events_base.c 37.2 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_irqs_dynamic(int nvec)
392
{
393
	int i, irq = irq_alloc_descs(-1, 0, nvec, -1);
394

395 396 397 398
	if (irq >= 0) {
		for (i = 0; i < nvec; i++)
			xen_irq_init(irq + i);
	}
399

J
Jeremy Fitzhardinge 已提交
400
	return irq;
401 402
}

403 404 405 406 407 408
static inline int __must_check xen_allocate_irq_dynamic(void)
{

	return xen_allocate_irqs_dynamic(1);
}

409
static int __must_check xen_allocate_irq_gsi(unsigned gsi)
410 411 412
{
	int irq;

413 414 415 416 417 418 419
	/*
	 * 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())
420 421 422 423
		return xen_allocate_irq_dynamic();

	/* Legacy IRQ descriptors are already allocated by the arch. */
	if (gsi < NR_IRQS_LEGACY)
424 425 426
		irq = gsi;
	else
		irq = irq_alloc_desc_at(gsi, -1);
427

428
	xen_irq_init(irq);
429 430 431 432 433 434

	return irq;
}

static void xen_free_irq(unsigned irq)
{
T
Thomas Gleixner 已提交
435
	struct irq_info *info = irq_get_handler_data(irq);
436

437 438 439
	if (WARN_ON(!info))
		return;

440
	list_del(&info->list);
441

T
Thomas Gleixner 已提交
442
	irq_set_handler_data(irq, NULL);
443

444 445
	WARN_ON(info->refcnt > 0);

446 447
	kfree(info);

448 449 450 451
	/* Legacy IRQ descriptors are managed by the arch. */
	if (irq < NR_IRQS_LEGACY)
		return;

452 453 454
	irq_free_desc(irq);
}

455 456 457 458 459 460 461 462 463
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();
}

464 465 466 467 468 469 470
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
471
	irq_status.irq = pirq_from_irq(irq);
472 473 474 475 476 477 478 479
	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;
}

480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502
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);
}

503
static unsigned int __startup_pirq(unsigned int irq)
504 505 506 507
{
	struct evtchn_bind_pirq bind_pirq;
	struct irq_info *info = info_for_irq(irq);
	int evtchn = evtchn_from_irq(irq);
508
	int rc;
509 510 511 512 513 514

	BUG_ON(info->type != IRQT_PIRQ);

	if (VALID_EVTCHN(evtchn))
		goto out;

S
Stefano Stabellini 已提交
515
	bind_pirq.pirq = pirq_from_irq(irq);
516
	/* NB. We are happy to share unless we are probing. */
517 518 519 520
	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) {
521
		pr_warn("Failed to obtain physical IRQ %d\n", irq);
522 523 524 525 526 527
		return 0;
	}
	evtchn = bind_pirq.port;

	pirq_query_unmask(irq);

528 529 530 531 532 533 534
	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;
	}
535 536 537 538 539
	bind_evtchn_to_cpu(evtchn, 0);
	info->evtchn = evtchn;

out:
	unmask_evtchn(evtchn);
540
	eoi_pirq(irq_get_irq_data(irq));
541 542 543 544

	return 0;
}

545 546 547 548 549 550
static unsigned int startup_pirq(struct irq_data *data)
{
	return __startup_pirq(data->irq);
}

static void shutdown_pirq(struct irq_data *data)
551
{
552
	unsigned int irq = data->irq;
553
	struct irq_info *info = info_for_irq(irq);
554
	unsigned evtchn = evtchn_from_irq(irq);
555 556 557 558 559 560 561

	BUG_ON(info->type != IRQT_PIRQ);

	if (!VALID_EVTCHN(evtchn))
		return;

	mask_evtchn(evtchn);
562 563
	xen_evtchn_close(evtchn);
	xen_irq_info_cleanup(info);
564 565
}

566
static void enable_pirq(struct irq_data *data)
567
{
568
	startup_pirq(data);
569 570
}

571
static void disable_pirq(struct irq_data *data)
572
{
573
	disable_dynirq(data);
574 575
}

576
int xen_irq_from_gsi(unsigned gsi)
577
{
578
	struct irq_info *info;
579

580 581
	list_for_each_entry(info, &xen_irq_list_head, list) {
		if (info->type != IRQT_PIRQ)
582 583
			continue;

584 585
		if (info->u.pirq.gsi == gsi)
			return info->irq;
586 587 588 589
	}

	return -1;
}
590
EXPORT_SYMBOL_GPL(xen_irq_from_gsi);
591

592 593 594 595 596 597 598 599 600 601 602 603
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)) {
604 605 606
		unsigned int cpu = cpu_from_irq(irq);

		xen_evtchn_close(evtchn);
607 608 609

		switch (type_from_irq(irq)) {
		case IRQT_VIRQ:
610
			per_cpu(virq_to_irq, cpu)[virq_from_irq(irq)] = -1;
611 612
			break;
		case IRQT_IPI:
613
			per_cpu(ipi_to_irq, cpu)[ipi_from_irq(irq)] = -1;
614 615 616 617 618
			break;
		default:
			break;
		}

619
		xen_irq_info_cleanup(info);
620 621 622 623 624 625 626
	}

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

	xen_free_irq(irq);
}

I
Ian Campbell 已提交
627 628 629
/*
 * 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
630 631 632
 *
 * 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.
633 634 635
 *
 * Shareable implies level triggered, not shareable implies edge
 * triggered here.
636
 */
637 638
int xen_bind_pirq_gsi_to_irq(unsigned gsi,
			     unsigned pirq, int shareable, char *name)
639
{
640
	int irq = -1;
641
	struct physdev_irq irq_op;
642
	int ret;
643

644
	mutex_lock(&irq_mapping_update_lock);
645

646
	irq = xen_irq_from_gsi(gsi);
647
	if (irq != -1) {
J
Joe Perches 已提交
648 649
		pr_info("%s: returning irq %d for gsi %u\n",
			__func__, irq, gsi);
650
		goto out;
651 652
	}

653
	irq = xen_allocate_irq_gsi(gsi);
654 655
	if (irq < 0)
		goto out;
656 657

	irq_op.irq = irq;
658 659 660 661 662 663 664
	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)) {
665
		xen_free_irq(irq);
666 667 668 669
		irq = -ENOSPC;
		goto out;
	}

670
	ret = xen_irq_info_pirq_setup(irq, 0, pirq, gsi, DOMID_SELF,
671
			       shareable ? PIRQ_SHAREABLE : 0);
672 673 674 675 676
	if (ret < 0) {
		__unbind_from_irq(irq);
		irq = ret;
		goto out;
	}
677

678 679
	pirq_query_unmask(irq);
	/* We try to use the handler with the appropriate semantic for the
680 681
	 * type of interrupt: if the interrupt is an edge triggered
	 * interrupt we use handle_edge_irq.
682
	 *
683 684
	 * On the other hand if the interrupt is level triggered we use
	 * handle_fasteoi_irq like the native code does for this kind of
685
	 * interrupts.
686
	 *
687 688 689 690 691 692 693
	 * 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.
	 */
694
	if (shareable)
695 696 697 698 699 700
		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);

701
out:
702
	mutex_unlock(&irq_mapping_update_lock);
703 704 705 706

	return irq;
}

707
#ifdef CONFIG_PCI_MSI
708
int xen_allocate_pirq_msi(struct pci_dev *dev, struct msi_desc *msidesc)
709
{
710
	int rc;
711 712
	struct physdev_get_free_pirq op_get_free_pirq;

713
	op_get_free_pirq.type = MAP_PIRQ_TYPE_MSI;
714 715
	rc = HYPERVISOR_physdev_op(PHYSDEVOP_get_free_pirq, &op_get_free_pirq);

716 717 718 719
	WARN_ONCE(rc == -ENOSYS,
		  "hypervisor does not support the PHYSDEVOP_get_free_pirq interface\n");

	return rc ? -1 : op_get_free_pirq.pirq;
720 721
}

722
int xen_bind_pirq_msi_to_irq(struct pci_dev *dev, struct msi_desc *msidesc,
723
			     int pirq, int nvec, const char *name, domid_t domid)
S
Stefano Stabellini 已提交
724
{
725
	int i, irq, ret;
726

727
	mutex_lock(&irq_mapping_update_lock);
S
Stefano Stabellini 已提交
728

729
	irq = xen_allocate_irqs_dynamic(nvec);
730
	if (irq < 0)
731
		goto out;
S
Stefano Stabellini 已提交
732

733 734 735 736 737 738 739 740
	for (i = 0; i < nvec; i++) {
		irq_set_chip_and_handler_name(irq + i, &xen_pirq_chip, handle_edge_irq, name);

		ret = xen_irq_info_pirq_setup(irq + i, 0, pirq + i, 0, domid,
					      i == 0 ? 0 : PIRQ_MSI_GROUP);
		if (ret < 0)
			goto error_irq;
	}
S
Stefano Stabellini 已提交
741

742
	ret = irq_set_msi_desc(irq, msidesc);
743 744
	if (ret < 0)
		goto error_irq;
S
Stefano Stabellini 已提交
745
out:
746
	mutex_unlock(&irq_mapping_update_lock);
747
	return irq;
748
error_irq:
749 750
	for (; i >= 0; i--)
		__unbind_from_irq(irq + i);
751
	mutex_unlock(&irq_mapping_update_lock);
752
	return ret;
S
Stefano Stabellini 已提交
753
}
754 755
#endif

756 757
int xen_destroy_irq(int irq)
{
758 759
	struct physdev_unmap_pirq unmap_irq;
	struct irq_info *info = info_for_irq(irq);
760 761
	int rc = -ENOENT;

762
	mutex_lock(&irq_mapping_update_lock);
763

764 765 766 767 768 769
	/*
	 * If trying to remove a vector in a MSI group different
	 * than the first one skip the PIRQ unmap unless this vector
	 * is the first one in the group.
	 */
	if (xen_initial_domain() && !(info->u.pirq.flags & PIRQ_MSI_GROUP)) {
770
		unmap_irq.pirq = info->u.pirq.pirq;
771
		unmap_irq.domid = info->u.pirq.domid;
772
		rc = HYPERVISOR_physdev_op(PHYSDEVOP_unmap_pirq, &unmap_irq);
773 774 775 776 777
		/* 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
778
			pr_info("domain %d does not have %d anymore\n",
779 780
				info->u.pirq.domid, info->u.pirq.pirq);
		else if (rc) {
J
Joe Perches 已提交
781
			pr_warn("unmap irq failed %d\n", rc);
782 783 784
			goto out;
		}
	}
785

786
	xen_free_irq(irq);
787 788

out:
789
	mutex_unlock(&irq_mapping_update_lock);
790 791 792
	return rc;
}

793
int xen_irq_from_pirq(unsigned pirq)
794
{
795
	int irq;
796

797
	struct irq_info *info;
J
Jeremy Fitzhardinge 已提交
798

799
	mutex_lock(&irq_mapping_update_lock);
800 801

	list_for_each_entry(info, &xen_irq_list_head, list) {
802
		if (info->type != IRQT_PIRQ)
803 804 805 806 807 808 809
			continue;
		irq = info->irq;
		if (info->u.pirq.pirq == pirq)
			goto out;
	}
	irq = -1;
out:
810
	mutex_unlock(&irq_mapping_update_lock);
811 812

	return irq;
813 814
}

815 816 817 818 819 820

int xen_pirq_from_irq(unsigned irq)
{
	return pirq_from_irq(irq);
}
EXPORT_SYMBOL_GPL(xen_pirq_from_irq);
821

822
int bind_evtchn_to_irq(unsigned int evtchn)
J
Jeremy Fitzhardinge 已提交
823 824
{
	int irq;
825
	int ret;
J
Jeremy Fitzhardinge 已提交
826

827 828 829
	if (evtchn >= xen_evtchn_max_channels())
		return -ENOMEM;

830
	mutex_lock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
831

832
	irq = get_evtchn_to_irq(evtchn);
J
Jeremy Fitzhardinge 已提交
833 834

	if (irq == -1) {
835
		irq = xen_allocate_irq_dynamic();
836
		if (irq < 0)
837
			goto out;
J
Jeremy Fitzhardinge 已提交
838

T
Thomas Gleixner 已提交
839
		irq_set_chip_and_handler_name(irq, &xen_dynamic_chip,
840
					      handle_edge_irq, "event");
J
Jeremy Fitzhardinge 已提交
841

842 843 844 845 846 847
		ret = xen_irq_info_evtchn_setup(irq, evtchn);
		if (ret < 0) {
			__unbind_from_irq(irq);
			irq = ret;
			goto out;
		}
848 849
		/* New interdomain events are bound to VCPU 0. */
		bind_evtchn_to_cpu(evtchn, 0);
850 851 852
	} else {
		struct irq_info *info = info_for_irq(irq);
		WARN_ON(info == NULL || info->type != IRQT_EVTCHN);
J
Jeremy Fitzhardinge 已提交
853 854
	}

855
out:
856
	mutex_unlock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
857 858 859

	return irq;
}
860
EXPORT_SYMBOL_GPL(bind_evtchn_to_irq);
J
Jeremy Fitzhardinge 已提交
861

J
Jeremy Fitzhardinge 已提交
862 863 864 865
static int bind_ipi_to_irq(unsigned int ipi, unsigned int cpu)
{
	struct evtchn_bind_ipi bind_ipi;
	int evtchn, irq;
866
	int ret;
J
Jeremy Fitzhardinge 已提交
867

868
	mutex_lock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
869 870

	irq = per_cpu(ipi_to_irq, cpu)[ipi];
871

J
Jeremy Fitzhardinge 已提交
872
	if (irq == -1) {
873
		irq = xen_allocate_irq_dynamic();
J
Jeremy Fitzhardinge 已提交
874 875 876
		if (irq < 0)
			goto out;

T
Thomas Gleixner 已提交
877
		irq_set_chip_and_handler_name(irq, &xen_percpu_chip,
878
					      handle_percpu_irq, "ipi");
J
Jeremy Fitzhardinge 已提交
879 880 881 882 883 884 885

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

886 887 888 889 890 891
		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
892
		bind_evtchn_to_cpu(evtchn, cpu);
893 894 895
	} else {
		struct irq_info *info = info_for_irq(irq);
		WARN_ON(info == NULL || info->type != IRQT_IPI);
J
Jeremy Fitzhardinge 已提交
896 897 898
	}

 out:
899
	mutex_unlock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
900 901 902
	return irq;
}

903 904 905 906 907 908 909 910 911 912 913 914 915 916 917
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);
}

918 919 920 921 922 923
static int find_virq(unsigned int virq, unsigned int cpu)
{
	struct evtchn_status status;
	int port, rc = -ENOENT;

	memset(&status, 0, sizeof(status));
924
	for (port = 0; port < xen_evtchn_max_channels(); port++) {
925 926 927 928 929 930 931 932 933 934 935 936 937 938
		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
939

940 941 942 943 944 945 946 947 948 949 950 951 952
/**
 * 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);

953
int bind_virq_to_irq(unsigned int virq, unsigned int cpu)
J
Jeremy Fitzhardinge 已提交
954 955
{
	struct evtchn_bind_virq bind_virq;
956
	int evtchn, irq, ret;
J
Jeremy Fitzhardinge 已提交
957

958
	mutex_lock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
959 960 961 962

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

	if (irq == -1) {
963
		irq = xen_allocate_irq_dynamic();
964
		if (irq < 0)
965
			goto out;
966

T
Thomas Gleixner 已提交
967
		irq_set_chip_and_handler_name(irq, &xen_percpu_chip,
968 969
					      handle_percpu_irq, "virq");

J
Jeremy Fitzhardinge 已提交
970 971
		bind_virq.virq = virq;
		bind_virq.vcpu = cpu;
972 973 974 975 976 977 978 979 980 981
		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
982

983 984 985 986 987 988
		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
989 990

		bind_evtchn_to_cpu(evtchn, cpu);
991 992 993
	} else {
		struct irq_info *info = info_for_irq(irq);
		WARN_ON(info == NULL || info->type != IRQT_VIRQ);
J
Jeremy Fitzhardinge 已提交
994 995
	}

996
out:
997
	mutex_unlock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
998 999 1000 1001 1002 1003

	return irq;
}

static void unbind_from_irq(unsigned int irq)
{
1004
	mutex_lock(&irq_mapping_update_lock);
1005
	__unbind_from_irq(irq);
1006
	mutex_unlock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
1007 1008 1009
}

int bind_evtchn_to_irqhandler(unsigned int evtchn,
1010
			      irq_handler_t handler,
J
Jeremy Fitzhardinge 已提交
1011 1012 1013
			      unsigned long irqflags,
			      const char *devname, void *dev_id)
{
1014
	int irq, retval;
J
Jeremy Fitzhardinge 已提交
1015 1016

	irq = bind_evtchn_to_irq(evtchn);
1017 1018
	if (irq < 0)
		return irq;
J
Jeremy Fitzhardinge 已提交
1019 1020 1021 1022 1023 1024 1025 1026 1027 1028
	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);

1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051
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
1052
int bind_virq_to_irqhandler(unsigned int virq, unsigned int cpu,
1053
			    irq_handler_t handler,
J
Jeremy Fitzhardinge 已提交
1054 1055
			    unsigned long irqflags, const char *devname, void *dev_id)
{
1056
	int irq, retval;
J
Jeremy Fitzhardinge 已提交
1057 1058

	irq = bind_virq_to_irq(virq, cpu);
1059 1060
	if (irq < 0)
		return irq;
J
Jeremy Fitzhardinge 已提交
1061 1062 1063 1064 1065 1066 1067 1068 1069 1070
	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
1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083
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;

1084
	irqflags |= IRQF_NO_SUSPEND | IRQF_FORCE_RESUME | IRQF_EARLY_RESUME;
J
Jeremy Fitzhardinge 已提交
1085 1086 1087 1088 1089 1090 1091 1092 1093
	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
1094 1095
void unbind_from_irqhandler(unsigned int irq, void *dev_id)
{
1096 1097 1098 1099
	struct irq_info *info = irq_get_handler_data(irq);

	if (WARN_ON(!info))
		return;
J
Jeremy Fitzhardinge 已提交
1100 1101 1102 1103 1104
	free_irq(irq, dev_id);
	unbind_from_irq(irq);
}
EXPORT_SYMBOL_GPL(unbind_from_irqhandler);

1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121
/**
 * 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);

1122 1123
int evtchn_make_refcounted(unsigned int evtchn)
{
1124
	int irq = get_evtchn_to_irq(evtchn);
1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148
	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;

1149
	if (evtchn >= xen_evtchn_max_channels())
1150 1151
		return -EINVAL;

1152 1153
	mutex_lock(&irq_mapping_update_lock);

1154
	irq = get_evtchn_to_irq(evtchn);
1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177
	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)
{
1178
	int irq = get_evtchn_to_irq(evtchn);
1179 1180 1181 1182 1183 1184
	if (WARN_ON(irq == -1))
		return;
	unbind_from_irq(irq);
}
EXPORT_SYMBOL_GPL(evtchn_put);

J
Jeremy Fitzhardinge 已提交
1185 1186
void xen_send_IPI_one(unsigned int cpu, enum ipi_vector vector)
{
1187 1188
	int irq;

1189
#ifdef CONFIG_X86
1190 1191 1192 1193 1194 1195
	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;
	}
1196
#endif
1197
	irq = per_cpu(ipi_to_irq, cpu)[vector];
J
Jeremy Fitzhardinge 已提交
1198 1199 1200 1201
	BUG_ON(irq < 0);
	notify_remote_via_irq(irq);
}

1202 1203
static DEFINE_PER_CPU(unsigned, xed_nesting_count);

1204
static void __xen_evtchn_do_upcall(void)
J
Jeremy Fitzhardinge 已提交
1205
{
C
Christoph Lameter 已提交
1206
	struct vcpu_info *vcpu_info = __this_cpu_read(xen_vcpu);
1207
	int cpu = get_cpu();
1208
	unsigned count;
J
Jeremy Fitzhardinge 已提交
1209

1210 1211
	do {
		vcpu_info->evtchn_upcall_pending = 0;
J
Jeremy Fitzhardinge 已提交
1212

C
Christoph Lameter 已提交
1213
		if (__this_cpu_inc_return(xed_nesting_count) - 1)
1214
			goto out;
J
Jeremy Fitzhardinge 已提交
1215

1216
		xen_evtchn_handle_events(cpu);
J
Jeremy Fitzhardinge 已提交
1217

1218 1219
		BUG_ON(!irqs_disabled());

C
Christoph Lameter 已提交
1220 1221
		count = __this_cpu_read(xed_nesting_count);
		__this_cpu_write(xed_nesting_count, 0);
1222
	} while (count != 1 || vcpu_info->evtchn_upcall_pending);
1223 1224

out:
1225 1226 1227 1228 1229 1230 1231 1232

	put_cpu();
}

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

1233
	irq_enter();
1234
#ifdef CONFIG_X86
1235
	exit_idle();
1236
	inc_irq_stat(irq_hv_callback_count);
1237
#endif
1238 1239 1240

	__xen_evtchn_do_upcall();

1241 1242
	irq_exit();
	set_irq_regs(old_regs);
1243
}
1244

1245 1246 1247
void xen_hvm_evtchn_do_upcall(void)
{
	__xen_evtchn_do_upcall();
J
Jeremy Fitzhardinge 已提交
1248
}
1249
EXPORT_SYMBOL_GPL(xen_hvm_evtchn_do_upcall);
J
Jeremy Fitzhardinge 已提交
1250

J
Jeremy Fitzhardinge 已提交
1251 1252 1253
/* Rebind a new event channel to an existing irq. */
void rebind_evtchn_irq(int evtchn, int irq)
{
J
Jeremy Fitzhardinge 已提交
1254 1255
	struct irq_info *info = info_for_irq(irq);

1256 1257 1258
	if (WARN_ON(!info))
		return;

J
Jeremy Fitzhardinge 已提交
1259 1260 1261 1262
	/* Make sure the irq is masked, since the new event channel
	   will also be masked. */
	disable_irq(irq);

1263
	mutex_lock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
1264 1265

	/* After resume the irq<->evtchn mappings are all cleared out */
1266
	BUG_ON(get_evtchn_to_irq(evtchn) != -1);
J
Jeremy Fitzhardinge 已提交
1267
	/* Expect irq to have been bound before,
J
Jeremy Fitzhardinge 已提交
1268 1269
	   so there should be a proper type */
	BUG_ON(info->type == IRQT_UNBOUND);
J
Jeremy Fitzhardinge 已提交
1270

1271
	(void)xen_irq_info_evtchn_setup(irq, evtchn);
J
Jeremy Fitzhardinge 已提交
1272

1273
	mutex_unlock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
1274 1275

	/* new event channels are always bound to cpu 0 */
1276
	irq_set_affinity(irq, cpumask_of(0));
J
Jeremy Fitzhardinge 已提交
1277 1278 1279 1280 1281

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

J
Jeremy Fitzhardinge 已提交
1282
/* Rebind an evtchn so that it gets delivered to a specific cpu */
1283
static int rebind_irq_to_cpu(unsigned irq, unsigned tcpu)
J
Jeremy Fitzhardinge 已提交
1284 1285 1286
{
	struct evtchn_bind_vcpu bind_vcpu;
	int evtchn = evtchn_from_irq(irq);
1287
	int masked;
J
Jeremy Fitzhardinge 已提交
1288

1289 1290 1291 1292 1293 1294 1295 1296
	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)
1297
		return -1;
J
Jeremy Fitzhardinge 已提交
1298 1299 1300 1301 1302

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

1303 1304 1305 1306
	/*
	 * Mask the event while changing the VCPU binding to prevent
	 * it being delivered on an unexpected VCPU.
	 */
1307
	masked = test_and_set_mask(evtchn);
1308

J
Jeremy Fitzhardinge 已提交
1309 1310 1311 1312 1313 1314 1315 1316
	/*
	 * 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);

1317 1318 1319
	if (!masked)
		unmask_evtchn(evtchn);

1320 1321
	return 0;
}
J
Jeremy Fitzhardinge 已提交
1322

1323 1324
static int set_affinity_irq(struct irq_data *data, const struct cpumask *dest,
			    bool force)
J
Jeremy Fitzhardinge 已提交
1325
{
1326
	unsigned tcpu = cpumask_first_and(dest, cpu_online_mask);
1327

1328
	return rebind_irq_to_cpu(data->irq, tcpu);
J
Jeremy Fitzhardinge 已提交
1329 1330
}

1331
static void enable_dynirq(struct irq_data *data)
J
Jeremy Fitzhardinge 已提交
1332
{
1333
	int evtchn = evtchn_from_irq(data->irq);
J
Jeremy Fitzhardinge 已提交
1334 1335 1336 1337 1338

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

1339
static void disable_dynirq(struct irq_data *data)
J
Jeremy Fitzhardinge 已提交
1340
{
1341
	int evtchn = evtchn_from_irq(data->irq);
J
Jeremy Fitzhardinge 已提交
1342 1343 1344 1345 1346

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

1347
static void ack_dynirq(struct irq_data *data)
J
Jeremy Fitzhardinge 已提交
1348
{
1349
	int evtchn = evtchn_from_irq(data->irq);
J
Jeremy Fitzhardinge 已提交
1350

1351
	irq_move_irq(data);
J
Jeremy Fitzhardinge 已提交
1352 1353

	if (VALID_EVTCHN(evtchn))
1354 1355 1356 1357 1358 1359 1360
		clear_evtchn(evtchn);
}

static void mask_ack_dynirq(struct irq_data *data)
{
	disable_dynirq(data);
	ack_dynirq(data);
J
Jeremy Fitzhardinge 已提交
1361 1362
}

1363
static int retrigger_dynirq(struct irq_data *data)
J
Jeremy Fitzhardinge 已提交
1364
{
1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376
	unsigned int evtchn = evtchn_from_irq(data->irq);
	int masked;

	if (!VALID_EVTCHN(evtchn))
		return 0;

	masked = test_and_set_mask(evtchn);
	set_evtchn(evtchn);
	if (!masked)
		unmask_evtchn(evtchn);

	return 1;
J
Jeremy Fitzhardinge 已提交
1377 1378
}

1379
static void restore_pirqs(void)
1380 1381 1382
{
	int pirq, rc, irq, gsi;
	struct physdev_map_pirq map_irq;
1383
	struct irq_info *info;
1384

1385 1386
	list_for_each_entry(info, &xen_irq_list_head, list) {
		if (info->type != IRQT_PIRQ)
1387 1388
			continue;

1389 1390 1391 1392
		pirq = info->u.pirq.pirq;
		gsi = info->u.pirq.gsi;
		irq = info->irq;

1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404
		/* 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
1405 1406
			pr_warn("xen map irq failed gsi=%d irq=%d pirq=%d rc=%d\n",
				gsi, irq, pirq, rc);
1407
			xen_free_irq(irq);
1408 1409 1410 1411 1412
			continue;
		}

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

1413
		__startup_pirq(irq);
1414 1415 1416
	}
}

1417 1418 1419 1420 1421 1422 1423 1424 1425
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;

1426
		BUG_ON(virq_from_irq(irq) != virq);
1427 1428 1429 1430 1431 1432 1433 1434 1435 1436

		/* 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. */
1437
		(void)xen_irq_info_virq_setup(cpu, irq, evtchn, virq);
1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450
		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;

1451
		BUG_ON(ipi_from_irq(irq) != ipi);
1452 1453 1454 1455 1456 1457 1458 1459 1460

		/* 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. */
1461
		(void)xen_irq_info_ipi_setup(cpu, irq, evtchn, ipi);
1462 1463 1464 1465
		bind_evtchn_to_cpu(evtchn, cpu);
	}
}

1466 1467 1468 1469 1470 1471 1472 1473
/* 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);
}
1474
EXPORT_SYMBOL(xen_clear_irq_pending);
1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493
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;
}

1494 1495 1496
/* 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)
1497 1498 1499 1500 1501 1502 1503
{
	evtchn_port_t evtchn = evtchn_from_irq(irq);

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

		poll.nr_ports = 1;
1504
		poll.timeout = timeout;
1505
		set_xen_guest_handle(poll.ports, &evtchn);
1506 1507 1508 1509 1510

		if (HYPERVISOR_sched_op(SCHEDOP_poll, &poll) != 0)
			BUG();
	}
}
1511 1512 1513 1514 1515 1516 1517
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 */);
}
1518

1519 1520 1521 1522
/* 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);
1523 1524 1525 1526 1527 1528
	struct physdev_irq_status_query irq_status;

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

	irq_status.irq = info->u.pirq.pirq;
1529 1530 1531 1532 1533 1534 1535

	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);

1536 1537
void xen_irq_resume(void)
{
1538
	unsigned int cpu;
1539
	struct irq_info *info;
1540 1541

	/* New event-channel space is not 'live' yet. */
1542
	xen_evtchn_mask_all();
1543
	xen_evtchn_resume();
1544 1545

	/* No IRQ <-> event-channel mappings. */
1546 1547
	list_for_each_entry(info, &xen_irq_list_head, list)
		info->evtchn = 0; /* zap event-channel binding */
1548

1549
	clear_evtchn_to_irq_all();
1550 1551 1552 1553 1554

	for_each_possible_cpu(cpu) {
		restore_cpu_virqs(cpu);
		restore_cpu_ipis(cpu);
	}
1555

1556
	restore_pirqs();
1557 1558
}

J
Jeremy Fitzhardinge 已提交
1559
static struct irq_chip xen_dynamic_chip __read_mostly = {
1560
	.name			= "xen-dyn",
J
Jeremy Fitzhardinge 已提交
1561

1562 1563 1564
	.irq_disable		= disable_dynirq,
	.irq_mask		= disable_dynirq,
	.irq_unmask		= enable_dynirq,
J
Jeremy Fitzhardinge 已提交
1565

1566 1567 1568
	.irq_ack		= ack_dynirq,
	.irq_mask_ack		= mask_ack_dynirq,

1569 1570
	.irq_set_affinity	= set_affinity_irq,
	.irq_retrigger		= retrigger_dynirq,
J
Jeremy Fitzhardinge 已提交
1571 1572
};

1573
static struct irq_chip xen_pirq_chip __read_mostly = {
1574
	.name			= "xen-pirq",
1575

1576 1577 1578 1579
	.irq_startup		= startup_pirq,
	.irq_shutdown		= shutdown_pirq,
	.irq_enable		= enable_pirq,
	.irq_disable		= disable_pirq,
1580

1581 1582 1583 1584 1585 1586
	.irq_mask		= disable_dynirq,
	.irq_unmask		= enable_dynirq,

	.irq_ack		= eoi_pirq,
	.irq_eoi		= eoi_pirq,
	.irq_mask_ack		= mask_ack_pirq,
1587

1588
	.irq_set_affinity	= set_affinity_irq,
1589

1590
	.irq_retrigger		= retrigger_dynirq,
1591 1592
};

1593
static struct irq_chip xen_percpu_chip __read_mostly = {
1594
	.name			= "xen-percpu",
1595

1596 1597 1598
	.irq_disable		= disable_dynirq,
	.irq_mask		= disable_dynirq,
	.irq_unmask		= enable_dynirq,
1599

1600
	.irq_ack		= ack_dynirq,
1601 1602
};

1603 1604 1605 1606 1607 1608 1609 1610 1611 1612
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);

1613
#ifdef CONFIG_XEN_PVHVM
1614 1615 1616 1617 1618 1619 1620 1621
/* 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) {
1622
		callback_via = HVM_CALLBACK_VECTOR(HYPERVISOR_CALLBACK_VECTOR);
1623 1624
		rc = xen_set_callback_via(callback_via);
		if (rc) {
J
Joe Perches 已提交
1625
			pr_err("Request for Xen HVM callback vector failed\n");
1626 1627 1628
			xen_have_vector_callback = 0;
			return;
		}
J
Joe Perches 已提交
1629
		pr_info("Xen HVM callback vector for event delivery is enabled\n");
1630
		/* in the restore case the vector has already been allocated */
1631 1632 1633
		if (!test_bit(HYPERVISOR_CALLBACK_VECTOR, used_vectors))
			alloc_intr_gate(HYPERVISOR_CALLBACK_VECTOR,
					xen_hvm_callback_vector);
1634 1635
	}
}
1636 1637 1638
#else
void xen_callback_vector(void) {}
#endif
1639

1640 1641 1642 1643 1644 1645
#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
1646
void __init xen_init_IRQ(void)
J
Jeremy Fitzhardinge 已提交
1647
{
1648 1649 1650 1651 1652 1653
	int ret = -EINVAL;

	if (fifo_events)
		ret = xen_evtchn_fifo_init();
	if (ret < 0)
		xen_evtchn_2l_init();
1654

1655 1656
	evtchn_to_irq = kcalloc(EVTCHN_ROW(xen_evtchn_max_channels()),
				sizeof(*evtchn_to_irq), GFP_KERNEL);
1657
	BUG_ON(!evtchn_to_irq);
J
Jeremy Fitzhardinge 已提交
1658 1659

	/* No event channels are 'live' right now. */
1660
	xen_evtchn_mask_all();
J
Jeremy Fitzhardinge 已提交
1661

S
Stefano Stabellini 已提交
1662 1663
	pirq_needs_eoi = pirq_needs_eoi_flag;

1664
#ifdef CONFIG_X86
1665 1666 1667 1668 1669 1670
	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))
1671
		xen_callback_vector();
1672 1673

	if (xen_hvm_domain()) {
1674
		native_init_IRQ();
1675 1676 1677
		/* 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();
1678
	} else {
1679
		int rc;
S
Stefano Stabellini 已提交
1680 1681 1682 1683 1684
		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);
1685
		/* TODO: No PVH support for PIRQ EOI */
S
Stefano Stabellini 已提交
1686 1687 1688 1689 1690
		if (rc != 0) {
			free_page((unsigned long) pirq_eoi_map);
			pirq_eoi_map = NULL;
		} else
			pirq_needs_eoi = pirq_check_eoi_map;
1691
	}
1692
#endif
J
Jeremy Fitzhardinge 已提交
1693
}