
	irq = xen_irq_from_gsi(gsi);
657
	if (irq != -1) {
J
Joe Perches 已提交
658 659
		pr_info("%s: returning irq %d for gsi %u\n",
			__func__, irq, gsi);
660
		goto out;
661 662
	}

663
	irq = xen_allocate_irq_gsi(gsi);
664 665
	if (irq < 0)
		goto out;
666 667

	irq_op.irq = irq;
668 669 670 671 672 673 674
	irq_op.vector = 0;

	/* Only the privileged domain can do this. For non-priv, the pcifront
	 * driver provides a PCI bus that does the call to do exactly
	 * this in the priv domain. */
	if (xen_initial_domain() &&
	    HYPERVISOR_physdev_op(PHYSDEVOP_alloc_irq_vector, &irq_op)) {
675
		xen_free_irq(irq);
676 677 678 679
		irq = -ENOSPC;
		goto out;
	}

680
	ret = xen_irq_info_pirq_setup(irq, 0, pirq, gsi, DOMID_SELF,
681
			       shareable ? PIRQ_SHAREABLE : 0);
682 683 684 685 686
	if (ret < 0) {
		__unbind_from_irq(irq);
		irq = ret;
		goto out;
	}
687

688 689
	pirq_query_unmask(irq);
	/* We try to use the handler with the appropriate semantic for the
690 691
	 * type of interrupt: if the interrupt is an edge triggered
	 * interrupt we use handle_edge_irq.
692
	 *
693 694
	 * On the other hand if the interrupt is level triggered we use
	 * handle_fasteoi_irq like the native code does for this kind of
695
	 * interrupts.
696
	 *
697 698 699 700 701 702 703
	 * Depending on the Xen version, pirq_needs_eoi might return true
	 * not only for level triggered interrupts but for edge triggered
	 * interrupts too. In any case Xen always honors the eoi mechanism,
	 * not injecting any more pirqs of the same kind if the first one
	 * hasn't received an eoi yet. Therefore using the fasteoi handler
	 * is the right choice either way.
	 */
704
	if (shareable)
705 706 707 708 709 710
		irq_set_chip_and_handler_name(irq, &xen_pirq_chip,
				handle_fasteoi_irq, name);
	else
		irq_set_chip_and_handler_name(irq, &xen_pirq_chip,
				handle_edge_irq, name);

711
out:
712
	mutex_unlock(&irq_mapping_update_lock);
713 714 715 716

	return irq;
}

717
#ifdef CONFIG_PCI_MSI
718
int xen_allocate_pirq_msi(struct pci_dev *dev, struct msi_desc *msidesc)
719
{
720
	int rc;
721 722
	struct physdev_get_free_pirq op_get_free_pirq;

723
	op_get_free_pirq.type = MAP_PIRQ_TYPE_MSI;
724 725
	rc = HYPERVISOR_physdev_op(PHYSDEVOP_get_free_pirq, &op_get_free_pirq);

726 727 728 729
	WARN_ONCE(rc == -ENOSYS,
		  "hypervisor does not support the PHYSDEVOP_get_free_pirq interface\n");

	return rc ? -1 : op_get_free_pirq.pirq;
730 731
}

732
int xen_bind_pirq_msi_to_irq(struct pci_dev *dev, struct msi_desc *msidesc,
J
Jan Beulich 已提交
733
			     int pirq, const char *name, domid_t domid)
S
Stefano Stabellini 已提交
734
{
735
	int irq, ret;
736

737
	mutex_lock(&irq_mapping_update_lock);
S
Stefano Stabellini 已提交
738

739
	irq = xen_allocate_irq_dynamic();
740
	if (irq < 0)
741
		goto out;
S
Stefano Stabellini 已提交
742

743 744
	irq_set_chip_and_handler_name(irq, &xen_pirq_chip, handle_edge_irq,
			name);
S
Stefano Stabellini 已提交
745

746 747 748
	ret = xen_irq_info_pirq_setup(irq, 0, pirq, 0, domid, 0);
	if (ret < 0)
		goto error_irq;
749
	ret = irq_set_msi_desc(irq, msidesc);
750 751
	if (ret < 0)
		goto error_irq;
S
Stefano Stabellini 已提交
752
out:
753
	mutex_unlock(&irq_mapping_update_lock);
754
	return irq;
755
error_irq:
756
	__unbind_from_irq(irq);
757
	mutex_unlock(&irq_mapping_update_lock);
758
	return ret;
S
Stefano Stabellini 已提交
759
}
760 761
#endif

762 763
int xen_destroy_irq(int irq)
{
764 765
	struct physdev_unmap_pirq unmap_irq;
	struct irq_info *info = info_for_irq(irq);
766 767
	int rc = -ENOENT;

768
	mutex_lock(&irq_mapping_update_lock);
769

770
	if (xen_initial_domain()) {
771
		unmap_irq.pirq = info->u.pirq.pirq;
772
		unmap_irq.domid = info->u.pirq.domid;
773
		rc = HYPERVISOR_physdev_op(PHYSDEVOP_unmap_pirq, &unmap_irq);
774 775 776 777 778
		/* If another domain quits without making the pci_disable_msix
		 * call, the Xen hypervisor takes care of freeing the PIRQs
		 * (free_domain_pirqs).
		 */
		if ((rc == -ESRCH && info->u.pirq.domid != DOMID_SELF))
J
Joe Perches 已提交
779
			pr_info("domain %d does not have %d anymore\n",
780 781
				info->u.pirq.domid, info->u.pirq.pirq);
		else if (rc) {
J
Joe Perches 已提交
782
			pr_warn("unmap irq failed %d\n", rc);
783 784 785
			goto out;
		}
	}
786

787
	xen_free_irq(irq);
788 789

out:
790
	mutex_unlock(&irq_mapping_update_lock);
791 792 793
	return rc;
}

794
int xen_irq_from_pirq(unsigned pirq)
795
{
796
	int irq;
797

798
	struct irq_info *info;
J
Jeremy Fitzhardinge 已提交
799

800
	mutex_lock(&irq_mapping_update_lock);
801 802

	list_for_each_entry(info, &xen_irq_list_head, list) {
803
		if (info->type != IRQT_PIRQ)
804 805 806 807 808 809 810
			continue;
		irq = info->irq;
		if (info->u.pirq.pirq == pirq)
			goto out;
	}
	irq = -1;
out:
811
	mutex_unlock(&irq_mapping_update_lock);
812 813

	return irq;
814 815
}

816 817 818 819 820 821

int xen_pirq_from_irq(unsigned irq)
{
	return pirq_from_irq(irq);
}
EXPORT_SYMBOL_GPL(xen_pirq_from_irq);
822

823
int bind_evtchn_to_irq(unsigned int evtchn)
J
Jeremy Fitzhardinge 已提交
824 825
{
	int irq;
826
	int ret;
J
Jeremy Fitzhardinge 已提交
827

828 829 830
	if (evtchn >= xen_evtchn_max_channels())
		return -ENOMEM;

831
	mutex_lock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
832

833
	irq = get_evtchn_to_irq(evtchn);
J
Jeremy Fitzhardinge 已提交
834 835

	if (irq == -1) {
836
		irq = xen_allocate_irq_dynamic();
837
		if (irq < 0)
838
			goto out;
J
Jeremy Fitzhardinge 已提交
839

T
Thomas Gleixner 已提交
840
		irq_set_chip_and_handler_name(irq, &xen_dynamic_chip,
841
					      handle_edge_irq, "event");
J
Jeremy Fitzhardinge 已提交
842

843 844 845 846 847 848
		ret = xen_irq_info_evtchn_setup(irq, evtchn);
		if (ret < 0) {
			__unbind_from_irq(irq);
			irq = ret;
			goto out;
		}
849 850
		/* New interdomain events are bound to VCPU 0. */
		bind_evtchn_to_cpu(evtchn, 0);
851 852 853
	} else {
		struct irq_info *info = info_for_irq(irq);
		WARN_ON(info == NULL || info->type != IRQT_EVTCHN);
J
Jeremy Fitzhardinge 已提交
854 855
	}

856
out:
857
	mutex_unlock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
858 859 860

	return irq;
}
861
EXPORT_SYMBOL_GPL(bind_evtchn_to_irq);
J
Jeremy Fitzhardinge 已提交
862

J
Jeremy Fitzhardinge 已提交
863 864 865 866
static int bind_ipi_to_irq(unsigned int ipi, unsigned int cpu)
{
	struct evtchn_bind_ipi bind_ipi;
	int evtchn, irq;
867
	int ret;
J
Jeremy Fitzhardinge 已提交
868

869
	mutex_lock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
870 871

	irq = per_cpu(ipi_to_irq, cpu)[ipi];
872

J
Jeremy Fitzhardinge 已提交
873
	if (irq == -1) {
874
		irq = xen_allocate_irq_dynamic();
J
Jeremy Fitzhardinge 已提交
875 876 877
		if (irq < 0)
			goto out;

T
Thomas Gleixner 已提交
878
		irq_set_chip_and_handler_name(irq, &xen_percpu_chip,
879
					      handle_percpu_irq, "ipi");
J
Jeremy Fitzhardinge 已提交
880 881 882 883 884 885 886

		bind_ipi.vcpu = cpu;
		if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_ipi,
						&bind_ipi) != 0)
			BUG();
		evtchn = bind_ipi.port;

887 888 889 890 891 892
		ret = xen_irq_info_ipi_setup(cpu, irq, evtchn, ipi);
		if (ret < 0) {
			__unbind_from_irq(irq);
			irq = ret;
			goto out;
		}
J
Jeremy Fitzhardinge 已提交
893
		bind_evtchn_to_cpu(evtchn, cpu);
894 895 896
	} else {
		struct irq_info *info = info_for_irq(irq);
		WARN_ON(info == NULL || info->type != IRQT_IPI);
J
Jeremy Fitzhardinge 已提交
897 898 899
	}

 out:
900
	mutex_unlock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
901 902 903
	return irq;
}

904 905 906 907 908 909 910 911 912 913 914 915 916 917 918
static int bind_interdomain_evtchn_to_irq(unsigned int remote_domain,
					  unsigned int remote_port)
{
	struct evtchn_bind_interdomain bind_interdomain;
	int err;

	bind_interdomain.remote_dom  = remote_domain;
	bind_interdomain.remote_port = remote_port;

	err = HYPERVISOR_event_channel_op(EVTCHNOP_bind_interdomain,
					  &bind_interdomain);

	return err ? : bind_evtchn_to_irq(bind_interdomain.local_port);
}

919 920 921 922 923 924
static int find_virq(unsigned int virq, unsigned int cpu)
{
	struct evtchn_status status;
	int port, rc = -ENOENT;

	memset(&status, 0, sizeof(status));
925
	for (port = 0; port < xen_evtchn_max_channels(); port++) {
926 927 928 929 930 931 932 933 934 935 936 937 938 939
		status.dom = DOMID_SELF;
		status.port = port;
		rc = HYPERVISOR_event_channel_op(EVTCHNOP_status, &status);
		if (rc < 0)
			continue;
		if (status.status != EVTCHNSTAT_virq)
			continue;
		if (status.u.virq == virq && status.vcpu == cpu) {
			rc = port;
			break;
		}
	}
	return rc;
}
J
Jeremy Fitzhardinge 已提交
940

941 942 943 944 945 946 947 948 949 950 951 952 953
/**
 * xen_evtchn_nr_channels - number of usable event channel ports
 *
 * This may be less than the maximum supported by the current
 * hypervisor ABI. Use xen_evtchn_max_channels() for the maximum
 * supported.
 */
unsigned xen_evtchn_nr_channels(void)
{
        return evtchn_ops->nr_channels();
}
EXPORT_SYMBOL_GPL(xen_evtchn_nr_channels);

954
int bind_virq_to_irq(unsigned int virq, unsigned int cpu)
J
Jeremy Fitzhardinge 已提交
955 956
{
	struct evtchn_bind_virq bind_virq;
957
	int evtchn, irq, ret;
J
Jeremy Fitzhardinge 已提交
958

959
	mutex_lock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
960 961 962 963

	irq = per_cpu(virq_to_irq, cpu)[virq];

	if (irq == -1) {
964
		irq = xen_allocate_irq_dynamic();
965
		if (irq < 0)
966
			goto out;
967

T
Thomas Gleixner 已提交
968
		irq_set_chip_and_handler_name(irq, &xen_percpu_chip,
969 970
					      handle_percpu_irq, "virq");

J
Jeremy Fitzhardinge 已提交
971 972
		bind_virq.virq = virq;
		bind_virq.vcpu = cpu;
973 974 975 976 977 978 979 980 981 982
		ret = HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq,
						&bind_virq);
		if (ret == 0)
			evtchn = bind_virq.port;
		else {
			if (ret == -EEXIST)
				ret = find_virq(virq, cpu);
			BUG_ON(ret < 0);
			evtchn = ret;
		}
J
Jeremy Fitzhardinge 已提交
983

984 985 986 987 988 989
		ret = xen_irq_info_virq_setup(cpu, irq, evtchn, virq);
		if (ret < 0) {
			__unbind_from_irq(irq);
			irq = ret;
			goto out;
		}
J
Jeremy Fitzhardinge 已提交
990 991

		bind_evtchn_to_cpu(evtchn, cpu);
992 993 994
	} else {
		struct irq_info *info = info_for_irq(irq);
		WARN_ON(info == NULL || info->type != IRQT_VIRQ);
J
Jeremy Fitzhardinge 已提交
995 996
	}

997
out:
998
	mutex_unlock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
999 1000 1001 1002 1003 1004

	return irq;
}

static void unbind_from_irq(unsigned int irq)
{
1005
	mutex_lock(&irq_mapping_update_lock);
1006
	__unbind_from_irq(irq);
1007
	mutex_unlock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
1008 1009 1010
}

int bind_evtchn_to_irqhandler(unsigned int evtchn,
1011
			      irq_handler_t handler,
J
Jeremy Fitzhardinge 已提交
1012 1013 1014
			      unsigned long irqflags,
			      const char *devname, void *dev_id)
{
1015
	int irq, retval;
J
Jeremy Fitzhardinge 已提交
1016 1017

	irq = bind_evtchn_to_irq(evtchn);
1018 1019
	if (irq < 0)
		return irq;
J
Jeremy Fitzhardinge 已提交
1020 1021 1022 1023 1024 1025 1026 1027 1028 1029
	retval = request_irq(irq, handler, irqflags, devname, dev_id);
	if (retval != 0) {
		unbind_from_irq(irq);
		return retval;
	}

	return irq;
}
EXPORT_SYMBOL_GPL(bind_evtchn_to_irqhandler);

1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052
int bind_interdomain_evtchn_to_irqhandler(unsigned int remote_domain,
					  unsigned int remote_port,
					  irq_handler_t handler,
					  unsigned long irqflags,
					  const char *devname,
					  void *dev_id)
{
	int irq, retval;

	irq = bind_interdomain_evtchn_to_irq(remote_domain, remote_port);
	if (irq < 0)
		return irq;

	retval = request_irq(irq, handler, irqflags, devname, dev_id);
	if (retval != 0) {
		unbind_from_irq(irq);
		return retval;
	}

	return irq;
}
EXPORT_SYMBOL_GPL(bind_interdomain_evtchn_to_irqhandler);

J
Jeremy Fitzhardinge 已提交
1053
int bind_virq_to_irqhandler(unsigned int virq, unsigned int cpu,
1054
			    irq_handler_t handler,
J
Jeremy Fitzhardinge 已提交
1055 1056
			    unsigned long irqflags, const char *devname, void *dev_id)
{
1057
	int irq, retval;
J
Jeremy Fitzhardinge 已提交
1058 1059

	irq = bind_virq_to_irq(virq, cpu);
1060 1061
	if (irq < 0)
		return irq;
J
Jeremy Fitzhardinge 已提交
1062 1063 1064 1065 1066 1067 1068 1069 1070 1071
	retval = request_irq(irq, handler, irqflags, devname, dev_id);
	if (retval != 0) {
		unbind_from_irq(irq);
		return retval;
	}

	return irq;
}
EXPORT_SYMBOL_GPL(bind_virq_to_irqhandler);

J
Jeremy Fitzhardinge 已提交
1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084
int bind_ipi_to_irqhandler(enum ipi_vector ipi,
			   unsigned int cpu,
			   irq_handler_t handler,
			   unsigned long irqflags,
			   const char *devname,
			   void *dev_id)
{
	int irq, retval;

	irq = bind_ipi_to_irq(ipi, cpu);
	if (irq < 0)
		return irq;

1085
	irqflags |= IRQF_NO_SUSPEND | IRQF_FORCE_RESUME | IRQF_EARLY_RESUME;
J
Jeremy Fitzhardinge 已提交
1086 1087 1088 1089 1090 1091 1092 1093 1094
	retval = request_irq(irq, handler, irqflags, devname, dev_id);
	if (retval != 0) {
		unbind_from_irq(irq);
		return retval;
	}

	return irq;
}

J
Jeremy Fitzhardinge 已提交
1095 1096
void unbind_from_irqhandler(unsigned int irq, void *dev_id)
{
1097 1098 1099 1100
	struct irq_info *info = irq_get_handler_data(irq);

	if (WARN_ON(!info))
		return;
J
Jeremy Fitzhardinge 已提交
1101 1102 1103 1104 1105
	free_irq(irq, dev_id);
	unbind_from_irq(irq);
}
EXPORT_SYMBOL_GPL(unbind_from_irqhandler);

1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122
/**
 * xen_set_irq_priority() - set an event channel priority.
 * @irq:irq bound to an event channel.
 * @priority: priority between XEN_IRQ_PRIORITY_MAX and XEN_IRQ_PRIORITY_MIN.
 */
int xen_set_irq_priority(unsigned irq, unsigned priority)
{
	struct evtchn_set_priority set_priority;

	set_priority.port = evtchn_from_irq(irq);
	set_priority.priority = priority;

	return HYPERVISOR_event_channel_op(EVTCHNOP_set_priority,
					   &set_priority);
}
EXPORT_SYMBOL_GPL(xen_set_irq_priority);

1123 1124
int evtchn_make_refcounted(unsigned int evtchn)
{
1125
	int irq = get_evtchn_to_irq(evtchn);
1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149
	struct irq_info *info;

	if (irq == -1)
		return -ENOENT;

	info = irq_get_handler_data(irq);

	if (!info)
		return -ENOENT;

	WARN_ON(info->refcnt != -1);

	info->refcnt = 1;

	return 0;
}
EXPORT_SYMBOL_GPL(evtchn_make_refcounted);

int evtchn_get(unsigned int evtchn)
{
	int irq;
	struct irq_info *info;
	int err = -ENOENT;

1150
	if (evtchn >= xen_evtchn_max_channels())
1151 1152
		return -EINVAL;

1153 1154
	mutex_lock(&irq_mapping_update_lock);

1155
	irq = get_evtchn_to_irq(evtchn);
1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178
	if (irq == -1)
		goto done;

	info = irq_get_handler_data(irq);

	if (!info)
		goto done;

	err = -EINVAL;
	if (info->refcnt <= 0)
		goto done;

	info->refcnt++;
	err = 0;
 done:
	mutex_unlock(&irq_mapping_update_lock);

	return err;
}
EXPORT_SYMBOL_GPL(evtchn_get);

void evtchn_put(unsigned int evtchn)
{
1179
	int irq = get_evtchn_to_irq(evtchn);
1180 1181 1182 1183 1184 1185
	if (WARN_ON(irq == -1))
		return;
	unbind_from_irq(irq);
}
EXPORT_SYMBOL_GPL(evtchn_put);

J
Jeremy Fitzhardinge 已提交
1186 1187
void xen_send_IPI_one(unsigned int cpu, enum ipi_vector vector)
{
1188 1189
	int irq;

1190
#ifdef CONFIG_X86
1191 1192 1193 1194 1195 1196
	if (unlikely(vector == XEN_NMI_VECTOR)) {
		int rc =  HYPERVISOR_vcpu_op(VCPUOP_send_nmi, cpu, NULL);
		if (rc < 0)
			printk(KERN_WARNING "Sending nmi to CPU%d failed (rc:%d)\n", cpu, rc);
		return;
	}
1197
#endif
1198
	irq = per_cpu(ipi_to_irq, cpu)[vector];
J
Jeremy Fitzhardinge 已提交
1199 1200 1201 1202
	BUG_ON(irq < 0);
	notify_remote_via_irq(irq);
}

1203 1204
static DEFINE_PER_CPU(unsigned, xed_nesting_count);

1205
static void __xen_evtchn_do_upcall(void)
J
Jeremy Fitzhardinge 已提交
1206
{
C
Christoph Lameter 已提交
1207
	struct vcpu_info *vcpu_info = __this_cpu_read(xen_vcpu);
1208
	int cpu = get_cpu();
1209
	unsigned count;
J
Jeremy Fitzhardinge 已提交
1210

1211 1212
	do {
		vcpu_info->evtchn_upcall_pending = 0;
J
Jeremy Fitzhardinge 已提交
1213

C
Christoph Lameter 已提交
1214
		if (__this_cpu_inc_return(xed_nesting_count) - 1)
1215
			goto out;
J
Jeremy Fitzhardinge 已提交
1216

1217
		xen_evtchn_handle_events(cpu);
J
Jeremy Fitzhardinge 已提交
1218

1219 1220
		BUG_ON(!irqs_disabled());

C
Christoph Lameter 已提交
1221 1222
		count = __this_cpu_read(xed_nesting_count);
		__this_cpu_write(xed_nesting_count, 0);
1223
	} while (count != 1 || vcpu_info->evtchn_upcall_pending);
1224 1225

out:
1226 1227 1228 1229 1230 1231 1232 1233

	put_cpu();
}

void xen_evtchn_do_upcall(struct pt_regs *regs)
{
	struct pt_regs *old_regs = set_irq_regs(regs);

1234
	irq_enter();
1235
#ifdef CONFIG_X86
1236
	exit_idle();
1237
#endif
1238
	inc_irq_stat(irq_hv_callback_count);
1239 1240 1241

	__xen_evtchn_do_upcall();

1242 1243
	irq_exit();
	set_irq_regs(old_regs);
1244
}
1245

1246 1247 1248
void xen_hvm_evtchn_do_upcall(void)
{
	__xen_evtchn_do_upcall();
J
Jeremy Fitzhardinge 已提交
1249
}
1250
EXPORT_SYMBOL_GPL(xen_hvm_evtchn_do_upcall);
J
Jeremy Fitzhardinge 已提交
1251

J
Jeremy Fitzhardinge 已提交
1252 1253 1254
/* Rebind a new event channel to an existing irq. */
void rebind_evtchn_irq(int evtchn, int irq)
{
J
Jeremy Fitzhardinge 已提交
1255 1256
	struct irq_info *info = info_for_irq(irq);

1257 1258 1259
	if (WARN_ON(!info))
		return;

J
Jeremy Fitzhardinge 已提交
1260 1261 1262 1263
	/* Make sure the irq is masked, since the new event channel
	   will also be masked. */
	disable_irq(irq);

1264
	mutex_lock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
1265 1266

	/* After resume the irq<->evtchn mappings are all cleared out */
1267
	BUG_ON(get_evtchn_to_irq(evtchn) != -1);
J
Jeremy Fitzhardinge 已提交
1268
	/* Expect irq to have been bound before,
J
Jeremy Fitzhardinge 已提交
1269 1270
	   so there should be a proper type */
	BUG_ON(info->type == IRQT_UNBOUND);
J
Jeremy Fitzhardinge 已提交
1271

1272
	(void)xen_irq_info_evtchn_setup(irq, evtchn);
J
Jeremy Fitzhardinge 已提交
1273

1274
	mutex_unlock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
1275 1276

	/* new event channels are always bound to cpu 0 */
1277
	irq_set_affinity(irq, cpumask_of(0));
J
Jeremy Fitzhardinge 已提交
1278 1279 1280 1281 1282

	/* Unmask the event channel. */
	enable_irq(irq);
}

J
Jeremy Fitzhardinge 已提交
1283
/* Rebind an evtchn so that it gets delivered to a specific cpu */
1284
static int rebind_irq_to_cpu(unsigned irq, unsigned tcpu)
J
Jeremy Fitzhardinge 已提交
1285 1286 1287
{
	struct evtchn_bind_vcpu bind_vcpu;
	int evtchn = evtchn_from_irq(irq);
1288
	int masked;
J
Jeremy Fitzhardinge 已提交
1289

1290 1291 1292 1293 1294 1295 1296 1297
	if (!VALID_EVTCHN(evtchn))
		return -1;

	/*
	 * Events delivered via platform PCI interrupts are always
	 * routed to vcpu 0 and hence cannot be rebound.
	 */
	if (xen_hvm_domain() && !xen_have_vector_callback)
1298
		return -1;
J
Jeremy Fitzhardinge 已提交
1299 1300 1301 1302 1303

	/* Send future instances of this interrupt to other vcpu. */
	bind_vcpu.port = evtchn;
	bind_vcpu.vcpu = tcpu;

1304 1305 1306 1307
	/*
	 * Mask the event while changing the VCPU binding to prevent
	 * it being delivered on an unexpected VCPU.
	 */
1308
	masked = test_and_set_mask(evtchn);
1309

J
Jeremy Fitzhardinge 已提交
1310 1311 1312 1313 1314 1315 1316 1317
	/*
	 * If this fails, it usually just indicates that we're dealing with a
	 * virq or IPI channel, which don't actually need to be rebound. Ignore
	 * it, but don't do the xenlinux-level rebind in that case.
	 */
	if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_vcpu, &bind_vcpu) >= 0)
		bind_evtchn_to_cpu(evtchn, tcpu);

1318 1319 1320
	if (!masked)
		unmask_evtchn(evtchn);

1321 1322
	return 0;
}
J
Jeremy Fitzhardinge 已提交
1323

1324 1325
static int set_affinity_irq(struct irq_data *data, const struct cpumask *dest,
			    bool force)
J
Jeremy Fitzhardinge 已提交
1326
{
1327
	unsigned tcpu = cpumask_first(dest);
1328

1329
	return rebind_irq_to_cpu(data->irq, tcpu);
J
Jeremy Fitzhardinge 已提交
1330 1331
}

1332
static int retrigger_evtchn(int evtchn)
1333
{
1334
	int masked;
1335 1336

	if (!VALID_EVTCHN(evtchn))
1337
		return 0;
1338

1339
	masked = test_and_set_mask(evtchn);
1340
	set_evtchn(evtchn);
1341 1342 1343 1344 1345 1346
	if (!masked)
		unmask_evtchn(evtchn);

	return 1;
}

1347 1348 1349 1350 1351
int resend_irq_on_evtchn(unsigned int irq)
{
	return retrigger_evtchn(evtchn_from_irq(irq));
}

1352
static void enable_dynirq(struct irq_data *data)
J
Jeremy Fitzhardinge 已提交
1353
{
1354
	int evtchn = evtchn_from_irq(data->irq);
J
Jeremy Fitzhardinge 已提交
1355 1356 1357 1358 1359

	if (VALID_EVTCHN(evtchn))
		unmask_evtchn(evtchn);
}

1360
static void disable_dynirq(struct irq_data *data)
J
Jeremy Fitzhardinge 已提交
1361
{
1362
	int evtchn = evtchn_from_irq(data->irq);
J
Jeremy Fitzhardinge 已提交
1363 1364 1365 1366 1367

	if (VALID_EVTCHN(evtchn))
		mask_evtchn(evtchn);
}

1368
static void ack_dynirq(struct irq_data *data)
J
Jeremy Fitzhardinge 已提交
1369
{
1370
	int evtchn = evtchn_from_irq(data->irq);
J
Jeremy Fitzhardinge 已提交
1371

1372
	irq_move_irq(data);
J
Jeremy Fitzhardinge 已提交
1373 1374

	if (VALID_EVTCHN(evtchn))
1375 1376 1377 1378 1379 1380 1381
		clear_evtchn(evtchn);
}

static void mask_ack_dynirq(struct irq_data *data)
{
	disable_dynirq(data);
	ack_dynirq(data);
J
Jeremy Fitzhardinge 已提交
1382 1383
}

1384
static int retrigger_dynirq(struct irq_data *data)
J
Jeremy Fitzhardinge 已提交
1385
{
1386
	return retrigger_evtchn(evtchn_from_irq(data->irq));
J
Jeremy Fitzhardinge 已提交
1387 1388
}

1389
static void restore_pirqs(void)
1390 1391 1392
{
	int pirq, rc, irq, gsi;
	struct physdev_map_pirq map_irq;
1393
	struct irq_info *info;
1394

1395 1396
	list_for_each_entry(info, &xen_irq_list_head, list) {
		if (info->type != IRQT_PIRQ)
1397 1398
			continue;

1399 1400 1401 1402
		pirq = info->u.pirq.pirq;
		gsi = info->u.pirq.gsi;
		irq = info->irq;

1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414
		/* save/restore of PT devices doesn't work, so at this point the
		 * only devices present are GSI based emulated devices */
		if (!gsi)
			continue;

		map_irq.domid = DOMID_SELF;
		map_irq.type = MAP_PIRQ_TYPE_GSI;
		map_irq.index = gsi;
		map_irq.pirq = pirq;

		rc = HYPERVISOR_physdev_op(PHYSDEVOP_map_pirq, &map_irq);
		if (rc) {
J
Joe Perches 已提交
1415 1416
			pr_warn("xen map irq failed gsi=%d irq=%d pirq=%d rc=%d\n",
				gsi, irq, pirq, rc);
1417
			xen_free_irq(irq);
1418 1419 1420 1421 1422
			continue;
		}

		printk(KERN_DEBUG "xen: --> irq=%d, pirq=%d\n", irq, map_irq.pirq);

1423
		__startup_pirq(irq);
1424 1425 1426
	}
}

1427 1428 1429 1430 1431 1432 1433 1434 1435
static void restore_cpu_virqs(unsigned int cpu)
{
	struct evtchn_bind_virq bind_virq;
	int virq, irq, evtchn;

	for (virq = 0; virq < NR_VIRQS; virq++) {
		if ((irq = per_cpu(virq_to_irq, cpu)[virq]) == -1)
			continue;

1436
		BUG_ON(virq_from_irq(irq) != virq);
1437 1438 1439 1440 1441 1442 1443 1444 1445 1446

		/* Get a new binding from Xen. */
		bind_virq.virq = virq;
		bind_virq.vcpu = cpu;
		if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq,
						&bind_virq) != 0)
			BUG();
		evtchn = bind_virq.port;

		/* Record the new mapping. */
1447
		(void)xen_irq_info_virq_setup(cpu, irq, evtchn, virq);
1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460
		bind_evtchn_to_cpu(evtchn, cpu);
	}
}

static void restore_cpu_ipis(unsigned int cpu)
{
	struct evtchn_bind_ipi bind_ipi;
	int ipi, irq, evtchn;

	for (ipi = 0; ipi < XEN_NR_IPIS; ipi++) {
		if ((irq = per_cpu(ipi_to_irq, cpu)[ipi]) == -1)
			continue;

1461
		BUG_ON(ipi_from_irq(irq) != ipi);
1462 1463 1464 1465 1466 1467 1468 1469 1470

		/* Get a new binding from Xen. */
		bind_ipi.vcpu = cpu;
		if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_ipi,
						&bind_ipi) != 0)
			BUG();
		evtchn = bind_ipi.port;

		/* Record the new mapping. */
1471
		(void)xen_irq_info_ipi_setup(cpu, irq, evtchn, ipi);
1472 1473 1474 1475
		bind_evtchn_to_cpu(evtchn, cpu);
	}
}

1476 1477 1478 1479 1480 1481 1482 1483
/* Clear an irq's pending state, in preparation for polling on it */
void xen_clear_irq_pending(int irq)
{
	int evtchn = evtchn_from_irq(irq);

	if (VALID_EVTCHN(evtchn))
		clear_evtchn(evtchn);
}
1484
EXPORT_SYMBOL(xen_clear_irq_pending);
1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503
void xen_set_irq_pending(int irq)
{
	int evtchn = evtchn_from_irq(irq);

	if (VALID_EVTCHN(evtchn))
		set_evtchn(evtchn);
}

bool xen_test_irq_pending(int irq)
{
	int evtchn = evtchn_from_irq(irq);
	bool ret = false;

	if (VALID_EVTCHN(evtchn))
		ret = test_evtchn(evtchn);

	return ret;
}

1504 1505 1506
/* Poll waiting for an irq to become pending with timeout.  In the usual case,
 * the irq will be disabled so it won't deliver an interrupt. */
void xen_poll_irq_timeout(int irq, u64 timeout)
1507 1508 1509 1510 1511 1512 1513
{
	evtchn_port_t evtchn = evtchn_from_irq(irq);

	if (VALID_EVTCHN(evtchn)) {
		struct sched_poll poll;

		poll.nr_ports = 1;
1514
		poll.timeout = timeout;
1515
		set_xen_guest_handle(poll.ports, &evtchn);
1516 1517 1518 1519 1520

		if (HYPERVISOR_sched_op(SCHEDOP_poll, &poll) != 0)
			BUG();
	}
}
1521 1522 1523 1524 1525 1526 1527
EXPORT_SYMBOL(xen_poll_irq_timeout);
/* Poll waiting for an irq to become pending.  In the usual case, the
 * irq will be disabled so it won't deliver an interrupt. */
void xen_poll_irq(int irq)
{
	xen_poll_irq_timeout(irq, 0 /* no timeout */);
}
1528

1529 1530 1531 1532
/* Check whether the IRQ line is shared with other guests. */
int xen_test_irq_shared(int irq)
{
	struct irq_info *info = info_for_irq(irq);
1533 1534 1535 1536 1537 1538
	struct physdev_irq_status_query irq_status;

	if (WARN_ON(!info))
		return -ENOENT;

	irq_status.irq = info->u.pirq.pirq;
1539 1540 1541 1542 1543 1544 1545

	if (HYPERVISOR_physdev_op(PHYSDEVOP_irq_status_query, &irq_status))
		return 0;
	return !(irq_status.flags & XENIRQSTAT_shared);
}
EXPORT_SYMBOL_GPL(xen_test_irq_shared);

1546 1547
void xen_irq_resume(void)
{
1548
	unsigned int cpu;
1549
	struct irq_info *info;
1550 1551

	/* New event-channel space is not 'live' yet. */
1552
	xen_evtchn_mask_all();
1553
	xen_evtchn_resume();
1554 1555

	/* No IRQ <-> event-channel mappings. */
1556 1557
	list_for_each_entry(info, &xen_irq_list_head, list)
		info->evtchn = 0; /* zap event-channel binding */
1558

1559
	clear_evtchn_to_irq_all();
1560 1561 1562 1563 1564

	for_each_possible_cpu(cpu) {
		restore_cpu_virqs(cpu);
		restore_cpu_ipis(cpu);
	}
1565

1566
	restore_pirqs();
1567 1568
}

J
Jeremy Fitzhardinge 已提交
1569
static struct irq_chip xen_dynamic_chip __read_mostly = {
1570
	.name			= "xen-dyn",
J
Jeremy Fitzhardinge 已提交
1571

1572 1573 1574
	.irq_disable		= disable_dynirq,
	.irq_mask		= disable_dynirq,
	.irq_unmask		= enable_dynirq,
J
Jeremy Fitzhardinge 已提交
1575

1576 1577 1578
	.irq_ack		= ack_dynirq,
	.irq_mask_ack		= mask_ack_dynirq,

1579 1580
	.irq_set_affinity	= set_affinity_irq,
	.irq_retrigger		= retrigger_dynirq,
J
Jeremy Fitzhardinge 已提交
1581 1582
};

1583
static struct irq_chip xen_pirq_chip __read_mostly = {
1584
	.name			= "xen-pirq",
1585

1586 1587 1588 1589
	.irq_startup		= startup_pirq,
	.irq_shutdown		= shutdown_pirq,
	.irq_enable		= enable_pirq,
	.irq_disable		= disable_pirq,
1590

1591 1592 1593 1594 1595 1596
	.irq_mask		= disable_dynirq,
	.irq_unmask		= enable_dynirq,

	.irq_ack		= eoi_pirq,
	.irq_eoi		= eoi_pirq,
	.irq_mask_ack		= mask_ack_pirq,
1597

1598
	.irq_set_affinity	= set_affinity_irq,
1599

1600
	.irq_retrigger		= retrigger_dynirq,
1601 1602
};

1603
static struct irq_chip xen_percpu_chip __read_mostly = {
1604
	.name			= "xen-percpu",
1605

1606 1607 1608
	.irq_disable		= disable_dynirq,
	.irq_mask		= disable_dynirq,
	.irq_unmask		= enable_dynirq,
1609

1610
	.irq_ack		= ack_dynirq,
1611 1612
};

1613 1614 1615 1616 1617 1618 1619 1620 1621 1622
int xen_set_callback_via(uint64_t via)
{
	struct xen_hvm_param a;
	a.domid = DOMID_SELF;
	a.index = HVM_PARAM_CALLBACK_IRQ;
	a.value = via;
	return HYPERVISOR_hvm_op(HVMOP_set_param, &a);
}
EXPORT_SYMBOL_GPL(xen_set_callback_via);

1623
#ifdef CONFIG_XEN_PVHVM
1624 1625 1626 1627 1628 1629 1630 1631
/* Vector callbacks are better than PCI interrupts to receive event
 * channel notifications because we can receive vector callbacks on any
 * vcpu and we don't need PCI support or APIC interactions. */
void xen_callback_vector(void)
{
	int rc;
	uint64_t callback_via;
	if (xen_have_vector_callback) {
1632
		callback_via = HVM_CALLBACK_VECTOR(HYPERVISOR_CALLBACK_VECTOR);
1633 1634
		rc = xen_set_callback_via(callback_via);
		if (rc) {
J
Joe Perches 已提交
1635
			pr_err("Request for Xen HVM callback vector failed\n");
1636 1637 1638
			xen_have_vector_callback = 0;
			return;
		}
J
Joe Perches 已提交
1639
		pr_info("Xen HVM callback vector for event delivery is enabled\n");
1640
		/* in the restore case the vector has already been allocated */
1641 1642 1643
		if (!test_bit(HYPERVISOR_CALLBACK_VECTOR, used_vectors))
			alloc_intr_gate(HYPERVISOR_CALLBACK_VECTOR,
					xen_hvm_callback_vector);
1644 1645
	}
}
1646 1647 1648
#else
void xen_callback_vector(void) {}
#endif
1649

1650 1651 1652 1653 1654 1655
#undef MODULE_PARAM_PREFIX
#define MODULE_PARAM_PREFIX "xen."

static bool fifo_events = true;
module_param(fifo_events, bool, 0);

S
Stefano Stabellini 已提交
1656
void __init xen_init_IRQ(void)
J
Jeremy Fitzhardinge 已提交
1657
{
1658 1659 1660 1661 1662 1663
	int ret = -EINVAL;

	if (fifo_events)
		ret = xen_evtchn_fifo_init();
	if (ret < 0)
		xen_evtchn_2l_init();
1664

1665 1666
	evtchn_to_irq = kcalloc(EVTCHN_ROW(xen_evtchn_max_channels()),
				sizeof(*evtchn_to_irq), GFP_KERNEL);
1667
	BUG_ON(!evtchn_to_irq);
J
Jeremy Fitzhardinge 已提交
1668 1669

	/* No event channels are 'live' right now. */
1670
	xen_evtchn_mask_all();
J
Jeremy Fitzhardinge 已提交
1671

S
Stefano Stabellini 已提交
1672 1673
	pirq_needs_eoi = pirq_needs_eoi_flag;

1674
#ifdef CONFIG_X86
1675 1676 1677 1678 1679 1680
	if (xen_pv_domain()) {
		irq_ctx_init(smp_processor_id());
		if (xen_initial_domain())
			pci_xen_initial_domain();
	}
	if (xen_feature(XENFEAT_hvm_callback_vector))
1681
		xen_callback_vector();
1682 1683

	if (xen_hvm_domain()) {
1684
		native_init_IRQ();
1685 1686 1687
		/* pci_xen_hvm_init must be called after native_init_IRQ so that
		 * __acpi_register_gsi can point at the right function */
		pci_xen_hvm_init();
1688
	} else {
1689
		int rc;
S
Stefano Stabellini 已提交
1690 1691 1692 1693 1694
		struct physdev_pirq_eoi_gmfn eoi_gmfn;

		pirq_eoi_map = (void *)__get_free_page(GFP_KERNEL|__GFP_ZERO);
		eoi_gmfn.gmfn = virt_to_mfn(pirq_eoi_map);
		rc = HYPERVISOR_physdev_op(PHYSDEVOP_pirq_eoi_gmfn_v2, &eoi_gmfn);
1695
		/* TODO: No PVH support for PIRQ EOI */
S
Stefano Stabellini 已提交
1696 1697 1698 1699 1700
		if (rc != 0) {
			free_page((unsigned long) pirq_eoi_map);
			pirq_eoi_map = NULL;
		} else
			pirq_needs_eoi = pirq_check_eoi_map;
1701
	}
1702
#endif
J
Jeremy Fitzhardinge 已提交
1703
}