events_base.c 49.0 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0-only
J
Jeremy Fitzhardinge 已提交
2 3 4 5 6 7 8
/*
 * 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
9
 * chip.  When an event is received, it is mapped to an irq and sent
J
Jeremy Fitzhardinge 已提交
10 11 12 13 14 15 16 17 18 19
 * 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.
20
 * 4. PIRQs - Hardware interrupts.
J
Jeremy Fitzhardinge 已提交
21 22 23 24
 *
 * Jeremy Fitzhardinge <jeremy@xensource.com>, XenSource Inc, 2007
 */

25 26
#define pr_fmt(fmt) "xen:" KBUILD_MODNAME ": " fmt

J
Jeremy Fitzhardinge 已提交
27 28 29
#include <linux/linkage.h>
#include <linux/interrupt.h>
#include <linux/irq.h>
30
#include <linux/moduleparam.h>
J
Jeremy Fitzhardinge 已提交
31
#include <linux/string.h>
32
#include <linux/memblock.h>
33
#include <linux/slab.h>
34
#include <linux/irqnr.h>
35
#include <linux/pci.h>
36
#include <linux/spinlock.h>
37
#include <linux/cpuhotplug.h>
38 39
#include <linux/atomic.h>
#include <linux/ktime.h>
J
Jeremy Fitzhardinge 已提交
40

41
#ifdef CONFIG_X86
42
#include <asm/desc.h>
J
Jeremy Fitzhardinge 已提交
43
#include <asm/ptrace.h>
44
#include <asm/idtentry.h>
J
Jeremy Fitzhardinge 已提交
45
#include <asm/irq.h>
46
#include <asm/io_apic.h>
47
#include <asm/i8259.h>
48
#include <asm/xen/pci.h>
49 50
#endif
#include <asm/sync_bitops.h>
J
Jeremy Fitzhardinge 已提交
51
#include <asm/xen/hypercall.h>
52
#include <asm/xen/hypervisor.h>
53
#include <xen/page.h>
J
Jeremy Fitzhardinge 已提交
54

55 56
#include <xen/xen.h>
#include <xen/hvm.h>
57
#include <xen/xen-ops.h>
J
Jeremy Fitzhardinge 已提交
58 59 60
#include <xen/events.h>
#include <xen/interface/xen.h>
#include <xen/interface/event_channel.h>
61 62
#include <xen/interface/hvm/hvm_op.h>
#include <xen/interface/hvm/params.h>
63 64
#include <xen/interface/physdev.h>
#include <xen/interface/sched.h>
65
#include <xen/interface/vcpu.h>
66
#include <asm/hw_irq.h>
J
Jeremy Fitzhardinge 已提交
67

68 69
#include "events_internal.h"

70 71 72
#undef MODULE_PARAM_PREFIX
#define MODULE_PARAM_PREFIX "xen."

73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97
/* Interrupt types. */
enum xen_irq_type {
	IRQT_UNBOUND = 0,
	IRQT_PIRQ,
	IRQT_VIRQ,
	IRQT_IPI,
	IRQT_EVTCHN
};

/*
 * Packed IRQ information:
 * type - enum xen_irq_type
 * event channel - irq->event channel mapping
 * cpu - cpu this event channel is bound to
 * index - type-specific information:
 *    PIRQ - vector, with MSB being "needs EIO", or physical IRQ of the HVM
 *           guest, or GSI (real passthrough IRQ) of the device.
 *    VIRQ - virq number
 *    IPI - IPI vector
 *    EVTCHN -
 */
struct irq_info {
	struct list_head list;
	struct list_head eoi_list;
	short refcnt;
98
	u8 spurious_cnt;
99
	u8 is_accounted;
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
	enum xen_irq_type type; /* type */
	unsigned irq;
	evtchn_port_t evtchn;   /* event channel */
	unsigned short cpu;     /* cpu bound */
	unsigned short eoi_cpu; /* EOI must happen on this cpu-1 */
	unsigned int irq_epoch; /* If eoi_cpu valid: irq_epoch of event */
	u64 eoi_time;           /* Time in jiffies when to EOI. */

	union {
		unsigned short virq;
		enum ipi_vector ipi;
		struct {
			unsigned short pirq;
			unsigned short gsi;
			unsigned char vector;
			unsigned char flags;
			uint16_t domid;
		} pirq;
	} u;
};

#define PIRQ_NEEDS_EOI	(1 << 0)
#define PIRQ_SHAREABLE	(1 << 1)
#define PIRQ_MSI_GROUP	(1 << 2)

125 126 127 128 129 130
static uint __read_mostly event_loop_timeout = 2;
module_param(event_loop_timeout, uint, 0644);

static uint __read_mostly event_eoi_delay = 10;
module_param(event_eoi_delay, uint, 0644);

131 132
const struct evtchn_ops *evtchn_ops;

J
Jeremy Fitzhardinge 已提交
133 134 135 136
/*
 * 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.
 */
137
static DEFINE_MUTEX(irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
138

139 140 141 142 143 144 145 146 147 148 149 150 151 152 153
/*
 * Lock protecting event handling loop against removing event channels.
 * Adding of event channels is no issue as the associated IRQ becomes active
 * only after everything is setup (before request_[threaded_]irq() the handler
 * can't be entered for an event, as the event channel will be unmasked only
 * then).
 */
static DEFINE_RWLOCK(evtchn_rwlock);

/*
 * Lock hierarchy:
 *
 * irq_mapping_update_lock
 *   evtchn_rwlock
 *     IRQ-desc lock
154
 *       percpu eoi_list_lock
155 156
 */

157 158
static LIST_HEAD(xen_irq_list_head);

J
Jeremy Fitzhardinge 已提交
159
/* IRQ <-> VIRQ mapping. */
160
static DEFINE_PER_CPU(int [NR_VIRQS], virq_to_irq) = {[0 ... NR_VIRQS-1] = -1};
J
Jeremy Fitzhardinge 已提交
161

162
/* IRQ <-> IPI mapping */
163
static DEFINE_PER_CPU(int [XEN_NR_IPIS], ipi_to_irq) = {[0 ... XEN_NR_IPIS-1] = -1};
164

165 166 167
/* Event channel distribution data */
static atomic_t channels_on_cpu[NR_CPUS];

168
static int **evtchn_to_irq;
169
#ifdef CONFIG_X86
170
static unsigned long *pirq_eoi_map;
171
#endif
172
static bool (*pirq_needs_eoi)(unsigned irq);
173

174 175 176 177
#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
178 179 180
/* Xen will never allocate port zero for any purpose. */
#define VALID_EVTCHN(chn)	((chn) != 0)

181 182
static struct irq_info *legacy_info_ptrs[NR_IRQS_LEGACY];

J
Jeremy Fitzhardinge 已提交
183
static struct irq_chip xen_dynamic_chip;
184
static struct irq_chip xen_lateeoi_chip;
185
static struct irq_chip xen_percpu_chip;
186
static struct irq_chip xen_pirq_chip;
187 188
static void enable_dynirq(struct irq_data *data);
static void disable_dynirq(struct irq_data *data);
J
Jeremy Fitzhardinge 已提交
189

190 191
static DEFINE_PER_CPU(unsigned int, irq_epoch);

192 193 194 195 196
static void clear_evtchn_to_irq_row(unsigned row)
{
	unsigned col;

	for (col = 0; col < EVTCHN_PER_ROW; col++)
197
		WRITE_ONCE(evtchn_to_irq[row][col], -1);
198 199 200 201 202 203 204 205 206 207 208 209 210
}

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);
	}
}

211
static int set_evtchn_to_irq(evtchn_port_t evtchn, unsigned int irq)
212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233
{
	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);
	}

234
	WRITE_ONCE(evtchn_to_irq[row][col], irq);
235 236 237
	return 0;
}

238
int get_evtchn_to_irq(evtchn_port_t evtchn)
239 240 241 242 243
{
	if (evtchn >= xen_evtchn_max_channels())
		return -1;
	if (evtchn_to_irq[EVTCHN_ROW(evtchn)] == NULL)
		return -1;
244
	return READ_ONCE(evtchn_to_irq[EVTCHN_ROW(evtchn)][EVTCHN_COL(evtchn)]);
245 246
}

247
/* Get info for IRQ */
248
static struct irq_info *info_for_irq(unsigned irq)
249
{
250 251 252 253 254 255 256 257 258 259 260 261
	if (irq < nr_legacy_irqs())
		return legacy_info_ptrs[irq];
	else
		return irq_get_chip_data(irq);
}

static void set_info_for_irq(unsigned int irq, struct irq_info *info)
{
	if (irq < nr_legacy_irqs())
		legacy_info_ptrs[irq] = info;
	else
		irq_set_chip_data(irq, info);
262 263
}

264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289
/* Per CPU channel accounting */
static void channels_on_cpu_dec(struct irq_info *info)
{
	if (!info->is_accounted)
		return;

	info->is_accounted = 0;

	if (WARN_ON_ONCE(info->cpu >= nr_cpu_ids))
		return;

	WARN_ON_ONCE(!atomic_add_unless(&channels_on_cpu[info->cpu], -1 , 0));
}

static void channels_on_cpu_inc(struct irq_info *info)
{
	if (WARN_ON_ONCE(info->cpu >= nr_cpu_ids))
		return;

	if (WARN_ON_ONCE(!atomic_add_unless(&channels_on_cpu[info->cpu], 1,
					    INT_MAX)))
		return;

	info->is_accounted = 1;
}

290
/* Constructors for packed IRQ information. */
291
static int xen_irq_info_common_setup(struct irq_info *info,
292
				     unsigned irq,
293
				     enum xen_irq_type type,
294
				     evtchn_port_t evtchn,
295
				     unsigned short cpu)
296
{
297
	int ret;
298 299 300 301

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

	info->type = type;
302
	info->irq = irq;
303 304
	info->evtchn = evtchn;
	info->cpu = cpu;
305

306 307 308
	ret = set_evtchn_to_irq(evtchn, irq);
	if (ret < 0)
		return ret;
309 310

	irq_clear_status_flags(irq, IRQ_NOREQUEST|IRQ_NOAUTOEN);
311

312
	return xen_evtchn_port_setup(evtchn);
313 314
}

315
static int xen_irq_info_evtchn_setup(unsigned irq,
316
				     evtchn_port_t evtchn)
317
{
318 319
	struct irq_info *info = info_for_irq(irq);

320
	return xen_irq_info_common_setup(info, irq, IRQT_EVTCHN, evtchn, 0);
321 322
}

323
static int xen_irq_info_ipi_setup(unsigned cpu,
324
				  unsigned irq,
325
				  evtchn_port_t evtchn,
326
				  enum ipi_vector ipi)
J
Jeremy Fitzhardinge 已提交
327
{
328 329 330
	struct irq_info *info = info_for_irq(irq);

	info->u.ipi = ipi;
331 332

	per_cpu(ipi_to_irq, cpu)[ipi] = irq;
333 334

	return xen_irq_info_common_setup(info, irq, IRQT_IPI, evtchn, 0);
335 336
}

337
static int xen_irq_info_virq_setup(unsigned cpu,
338
				   unsigned irq,
339
				   evtchn_port_t evtchn,
340
				   unsigned virq)
341
{
342 343 344
	struct irq_info *info = info_for_irq(irq);

	info->u.virq = virq;
345 346

	per_cpu(virq_to_irq, cpu)[virq] = irq;
347 348

	return xen_irq_info_common_setup(info, irq, IRQT_VIRQ, evtchn, 0);
349 350
}

351
static int xen_irq_info_pirq_setup(unsigned irq,
352
				   evtchn_port_t evtchn,
353 354
				   unsigned pirq,
				   unsigned gsi,
355
				   uint16_t domid,
356
				   unsigned char flags)
357
{
358 359 360 361
	struct irq_info *info = info_for_irq(irq);

	info->u.pirq.pirq = pirq;
	info->u.pirq.gsi = gsi;
362
	info->u.pirq.domid = domid;
363
	info->u.pirq.flags = flags;
364 365

	return xen_irq_info_common_setup(info, irq, IRQT_PIRQ, evtchn, 0);
J
Jeremy Fitzhardinge 已提交
366 367
}

368 369 370 371
static void xen_irq_info_cleanup(struct irq_info *info)
{
	set_evtchn_to_irq(info->evtchn, -1);
	info->evtchn = 0;
372
	channels_on_cpu_dec(info);
373 374
}

J
Jeremy Fitzhardinge 已提交
375 376 377
/*
 * Accessors for packed IRQ information.
 */
378
evtchn_port_t evtchn_from_irq(unsigned irq)
J
Jeremy Fitzhardinge 已提交
379
{
380 381 382 383 384
	const struct irq_info *info = NULL;

	if (likely(irq < nr_irqs))
		info = info_for_irq(irq);
	if (!info)
385 386
		return 0;

387
	return info->evtchn;
J
Jeremy Fitzhardinge 已提交
388 389
}

390
unsigned int irq_from_evtchn(evtchn_port_t evtchn)
I
Ian Campbell 已提交
391
{
392
	return get_evtchn_to_irq(evtchn);
I
Ian Campbell 已提交
393 394 395
}
EXPORT_SYMBOL_GPL(irq_from_evtchn);

396 397 398 399 400
int irq_from_virq(unsigned int cpu, unsigned int virq)
{
	return per_cpu(virq_to_irq, cpu)[virq];
}

401
static enum ipi_vector ipi_from_irq(unsigned irq)
J
Jeremy Fitzhardinge 已提交
402
{
403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420
	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;
}

421 422 423 424 425 426 427 428 429 430
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;
}

431 432 433 434 435
static enum xen_irq_type type_from_irq(unsigned irq)
{
	return info_for_irq(irq)->type;
}

436
static unsigned cpu_from_irq(unsigned irq)
437 438 439 440
{
	return info_for_irq(irq)->cpu;
}

441
unsigned int cpu_from_evtchn(evtchn_port_t evtchn)
442
{
443
	int irq = get_evtchn_to_irq(evtchn);
444 445 446 447 448 449
	unsigned ret = 0;

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

	return ret;
J
Jeremy Fitzhardinge 已提交
450 451
}

452
#ifdef CONFIG_X86
453
static bool pirq_check_eoi_map(unsigned irq)
454
{
455
	return test_bit(pirq_from_irq(irq), pirq_eoi_map);
456
}
457
#endif
458

459 460 461
static bool pirq_needs_eoi_flag(unsigned irq)
{
	struct irq_info *info = info_for_irq(irq);
462 463 464 465 466
	BUG_ON(info->type != IRQT_PIRQ);

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

467 468
static void bind_evtchn_to_cpu(evtchn_port_t evtchn, unsigned int cpu,
			       bool force_affinity)
J
Jeremy Fitzhardinge 已提交
469
{
470
	int irq = get_evtchn_to_irq(evtchn);
471
	struct irq_info *info = info_for_irq(irq);
J
Jeremy Fitzhardinge 已提交
472 473

	BUG_ON(irq == -1);
474 475 476 477 478 479 480

	if (IS_ENABLED(CONFIG_SMP) && force_affinity) {
		cpumask_copy(irq_get_affinity_mask(irq), cpumask_of(cpu));
		cpumask_copy(irq_get_effective_affinity_mask(irq),
			     cpumask_of(cpu));
	}

481
	xen_evtchn_port_bind_to_cpu(evtchn, cpu, info->cpu);
482

483
	channels_on_cpu_dec(info);
484
	info->cpu = cpu;
485
	channels_on_cpu_inc(info);
486 487
}

J
Jeremy Fitzhardinge 已提交
488 489 490 491 492 493 494 495 496 497
/**
 * 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)
{
498
	evtchn_port_t evtchn = evtchn_from_irq(irq);
J
Jeremy Fitzhardinge 已提交
499 500 501 502 503 504

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

505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552
struct lateeoi_work {
	struct delayed_work delayed;
	spinlock_t eoi_list_lock;
	struct list_head eoi_list;
};

static DEFINE_PER_CPU(struct lateeoi_work, lateeoi);

static void lateeoi_list_del(struct irq_info *info)
{
	struct lateeoi_work *eoi = &per_cpu(lateeoi, info->eoi_cpu);
	unsigned long flags;

	spin_lock_irqsave(&eoi->eoi_list_lock, flags);
	list_del_init(&info->eoi_list);
	spin_unlock_irqrestore(&eoi->eoi_list_lock, flags);
}

static void lateeoi_list_add(struct irq_info *info)
{
	struct lateeoi_work *eoi = &per_cpu(lateeoi, info->eoi_cpu);
	struct irq_info *elem;
	u64 now = get_jiffies_64();
	unsigned long delay;
	unsigned long flags;

	if (now < info->eoi_time)
		delay = info->eoi_time - now;
	else
		delay = 1;

	spin_lock_irqsave(&eoi->eoi_list_lock, flags);

	if (list_empty(&eoi->eoi_list)) {
		list_add(&info->eoi_list, &eoi->eoi_list);
		mod_delayed_work_on(info->eoi_cpu, system_wq,
				    &eoi->delayed, delay);
	} else {
		list_for_each_entry_reverse(elem, &eoi->eoi_list, eoi_list) {
			if (elem->eoi_time <= info->eoi_time)
				break;
		}
		list_add(&info->eoi_list, &elem->eoi_list);
	}

	spin_unlock_irqrestore(&eoi->eoi_list_lock, flags);
}

553
static void xen_irq_lateeoi_locked(struct irq_info *info, bool spurious)
554 555
{
	evtchn_port_t evtchn;
556
	unsigned int cpu;
557
	unsigned int delay = 0;
558 559

	evtchn = info->evtchn;
560
	if (!VALID_EVTCHN(evtchn) || !list_empty(&info->eoi_list))
561 562
		return;

563
	if (spurious) {
564 565 566 567
		if ((1 << info->spurious_cnt) < (HZ << 2)) {
			if (info->spurious_cnt != 0xFF)
				info->spurious_cnt++;
		}
568 569 570 571 572 573 574 575 576 577 578 579
		if (info->spurious_cnt > 1) {
			delay = 1 << (info->spurious_cnt - 2);
			if (delay > HZ)
				delay = HZ;
			if (!info->eoi_time)
				info->eoi_cpu = smp_processor_id();
			info->eoi_time = get_jiffies_64() + delay;
		}
	} else {
		info->spurious_cnt = 0;
	}

580
	cpu = info->eoi_cpu;
581 582
	if (info->eoi_time &&
	    (info->irq_epoch == per_cpu(irq_epoch, cpu) || delay)) {
583 584 585 586 587
		lateeoi_list_add(info);
		return;
	}

	info->eoi_time = 0;
588 589 590
	unmask_evtchn(evtchn);
}

591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618
static void xen_irq_lateeoi_worker(struct work_struct *work)
{
	struct lateeoi_work *eoi;
	struct irq_info *info;
	u64 now = get_jiffies_64();
	unsigned long flags;

	eoi = container_of(to_delayed_work(work), struct lateeoi_work, delayed);

	read_lock_irqsave(&evtchn_rwlock, flags);

	while (true) {
		spin_lock(&eoi->eoi_list_lock);

		info = list_first_entry_or_null(&eoi->eoi_list, struct irq_info,
						eoi_list);

		if (info == NULL || now < info->eoi_time) {
			spin_unlock(&eoi->eoi_list_lock);
			break;
		}

		list_del_init(&info->eoi_list);

		spin_unlock(&eoi->eoi_list_lock);

		info->eoi_time = 0;

619
		xen_irq_lateeoi_locked(info, false);
620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637
	}

	if (info)
		mod_delayed_work_on(info->eoi_cpu, system_wq,
				    &eoi->delayed, info->eoi_time - now);

	read_unlock_irqrestore(&evtchn_rwlock, flags);
}

static void xen_cpu_init_eoi(unsigned int cpu)
{
	struct lateeoi_work *eoi = &per_cpu(lateeoi, cpu);

	INIT_DELAYED_WORK(&eoi->delayed, xen_irq_lateeoi_worker);
	spin_lock_init(&eoi->eoi_list_lock);
	INIT_LIST_HEAD(&eoi->eoi_list);
}

638 639 640 641 642 643 644 645 646 647
void xen_irq_lateeoi(unsigned int irq, unsigned int eoi_flags)
{
	struct irq_info *info;
	unsigned long flags;

	read_lock_irqsave(&evtchn_rwlock, flags);

	info = info_for_irq(irq);

	if (info)
648
		xen_irq_lateeoi_locked(info, eoi_flags & XEN_EOI_FLAG_SPURIOUS);
649 650 651 652 653

	read_unlock_irqrestore(&evtchn_rwlock, flags);
}
EXPORT_SYMBOL_GPL(xen_irq_lateeoi);

654 655 656
static void xen_irq_init(unsigned irq)
{
	struct irq_info *info;
657

658 659 660
	info = kzalloc(sizeof(*info), GFP_KERNEL);
	if (info == NULL)
		panic("Unable to allocate metadata for IRQ%d\n", irq);
661 662

	info->type = IRQT_UNBOUND;
663
	info->refcnt = -1;
664

665
	set_info_for_irq(irq, info);
666 667 668 669 670
	/*
	 * Interrupt affinity setting can be immediate. No point
	 * in delaying it until an interrupt is handled.
	 */
	irq_set_status_flags(irq, IRQ_MOVE_PCNTXT);
671

672
	INIT_LIST_HEAD(&info->eoi_list);
673 674 675
	list_add_tail(&info->list, &xen_irq_list_head);
}

676
static int __must_check xen_allocate_irqs_dynamic(int nvec)
677
{
678
	int i, irq = irq_alloc_descs(-1, 0, nvec, -1);
679

680 681 682 683
	if (irq >= 0) {
		for (i = 0; i < nvec; i++)
			xen_irq_init(irq + i);
	}
684

J
Jeremy Fitzhardinge 已提交
685
	return irq;
686 687
}

688 689 690 691 692 693
static inline int __must_check xen_allocate_irq_dynamic(void)
{

	return xen_allocate_irqs_dynamic(1);
}

694
static int __must_check xen_allocate_irq_gsi(unsigned gsi)
695 696 697
{
	int irq;

698 699 700 701 702 703 704
	/*
	 * 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())
705 706 707
		return xen_allocate_irq_dynamic();

	/* Legacy IRQ descriptors are already allocated by the arch. */
708
	if (gsi < nr_legacy_irqs())
709 710 711
		irq = gsi;
	else
		irq = irq_alloc_desc_at(gsi, -1);
712

713
	xen_irq_init(irq);
714 715 716 717 718 719

	return irq;
}

static void xen_free_irq(unsigned irq)
{
720
	struct irq_info *info = info_for_irq(irq);
721
	unsigned long flags;
722

723 724 725
	if (WARN_ON(!info))
		return;

726 727
	write_lock_irqsave(&evtchn_rwlock, flags);

728 729 730
	if (!list_empty(&info->eoi_list))
		lateeoi_list_del(info);

731
	list_del(&info->list);
732

733
	set_info_for_irq(irq, NULL);
734

735 736
	WARN_ON(info->refcnt > 0);

737 738
	write_unlock_irqrestore(&evtchn_rwlock, flags);

739 740
	kfree(info);

741
	/* Legacy IRQ descriptors are managed by the arch. */
742
	if (irq < nr_legacy_irqs())
743 744
		return;

745 746 747
	irq_free_desc(irq);
}

748
static void xen_evtchn_close(evtchn_port_t port)
749 750 751 752 753 754 755 756
{
	struct evtchn_close close;

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

757 758 759 760 761 762 763
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);

764
	irq_status.irq = pirq_from_irq(irq);
765 766 767 768 769 770 771 772
	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;
}

773 774
static void eoi_pirq(struct irq_data *data)
{
775
	evtchn_port_t evtchn = evtchn_from_irq(data->irq);
776 777 778
	struct physdev_eoi eoi = { .irq = pirq_from_irq(data->irq) };
	int rc = 0;

779 780
	if (!VALID_EVTCHN(evtchn))
		return;
781

782
	clear_evtchn(evtchn);
783 784 785 786 787 788 789 790 791 792 793 794 795

	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);
}

796
static unsigned int __startup_pirq(unsigned int irq)
797 798 799
{
	struct evtchn_bind_pirq bind_pirq;
	struct irq_info *info = info_for_irq(irq);
800
	evtchn_port_t evtchn = evtchn_from_irq(irq);
801
	int rc;
802 803 804 805 806 807

	BUG_ON(info->type != IRQT_PIRQ);

	if (VALID_EVTCHN(evtchn))
		goto out;

808
	bind_pirq.pirq = pirq_from_irq(irq);
809
	/* NB. We are happy to share unless we are probing. */
810 811 812 813
	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) {
814
		pr_warn("Failed to obtain physical IRQ %d\n", irq);
815 816 817 818 819 820
		return 0;
	}
	evtchn = bind_pirq.port;

	pirq_query_unmask(irq);

821
	rc = set_evtchn_to_irq(evtchn, irq);
822 823 824
	if (rc)
		goto err;

825
	info->evtchn = evtchn;
826
	bind_evtchn_to_cpu(evtchn, 0, false);
827

828
	rc = xen_evtchn_port_setup(evtchn);
829 830 831
	if (rc)
		goto err;

832 833
out:
	unmask_evtchn(evtchn);
834
	eoi_pirq(irq_get_irq_data(irq));
835 836

	return 0;
837 838 839 840 841

err:
	pr_err("irq%d: Failed to set port to irq mapping (%d)\n", irq, rc);
	xen_evtchn_close(evtchn);
	return 0;
842 843
}

844 845 846 847 848 849
static unsigned int startup_pirq(struct irq_data *data)
{
	return __startup_pirq(data->irq);
}

static void shutdown_pirq(struct irq_data *data)
850
{
851
	unsigned int irq = data->irq;
852
	struct irq_info *info = info_for_irq(irq);
853
	evtchn_port_t evtchn = evtchn_from_irq(irq);
854 855 856 857 858 859 860

	BUG_ON(info->type != IRQT_PIRQ);

	if (!VALID_EVTCHN(evtchn))
		return;

	mask_evtchn(evtchn);
861 862
	xen_evtchn_close(evtchn);
	xen_irq_info_cleanup(info);
863 864
}

865
static void enable_pirq(struct irq_data *data)
866
{
867
	enable_dynirq(data);
868 869
}

870
static void disable_pirq(struct irq_data *data)
871
{
872
	disable_dynirq(data);
873 874
}

875
int xen_irq_from_gsi(unsigned gsi)
876
{
877
	struct irq_info *info;
878

879 880
	list_for_each_entry(info, &xen_irq_list_head, list) {
		if (info->type != IRQT_PIRQ)
881 882
			continue;

883 884
		if (info->u.pirq.gsi == gsi)
			return info->irq;
885 886 887 888
	}

	return -1;
}
889
EXPORT_SYMBOL_GPL(xen_irq_from_gsi);
890

891 892
static void __unbind_from_irq(unsigned int irq)
{
893
	evtchn_port_t evtchn = evtchn_from_irq(irq);
894
	struct irq_info *info = info_for_irq(irq);
895 896 897 898 899 900 901 902

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

	if (VALID_EVTCHN(evtchn)) {
903 904 905
		unsigned int cpu = cpu_from_irq(irq);

		xen_evtchn_close(evtchn);
906 907 908

		switch (type_from_irq(irq)) {
		case IRQT_VIRQ:
909
			per_cpu(virq_to_irq, cpu)[virq_from_irq(irq)] = -1;
910 911
			break;
		case IRQT_IPI:
912
			per_cpu(ipi_to_irq, cpu)[ipi_from_irq(irq)] = -1;
913 914 915 916 917
			break;
		default:
			break;
		}

918
		xen_irq_info_cleanup(info);
919 920 921 922 923
	}

	xen_free_irq(irq);
}

924 925 926
/*
 * Do not make any assumptions regarding the relationship between the
 * IRQ number returned here and the Xen pirq argument.
927 928 929
 *
 * 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.
930 931 932
 *
 * Shareable implies level triggered, not shareable implies edge
 * triggered here.
933
 */
934 935
int xen_bind_pirq_gsi_to_irq(unsigned gsi,
			     unsigned pirq, int shareable, char *name)
936
{
937
	int irq = -1;
938
	struct physdev_irq irq_op;
939
	int ret;
940

941
	mutex_lock(&irq_mapping_update_lock);
942

943
	irq = xen_irq_from_gsi(gsi);
944
	if (irq != -1) {
945 946
		pr_info("%s: returning irq %d for gsi %u\n",
			__func__, irq, gsi);
947
		goto out;
948 949
	}

950
	irq = xen_allocate_irq_gsi(gsi);
951 952
	if (irq < 0)
		goto out;
953 954

	irq_op.irq = irq;
955 956 957 958 959 960 961
	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)) {
962
		xen_free_irq(irq);
963 964 965 966
		irq = -ENOSPC;
		goto out;
	}

967
	ret = xen_irq_info_pirq_setup(irq, 0, pirq, gsi, DOMID_SELF,
968
			       shareable ? PIRQ_SHAREABLE : 0);
969 970 971 972 973
	if (ret < 0) {
		__unbind_from_irq(irq);
		irq = ret;
		goto out;
	}
974

975 976
	pirq_query_unmask(irq);
	/* We try to use the handler with the appropriate semantic for the
977 978
	 * type of interrupt: if the interrupt is an edge triggered
	 * interrupt we use handle_edge_irq.
979
	 *
980 981
	 * On the other hand if the interrupt is level triggered we use
	 * handle_fasteoi_irq like the native code does for this kind of
982
	 * interrupts.
983
	 *
984 985 986 987 988 989 990
	 * 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.
	 */
991
	if (shareable)
992 993 994 995 996 997
		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);

998
out:
999
	mutex_unlock(&irq_mapping_update_lock);
1000 1001 1002 1003

	return irq;
}

1004
#ifdef CONFIG_PCI_MSI
1005
int xen_allocate_pirq_msi(struct pci_dev *dev, struct msi_desc *msidesc)
1006
{
1007
	int rc;
1008 1009
	struct physdev_get_free_pirq op_get_free_pirq;

1010
	op_get_free_pirq.type = MAP_PIRQ_TYPE_MSI;
1011 1012
	rc = HYPERVISOR_physdev_op(PHYSDEVOP_get_free_pirq, &op_get_free_pirq);

1013 1014 1015 1016
	WARN_ONCE(rc == -ENOSYS,
		  "hypervisor does not support the PHYSDEVOP_get_free_pirq interface\n");

	return rc ? -1 : op_get_free_pirq.pirq;
1017 1018
}

1019
int xen_bind_pirq_msi_to_irq(struct pci_dev *dev, struct msi_desc *msidesc,
1020
			     int pirq, int nvec, const char *name, domid_t domid)
1021
{
1022
	int i, irq, ret;
1023

1024
	mutex_lock(&irq_mapping_update_lock);
1025

1026
	irq = xen_allocate_irqs_dynamic(nvec);
1027
	if (irq < 0)
1028
		goto out;
1029

1030 1031 1032 1033 1034 1035 1036 1037
	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;
	}
1038

1039
	ret = irq_set_msi_desc(irq, msidesc);
1040 1041
	if (ret < 0)
		goto error_irq;
1042
out:
1043
	mutex_unlock(&irq_mapping_update_lock);
1044
	return irq;
1045
error_irq:
1046 1047
	while (nvec--)
		__unbind_from_irq(irq + nvec);
1048
	mutex_unlock(&irq_mapping_update_lock);
1049
	return ret;
1050
}
1051 1052
#endif

1053 1054
int xen_destroy_irq(int irq)
{
1055 1056
	struct physdev_unmap_pirq unmap_irq;
	struct irq_info *info = info_for_irq(irq);
1057 1058
	int rc = -ENOENT;

1059
	mutex_lock(&irq_mapping_update_lock);
1060

1061 1062 1063 1064 1065 1066
	/*
	 * 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)) {
1067
		unmap_irq.pirq = info->u.pirq.pirq;
1068
		unmap_irq.domid = info->u.pirq.domid;
1069
		rc = HYPERVISOR_physdev_op(PHYSDEVOP_unmap_pirq, &unmap_irq);
1070 1071 1072 1073 1074
		/* 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))
1075
			pr_info("domain %d does not have %d anymore\n",
1076 1077
				info->u.pirq.domid, info->u.pirq.pirq);
		else if (rc) {
1078
			pr_warn("unmap irq failed %d\n", rc);
1079 1080 1081
			goto out;
		}
	}
1082

1083
	xen_free_irq(irq);
1084 1085

out:
1086
	mutex_unlock(&irq_mapping_update_lock);
1087 1088 1089
	return rc;
}

1090
int xen_irq_from_pirq(unsigned pirq)
1091
{
1092
	int irq;
1093

1094
	struct irq_info *info;
J
Jeremy Fitzhardinge 已提交
1095

1096
	mutex_lock(&irq_mapping_update_lock);
1097 1098

	list_for_each_entry(info, &xen_irq_list_head, list) {
1099
		if (info->type != IRQT_PIRQ)
1100 1101 1102 1103 1104 1105 1106
			continue;
		irq = info->irq;
		if (info->u.pirq.pirq == pirq)
			goto out;
	}
	irq = -1;
out:
1107
	mutex_unlock(&irq_mapping_update_lock);
1108 1109

	return irq;
1110 1111
}

1112 1113 1114 1115 1116 1117

int xen_pirq_from_irq(unsigned irq)
{
	return pirq_from_irq(irq);
}
EXPORT_SYMBOL_GPL(xen_pirq_from_irq);
1118

1119
static int bind_evtchn_to_irq_chip(evtchn_port_t evtchn, struct irq_chip *chip)
J
Jeremy Fitzhardinge 已提交
1120 1121
{
	int irq;
1122
	int ret;
J
Jeremy Fitzhardinge 已提交
1123

1124 1125 1126
	if (evtchn >= xen_evtchn_max_channels())
		return -ENOMEM;

1127
	mutex_lock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
1128

1129
	irq = get_evtchn_to_irq(evtchn);
J
Jeremy Fitzhardinge 已提交
1130 1131

	if (irq == -1) {
1132
		irq = xen_allocate_irq_dynamic();
1133
		if (irq < 0)
1134
			goto out;
J
Jeremy Fitzhardinge 已提交
1135

1136
		irq_set_chip_and_handler_name(irq, chip,
1137
					      handle_edge_irq, "event");
J
Jeremy Fitzhardinge 已提交
1138

1139 1140 1141 1142 1143 1144
		ret = xen_irq_info_evtchn_setup(irq, evtchn);
		if (ret < 0) {
			__unbind_from_irq(irq);
			irq = ret;
			goto out;
		}
1145 1146 1147 1148 1149 1150 1151 1152
		/*
		 * New interdomain events are initially bound to vCPU0 This
		 * is required to setup the event channel in the first
		 * place and also important for UP guests because the
		 * affinity setting is not invoked on them so nothing would
		 * bind the channel.
		 */
		bind_evtchn_to_cpu(evtchn, 0, false);
1153 1154 1155
	} else {
		struct irq_info *info = info_for_irq(irq);
		WARN_ON(info == NULL || info->type != IRQT_EVTCHN);
J
Jeremy Fitzhardinge 已提交
1156 1157
	}

1158
out:
1159
	mutex_unlock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
1160 1161 1162

	return irq;
}
1163 1164 1165 1166 1167

int bind_evtchn_to_irq(evtchn_port_t evtchn)
{
	return bind_evtchn_to_irq_chip(evtchn, &xen_dynamic_chip);
}
1168
EXPORT_SYMBOL_GPL(bind_evtchn_to_irq);
J
Jeremy Fitzhardinge 已提交
1169

1170 1171 1172
static int bind_ipi_to_irq(unsigned int ipi, unsigned int cpu)
{
	struct evtchn_bind_ipi bind_ipi;
1173 1174
	evtchn_port_t evtchn;
	int ret, irq;
1175

1176
	mutex_lock(&irq_mapping_update_lock);
1177 1178

	irq = per_cpu(ipi_to_irq, cpu)[ipi];
1179

1180
	if (irq == -1) {
1181
		irq = xen_allocate_irq_dynamic();
1182 1183 1184
		if (irq < 0)
			goto out;

1185
		irq_set_chip_and_handler_name(irq, &xen_percpu_chip,
1186
					      handle_percpu_irq, "ipi");
1187

1188
		bind_ipi.vcpu = xen_vcpu_nr(cpu);
1189 1190 1191 1192 1193
		if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_ipi,
						&bind_ipi) != 0)
			BUG();
		evtchn = bind_ipi.port;

1194 1195 1196 1197 1198 1199
		ret = xen_irq_info_ipi_setup(cpu, irq, evtchn, ipi);
		if (ret < 0) {
			__unbind_from_irq(irq);
			irq = ret;
			goto out;
		}
1200 1201 1202 1203 1204
		/*
		 * Force the affinity mask to the target CPU so proc shows
		 * the correct target.
		 */
		bind_evtchn_to_cpu(evtchn, cpu, true);
1205 1206 1207
	} else {
		struct irq_info *info = info_for_irq(irq);
		WARN_ON(info == NULL || info->type != IRQT_IPI);
1208 1209 1210
	}

 out:
1211
	mutex_unlock(&irq_mapping_update_lock);
1212 1213 1214
	return irq;
}

1215 1216 1217
static int bind_interdomain_evtchn_to_irq_chip(unsigned int remote_domain,
					       evtchn_port_t remote_port,
					       struct irq_chip *chip)
1218 1219 1220 1221 1222 1223 1224 1225 1226 1227
{
	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);

1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239
	return err ? : bind_evtchn_to_irq_chip(bind_interdomain.local_port,
					       chip);
}

int bind_interdomain_evtchn_to_irq_lateeoi(unsigned int remote_domain,
					   evtchn_port_t remote_port)
{
	return bind_interdomain_evtchn_to_irq_chip(remote_domain, remote_port,
						   &xen_lateeoi_chip);
}
EXPORT_SYMBOL_GPL(bind_interdomain_evtchn_to_irq_lateeoi);

1240
static int find_virq(unsigned int virq, unsigned int cpu, evtchn_port_t *evtchn)
1241 1242
{
	struct evtchn_status status;
1243 1244
	evtchn_port_t port;
	int rc = -ENOENT;
1245 1246

	memset(&status, 0, sizeof(status));
1247
	for (port = 0; port < xen_evtchn_max_channels(); port++) {
1248 1249 1250 1251 1252 1253 1254
		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;
1255
		if (status.u.virq == virq && status.vcpu == xen_vcpu_nr(cpu)) {
1256
			*evtchn = port;
1257 1258 1259 1260 1261
			break;
		}
	}
	return rc;
}
1262

1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275
/**
 * 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);

1276
int bind_virq_to_irq(unsigned int virq, unsigned int cpu, bool percpu)
J
Jeremy Fitzhardinge 已提交
1277 1278
{
	struct evtchn_bind_virq bind_virq;
1279 1280
	evtchn_port_t evtchn = 0;
	int irq, ret;
J
Jeremy Fitzhardinge 已提交
1281

1282
	mutex_lock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
1283 1284 1285 1286

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

	if (irq == -1) {
1287
		irq = xen_allocate_irq_dynamic();
1288
		if (irq < 0)
1289
			goto out;
1290

1291 1292 1293 1294 1295 1296
		if (percpu)
			irq_set_chip_and_handler_name(irq, &xen_percpu_chip,
						      handle_percpu_irq, "virq");
		else
			irq_set_chip_and_handler_name(irq, &xen_dynamic_chip,
						      handle_edge_irq, "virq");
1297

J
Jeremy Fitzhardinge 已提交
1298
		bind_virq.virq = virq;
1299
		bind_virq.vcpu = xen_vcpu_nr(cpu);
1300 1301 1302 1303 1304 1305
		ret = HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq,
						&bind_virq);
		if (ret == 0)
			evtchn = bind_virq.port;
		else {
			if (ret == -EEXIST)
1306
				ret = find_virq(virq, cpu, &evtchn);
1307 1308
			BUG_ON(ret < 0);
		}
J
Jeremy Fitzhardinge 已提交
1309

1310 1311 1312 1313 1314 1315
		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
1316

1317 1318 1319 1320 1321
		/*
		 * Force the affinity mask for percpu interrupts so proc
		 * shows the correct target.
		 */
		bind_evtchn_to_cpu(evtchn, cpu, percpu);
1322 1323 1324
	} else {
		struct irq_info *info = info_for_irq(irq);
		WARN_ON(info == NULL || info->type != IRQT_VIRQ);
J
Jeremy Fitzhardinge 已提交
1325 1326
	}

1327
out:
1328
	mutex_unlock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
1329 1330 1331 1332 1333 1334

	return irq;
}

static void unbind_from_irq(unsigned int irq)
{
1335
	mutex_lock(&irq_mapping_update_lock);
1336
	__unbind_from_irq(irq);
1337
	mutex_unlock(&irq_mapping_update_lock);
J
Jeremy Fitzhardinge 已提交
1338 1339
}

1340 1341 1342 1343 1344
static int bind_evtchn_to_irqhandler_chip(evtchn_port_t evtchn,
					  irq_handler_t handler,
					  unsigned long irqflags,
					  const char *devname, void *dev_id,
					  struct irq_chip *chip)
J
Jeremy Fitzhardinge 已提交
1345
{
1346
	int irq, retval;
J
Jeremy Fitzhardinge 已提交
1347

1348
	irq = bind_evtchn_to_irq_chip(evtchn, chip);
1349 1350
	if (irq < 0)
		return irq;
J
Jeremy Fitzhardinge 已提交
1351 1352 1353 1354 1355 1356 1357 1358
	retval = request_irq(irq, handler, irqflags, devname, dev_id);
	if (retval != 0) {
		unbind_from_irq(irq);
		return retval;
	}

	return irq;
}
1359 1360 1361 1362 1363 1364 1365 1366 1367 1368

int bind_evtchn_to_irqhandler(evtchn_port_t evtchn,
			      irq_handler_t handler,
			      unsigned long irqflags,
			      const char *devname, void *dev_id)
{
	return bind_evtchn_to_irqhandler_chip(evtchn, handler, irqflags,
					      devname, dev_id,
					      &xen_dynamic_chip);
}
J
Jeremy Fitzhardinge 已提交
1369 1370
EXPORT_SYMBOL_GPL(bind_evtchn_to_irqhandler);

1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385
int bind_evtchn_to_irqhandler_lateeoi(evtchn_port_t evtchn,
				      irq_handler_t handler,
				      unsigned long irqflags,
				      const char *devname, void *dev_id)
{
	return bind_evtchn_to_irqhandler_chip(evtchn, handler, irqflags,
					      devname, dev_id,
					      &xen_lateeoi_chip);
}
EXPORT_SYMBOL_GPL(bind_evtchn_to_irqhandler_lateeoi);

static int bind_interdomain_evtchn_to_irqhandler_chip(
		unsigned int remote_domain, evtchn_port_t remote_port,
		irq_handler_t handler, unsigned long irqflags,
		const char *devname, void *dev_id, struct irq_chip *chip)
1386 1387 1388
{
	int irq, retval;

1389 1390
	irq = bind_interdomain_evtchn_to_irq_chip(remote_domain, remote_port,
						  chip);
1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401
	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;
}
1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415

int bind_interdomain_evtchn_to_irqhandler_lateeoi(unsigned int remote_domain,
						  evtchn_port_t remote_port,
						  irq_handler_t handler,
						  unsigned long irqflags,
						  const char *devname,
						  void *dev_id)
{
	return bind_interdomain_evtchn_to_irqhandler_chip(remote_domain,
				remote_port, handler, irqflags, devname,
				dev_id, &xen_lateeoi_chip);
}
EXPORT_SYMBOL_GPL(bind_interdomain_evtchn_to_irqhandler_lateeoi);

J
Jeremy Fitzhardinge 已提交
1416
int bind_virq_to_irqhandler(unsigned int virq, unsigned int cpu,
1417
			    irq_handler_t handler,
J
Jeremy Fitzhardinge 已提交
1418 1419
			    unsigned long irqflags, const char *devname, void *dev_id)
{
1420
	int irq, retval;
J
Jeremy Fitzhardinge 已提交
1421

1422
	irq = bind_virq_to_irq(virq, cpu, irqflags & IRQF_PERCPU);
1423 1424
	if (irq < 0)
		return irq;
J
Jeremy Fitzhardinge 已提交
1425 1426 1427 1428 1429 1430 1431 1432 1433 1434
	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);

1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447
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;

1448
	irqflags |= IRQF_NO_SUSPEND | IRQF_FORCE_RESUME | IRQF_EARLY_RESUME;
1449 1450 1451 1452 1453 1454 1455 1456 1457
	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
1458 1459
void unbind_from_irqhandler(unsigned int irq, void *dev_id)
{
1460
	struct irq_info *info = info_for_irq(irq);
1461 1462 1463

	if (WARN_ON(!info))
		return;
J
Jeremy Fitzhardinge 已提交
1464 1465 1466 1467 1468
	free_irq(irq, dev_id);
	unbind_from_irq(irq);
}
EXPORT_SYMBOL_GPL(unbind_from_irqhandler);

1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485
/**
 * 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);

1486
int evtchn_make_refcounted(evtchn_port_t evtchn)
1487
{
1488
	int irq = get_evtchn_to_irq(evtchn);
1489 1490 1491 1492 1493
	struct irq_info *info;

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

1494
	info = info_for_irq(irq);
1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506

	if (!info)
		return -ENOENT;

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

	info->refcnt = 1;

	return 0;
}
EXPORT_SYMBOL_GPL(evtchn_make_refcounted);

1507
int evtchn_get(evtchn_port_t evtchn)
1508 1509 1510 1511 1512
{
	int irq;
	struct irq_info *info;
	int err = -ENOENT;

1513
	if (evtchn >= xen_evtchn_max_channels())
1514 1515
		return -EINVAL;

1516 1517
	mutex_lock(&irq_mapping_update_lock);

1518
	irq = get_evtchn_to_irq(evtchn);
1519 1520 1521
	if (irq == -1)
		goto done;

1522
	info = info_for_irq(irq);
1523 1524 1525 1526 1527

	if (!info)
		goto done;

	err = -EINVAL;
1528
	if (info->refcnt <= 0 || info->refcnt == SHRT_MAX)
1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539
		goto done;

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

	return err;
}
EXPORT_SYMBOL_GPL(evtchn_get);

1540
void evtchn_put(evtchn_port_t evtchn)
1541
{
1542
	int irq = get_evtchn_to_irq(evtchn);
1543 1544 1545 1546 1547 1548
	if (WARN_ON(irq == -1))
		return;
	unbind_from_irq(irq);
}
EXPORT_SYMBOL_GPL(evtchn_put);

1549 1550
void xen_send_IPI_one(unsigned int cpu, enum ipi_vector vector)
{
1551 1552
	int irq;

1553
#ifdef CONFIG_X86
1554
	if (unlikely(vector == XEN_NMI_VECTOR)) {
1555 1556
		int rc =  HYPERVISOR_vcpu_op(VCPUOP_send_nmi, xen_vcpu_nr(cpu),
					     NULL);
1557 1558 1559 1560
		if (rc < 0)
			printk(KERN_WARNING "Sending nmi to CPU%d failed (rc:%d)\n", cpu, rc);
		return;
	}
1561
#endif
1562
	irq = per_cpu(ipi_to_irq, cpu)[vector];
1563 1564 1565 1566
	BUG_ON(irq < 0);
	notify_remote_via_irq(irq);
}

1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614
struct evtchn_loop_ctrl {
	ktime_t timeout;
	unsigned count;
	bool defer_eoi;
};

void handle_irq_for_port(evtchn_port_t port, struct evtchn_loop_ctrl *ctrl)
{
	int irq;
	struct irq_info *info;

	irq = get_evtchn_to_irq(port);
	if (irq == -1)
		return;

	/*
	 * Check for timeout every 256 events.
	 * We are setting the timeout value only after the first 256
	 * events in order to not hurt the common case of few loop
	 * iterations. The 256 is basically an arbitrary value.
	 *
	 * In case we are hitting the timeout we need to defer all further
	 * EOIs in order to ensure to leave the event handling loop rather
	 * sooner than later.
	 */
	if (!ctrl->defer_eoi && !(++ctrl->count & 0xff)) {
		ktime_t kt = ktime_get();

		if (!ctrl->timeout) {
			kt = ktime_add_ms(kt,
					  jiffies_to_msecs(event_loop_timeout));
			ctrl->timeout = kt;
		} else if (kt > ctrl->timeout) {
			ctrl->defer_eoi = true;
		}
	}

	info = info_for_irq(irq);

	if (ctrl->defer_eoi) {
		info->eoi_cpu = smp_processor_id();
		info->irq_epoch = __this_cpu_read(irq_epoch);
		info->eoi_time = get_jiffies_64() + event_eoi_delay;
	}

	generic_handle_irq(irq);
}

1615
static void __xen_evtchn_do_upcall(void)
J
Jeremy Fitzhardinge 已提交
1616
{
C
Christoph Lameter 已提交
1617
	struct vcpu_info *vcpu_info = __this_cpu_read(xen_vcpu);
1618
	int cpu = smp_processor_id();
1619
	struct evtchn_loop_ctrl ctrl = { 0 };
J
Jeremy Fitzhardinge 已提交
1620

1621 1622
	read_lock(&evtchn_rwlock);

1623 1624
	do {
		vcpu_info->evtchn_upcall_pending = 0;
J
Jeremy Fitzhardinge 已提交
1625

1626
		xen_evtchn_handle_events(cpu, &ctrl);
J
Jeremy Fitzhardinge 已提交
1627

1628 1629
		BUG_ON(!irqs_disabled());

1630
		virt_rmb(); /* Hypervisor can set upcall pending. */
1631

1632
	} while (vcpu_info->evtchn_upcall_pending);
1633 1634

	read_unlock(&evtchn_rwlock);
1635 1636 1637 1638 1639 1640 1641

	/*
	 * Increment irq_epoch only now to defer EOIs only for
	 * xen_irq_lateeoi() invocations occurring from inside the loop
	 * above.
	 */
	__this_cpu_inc(irq_epoch);
1642 1643 1644 1645 1646 1647
}

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

1648
	irq_enter();
1649 1650 1651

	__xen_evtchn_do_upcall();

1652 1653
	irq_exit();
	set_irq_regs(old_regs);
1654
}
1655

1656 1657 1658
void xen_hvm_evtchn_do_upcall(void)
{
	__xen_evtchn_do_upcall();
J
Jeremy Fitzhardinge 已提交
1659
}
1660
EXPORT_SYMBOL_GPL(xen_hvm_evtchn_do_upcall);
J
Jeremy Fitzhardinge 已提交
1661

1662
/* Rebind a new event channel to an existing irq. */
1663
void rebind_evtchn_irq(evtchn_port_t evtchn, int irq)
1664
{
1665 1666
	struct irq_info *info = info_for_irq(irq);

1667 1668 1669
	if (WARN_ON(!info))
		return;

1670 1671 1672 1673
	/* Make sure the irq is masked, since the new event channel
	   will also be masked. */
	disable_irq(irq);

1674
	mutex_lock(&irq_mapping_update_lock);
1675 1676

	/* After resume the irq<->evtchn mappings are all cleared out */
1677
	BUG_ON(get_evtchn_to_irq(evtchn) != -1);
1678
	/* Expect irq to have been bound before,
1679 1680
	   so there should be a proper type */
	BUG_ON(info->type == IRQT_UNBOUND);
1681

1682
	(void)xen_irq_info_evtchn_setup(irq, evtchn);
1683

1684
	mutex_unlock(&irq_mapping_update_lock);
1685

1686
	bind_evtchn_to_cpu(evtchn, info->cpu, false);
1687 1688 1689 1690 1691

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

J
Jeremy Fitzhardinge 已提交
1692
/* Rebind an evtchn so that it gets delivered to a specific cpu */
1693
static int xen_rebind_evtchn_to_cpu(evtchn_port_t evtchn, unsigned int tcpu)
J
Jeremy Fitzhardinge 已提交
1694 1695
{
	struct evtchn_bind_vcpu bind_vcpu;
1696
	int masked;
J
Jeremy Fitzhardinge 已提交
1697

1698 1699 1700
	if (!VALID_EVTCHN(evtchn))
		return -1;

1701 1702 1703
	if (!xen_support_evtchn_rebind())
		return -1;

J
Jeremy Fitzhardinge 已提交
1704 1705
	/* Send future instances of this interrupt to other vcpu. */
	bind_vcpu.port = evtchn;
1706
	bind_vcpu.vcpu = xen_vcpu_nr(tcpu);
J
Jeremy Fitzhardinge 已提交
1707

1708 1709 1710 1711
	/*
	 * Mask the event while changing the VCPU binding to prevent
	 * it being delivered on an unexpected VCPU.
	 */
1712
	masked = test_and_set_mask(evtchn);
1713

J
Jeremy Fitzhardinge 已提交
1714 1715 1716 1717 1718 1719
	/*
	 * 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)
1720
		bind_evtchn_to_cpu(evtchn, tcpu, false);
J
Jeremy Fitzhardinge 已提交
1721

1722 1723 1724
	if (!masked)
		unmask_evtchn(evtchn);

1725 1726
	return 0;
}
J
Jeremy Fitzhardinge 已提交
1727

1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755
/*
 * Find the CPU within @dest mask which has the least number of channels
 * assigned. This is not precise as the per cpu counts can be modified
 * concurrently.
 */
static unsigned int select_target_cpu(const struct cpumask *dest)
{
	unsigned int cpu, best_cpu = UINT_MAX, minch = UINT_MAX;

	for_each_cpu_and(cpu, dest, cpu_online_mask) {
		unsigned int curch = atomic_read(&channels_on_cpu[cpu]);

		if (curch < minch) {
			minch = curch;
			best_cpu = cpu;
		}
	}

	/*
	 * Catch the unlikely case that dest contains no online CPUs. Can't
	 * recurse.
	 */
	if (best_cpu == UINT_MAX)
		return select_target_cpu(cpu_online_mask);

	return best_cpu;
}

1756 1757
static int set_affinity_irq(struct irq_data *data, const struct cpumask *dest,
			    bool force)
J
Jeremy Fitzhardinge 已提交
1758
{
1759
	unsigned int tcpu = select_target_cpu(dest);
1760
	int ret;
1761

1762
	ret = xen_rebind_evtchn_to_cpu(evtchn_from_irq(data->irq), tcpu);
1763 1764 1765 1766
	if (!ret)
		irq_data_update_effective_affinity(data, cpumask_of(tcpu));

	return ret;
J
Jeremy Fitzhardinge 已提交
1767 1768
}

1769
static void enable_dynirq(struct irq_data *data)
J
Jeremy Fitzhardinge 已提交
1770
{
1771
	evtchn_port_t evtchn = evtchn_from_irq(data->irq);
J
Jeremy Fitzhardinge 已提交
1772 1773 1774 1775 1776

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

1777
static void disable_dynirq(struct irq_data *data)
J
Jeremy Fitzhardinge 已提交
1778
{
1779
	evtchn_port_t evtchn = evtchn_from_irq(data->irq);
J
Jeremy Fitzhardinge 已提交
1780 1781 1782 1783 1784

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

1785
static void ack_dynirq(struct irq_data *data)
J
Jeremy Fitzhardinge 已提交
1786
{
1787
	evtchn_port_t evtchn = evtchn_from_irq(data->irq);
J
Jeremy Fitzhardinge 已提交
1788

1789 1790
	if (!VALID_EVTCHN(evtchn))
		return;
J
Jeremy Fitzhardinge 已提交
1791

1792
	clear_evtchn(evtchn);
1793 1794 1795 1796 1797 1798
}

static void mask_ack_dynirq(struct irq_data *data)
{
	disable_dynirq(data);
	ack_dynirq(data);
J
Jeremy Fitzhardinge 已提交
1799 1800
}

1801
static int retrigger_dynirq(struct irq_data *data)
J
Jeremy Fitzhardinge 已提交
1802
{
1803
	evtchn_port_t evtchn = evtchn_from_irq(data->irq);
1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814
	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
1815 1816
}

1817
static void restore_pirqs(void)
1818 1819 1820
{
	int pirq, rc, irq, gsi;
	struct physdev_map_pirq map_irq;
1821
	struct irq_info *info;
1822

1823 1824
	list_for_each_entry(info, &xen_irq_list_head, list) {
		if (info->type != IRQT_PIRQ)
1825 1826
			continue;

1827 1828 1829 1830
		pirq = info->u.pirq.pirq;
		gsi = info->u.pirq.gsi;
		irq = info->irq;

1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842
		/* 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) {
1843 1844
			pr_warn("xen map irq failed gsi=%d irq=%d pirq=%d rc=%d\n",
				gsi, irq, pirq, rc);
1845
			xen_free_irq(irq);
1846 1847 1848 1849 1850
			continue;
		}

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

1851
		__startup_pirq(irq);
1852 1853 1854
	}
}

1855 1856 1857
static void restore_cpu_virqs(unsigned int cpu)
{
	struct evtchn_bind_virq bind_virq;
1858 1859
	evtchn_port_t evtchn;
	int virq, irq;
1860 1861 1862 1863 1864

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

1865
		BUG_ON(virq_from_irq(irq) != virq);
1866 1867 1868

		/* Get a new binding from Xen. */
		bind_virq.virq = virq;
1869
		bind_virq.vcpu = xen_vcpu_nr(cpu);
1870 1871 1872 1873 1874 1875
		if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq,
						&bind_virq) != 0)
			BUG();
		evtchn = bind_virq.port;

		/* Record the new mapping. */
1876
		(void)xen_irq_info_virq_setup(cpu, irq, evtchn, virq);
1877 1878
		/* The affinity mask is still valid */
		bind_evtchn_to_cpu(evtchn, cpu, false);
1879 1880 1881 1882 1883 1884
	}
}

static void restore_cpu_ipis(unsigned int cpu)
{
	struct evtchn_bind_ipi bind_ipi;
1885 1886
	evtchn_port_t evtchn;
	int ipi, irq;
1887 1888 1889 1890 1891

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

1892
		BUG_ON(ipi_from_irq(irq) != ipi);
1893 1894

		/* Get a new binding from Xen. */
1895
		bind_ipi.vcpu = xen_vcpu_nr(cpu);
1896 1897 1898 1899 1900 1901
		if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_ipi,
						&bind_ipi) != 0)
			BUG();
		evtchn = bind_ipi.port;

		/* Record the new mapping. */
1902
		(void)xen_irq_info_ipi_setup(cpu, irq, evtchn, ipi);
1903 1904
		/* The affinity mask is still valid */
		bind_evtchn_to_cpu(evtchn, cpu, false);
1905 1906 1907
	}
}

1908 1909 1910
/* Clear an irq's pending state, in preparation for polling on it */
void xen_clear_irq_pending(int irq)
{
1911
	evtchn_port_t evtchn = evtchn_from_irq(irq);
1912 1913 1914 1915

	if (VALID_EVTCHN(evtchn))
		clear_evtchn(evtchn);
}
1916
EXPORT_SYMBOL(xen_clear_irq_pending);
1917 1918
void xen_set_irq_pending(int irq)
{
1919
	evtchn_port_t evtchn = evtchn_from_irq(irq);
1920 1921 1922 1923 1924 1925 1926

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

bool xen_test_irq_pending(int irq)
{
1927
	evtchn_port_t evtchn = evtchn_from_irq(irq);
1928 1929 1930 1931 1932 1933 1934 1935
	bool ret = false;

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

	return ret;
}

1936 1937 1938
/* 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)
1939 1940 1941 1942 1943 1944 1945
{
	evtchn_port_t evtchn = evtchn_from_irq(irq);

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

		poll.nr_ports = 1;
1946
		poll.timeout = timeout;
1947
		set_xen_guest_handle(poll.ports, &evtchn);
1948 1949 1950 1951 1952

		if (HYPERVISOR_sched_op(SCHEDOP_poll, &poll) != 0)
			BUG();
	}
}
1953 1954 1955 1956 1957 1958 1959
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 */);
}
1960

1961 1962 1963 1964
/* 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);
1965 1966 1967 1968 1969 1970
	struct physdev_irq_status_query irq_status;

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

	irq_status.irq = info->u.pirq.pirq;
1971 1972 1973 1974 1975 1976 1977

	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);

1978 1979
void xen_irq_resume(void)
{
1980
	unsigned int cpu;
1981
	struct irq_info *info;
1982 1983

	/* New event-channel space is not 'live' yet. */
1984
	xen_evtchn_resume();
1985 1986

	/* No IRQ <-> event-channel mappings. */
1987 1988 1989 1990 1991 1992
	list_for_each_entry(info, &xen_irq_list_head, list) {
		/* Zap event-channel binding */
		info->evtchn = 0;
		/* Adjust accounting */
		channels_on_cpu_dec(info);
	}
1993

1994
	clear_evtchn_to_irq_all();
1995 1996 1997 1998 1999

	for_each_possible_cpu(cpu) {
		restore_cpu_virqs(cpu);
		restore_cpu_ipis(cpu);
	}
2000

2001
	restore_pirqs();
2002 2003
}

J
Jeremy Fitzhardinge 已提交
2004
static struct irq_chip xen_dynamic_chip __read_mostly = {
2005
	.name			= "xen-dyn",
2006

2007 2008 2009
	.irq_disable		= disable_dynirq,
	.irq_mask		= disable_dynirq,
	.irq_unmask		= enable_dynirq,
2010

2011 2012 2013
	.irq_ack		= ack_dynirq,
	.irq_mask_ack		= mask_ack_dynirq,

2014 2015
	.irq_set_affinity	= set_affinity_irq,
	.irq_retrigger		= retrigger_dynirq,
J
Jeremy Fitzhardinge 已提交
2016 2017
};

2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032
static struct irq_chip xen_lateeoi_chip __read_mostly = {
	/* The chip name needs to contain "xen-dyn" for irqbalance to work. */
	.name			= "xen-dyn-lateeoi",

	.irq_disable		= disable_dynirq,
	.irq_mask		= disable_dynirq,
	.irq_unmask		= enable_dynirq,

	.irq_ack		= mask_ack_dynirq,
	.irq_mask_ack		= mask_ack_dynirq,

	.irq_set_affinity	= set_affinity_irq,
	.irq_retrigger		= retrigger_dynirq,
};

2033
static struct irq_chip xen_pirq_chip __read_mostly = {
2034
	.name			= "xen-pirq",
2035

2036 2037 2038 2039
	.irq_startup		= startup_pirq,
	.irq_shutdown		= shutdown_pirq,
	.irq_enable		= enable_pirq,
	.irq_disable		= disable_pirq,
2040

2041 2042 2043 2044 2045 2046
	.irq_mask		= disable_dynirq,
	.irq_unmask		= enable_dynirq,

	.irq_ack		= eoi_pirq,
	.irq_eoi		= eoi_pirq,
	.irq_mask_ack		= mask_ack_pirq,
2047

2048
	.irq_set_affinity	= set_affinity_irq,
2049

2050
	.irq_retrigger		= retrigger_dynirq,
2051 2052
};

2053
static struct irq_chip xen_percpu_chip __read_mostly = {
2054
	.name			= "xen-percpu",
2055

2056 2057 2058
	.irq_disable		= disable_dynirq,
	.irq_mask		= disable_dynirq,
	.irq_unmask		= enable_dynirq,
2059

2060
	.irq_ack		= ack_dynirq,
2061 2062
};

2063
#ifdef CONFIG_XEN_PVHVM
2064 2065 2066
/* 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. */
2067
void xen_setup_callback_vector(void)
2068 2069
{
	uint64_t callback_via;
2070

2071 2072
	if (xen_have_vector_callback) {
		callback_via = HVM_CALLBACK_VECTOR(HYPERVISOR_CALLBACK_VECTOR);
2073
		if (xen_set_callback_via(callback_via)) {
2074 2075 2076 2077
			pr_err("Request for Xen HVM callback vector failed\n");
			xen_have_vector_callback = 0;
		}
	}
2078
}
2079 2080 2081 2082 2083 2084 2085

static __init void xen_alloc_callback_vector(void)
{
	if (!xen_have_vector_callback)
		return;

	pr_info("Xen HVM callback vector for event delivery is enabled\n");
2086
	alloc_intr_gate(HYPERVISOR_CALLBACK_VECTOR, asm_sysvec_xen_hvm_callback);
2087
}
2088
#else
2089 2090
void xen_setup_callback_vector(void) {}
static inline void xen_alloc_callback_vector(void) {}
2091
#endif
2092

2093 2094
bool xen_fifo_events = true;
module_param_named(fifo_events, xen_fifo_events, bool, 0);
2095

2096 2097 2098 2099
static int xen_evtchn_cpu_prepare(unsigned int cpu)
{
	int ret = 0;

2100 2101
	xen_cpu_init_eoi(cpu);

2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117
	if (evtchn_ops->percpu_init)
		ret = evtchn_ops->percpu_init(cpu);

	return ret;
}

static int xen_evtchn_cpu_dead(unsigned int cpu)
{
	int ret = 0;

	if (evtchn_ops->percpu_deinit)
		ret = evtchn_ops->percpu_deinit(cpu);

	return ret;
}

2118
void __init xen_init_IRQ(void)
J
Jeremy Fitzhardinge 已提交
2119
{
2120
	int ret = -EINVAL;
2121
	evtchn_port_t evtchn;
2122

2123
	if (xen_fifo_events)
2124
		ret = xen_evtchn_fifo_init();
2125
	if (ret < 0) {
2126
		xen_evtchn_2l_init();
2127 2128
		xen_fifo_events = false;
	}
2129

2130 2131
	xen_cpu_init_eoi(smp_processor_id());

2132 2133 2134 2135
	cpuhp_setup_state_nocalls(CPUHP_XEN_EVTCHN_PREPARE,
				  "xen/evtchn:prepare",
				  xen_evtchn_cpu_prepare, xen_evtchn_cpu_dead);

2136 2137
	evtchn_to_irq = kcalloc(EVTCHN_ROW(xen_evtchn_max_channels()),
				sizeof(*evtchn_to_irq), GFP_KERNEL);
2138
	BUG_ON(!evtchn_to_irq);
J
Jeremy Fitzhardinge 已提交
2139 2140

	/* No event channels are 'live' right now. */
2141 2142
	for (evtchn = 0; evtchn < xen_evtchn_nr_channels(); evtchn++)
		mask_evtchn(evtchn);
J
Jeremy Fitzhardinge 已提交
2143

2144 2145
	pirq_needs_eoi = pirq_needs_eoi_flag;

2146
#ifdef CONFIG_X86
2147 2148 2149 2150
	if (xen_pv_domain()) {
		if (xen_initial_domain())
			pci_xen_initial_domain();
	}
2151 2152 2153 2154
	if (xen_feature(XENFEAT_hvm_callback_vector)) {
		xen_setup_callback_vector();
		xen_alloc_callback_vector();
	}
2155 2156

	if (xen_hvm_domain()) {
2157
		native_init_IRQ();
2158 2159 2160
		/* 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();
2161
	} else {
2162
		int rc;
2163 2164 2165
		struct physdev_pirq_eoi_gmfn eoi_gmfn;

		pirq_eoi_map = (void *)__get_free_page(GFP_KERNEL|__GFP_ZERO);
2166
		eoi_gmfn.gmfn = virt_to_gfn(pirq_eoi_map);
2167 2168 2169 2170 2171 2172
		rc = HYPERVISOR_physdev_op(PHYSDEVOP_pirq_eoi_gmfn_v2, &eoi_gmfn);
		if (rc != 0) {
			free_page((unsigned long) pirq_eoi_map);
			pirq_eoi_map = NULL;
		} else
			pirq_needs_eoi = pirq_check_eoi_map;
2173
	}
2174
#endif
J
Jeremy Fitzhardinge 已提交
2175
}
反馈
建议
客服 返回
顶部