[deliverable/linux.git] / drivers / xen / events / events_fifo.c

/*
 * Xen event channels (FIFO-based ABI)
 *
 * Copyright (C) 2013 Citrix Systems R&D ltd.
 *
 * This source code is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License as
 * published by the Free Software Foundation; either version 2 of the
 * License, or (at your option) any later version.
 *
 * Or, when distributed separately from the Linux kernel or
 * incorporated into other software packages, subject to the following
 * license:
 *
 * Permission is hereby granted, free of charge, to any person obtaining a copy
 * of this source file (the "Software"), to deal in the Software without
 * restriction, including without limitation the rights to use, copy, modify,
 * merge, publish, distribute, sublicense, and/or sell copies of the Software,
 * and to permit persons to whom the Software is furnished to do so, subject to
 * the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
 * IN THE SOFTWARE.
 */

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

#include <linux/linkage.h>
#include <linux/interrupt.h>
#include <linux/irq.h>
#include <linux/module.h>
#include <linux/smp.h>
#include <linux/percpu.h>
#include <linux/cpu.h>

#include <asm/sync_bitops.h>
#include <asm/xen/hypercall.h>
#include <asm/xen/hypervisor.h>

#include <xen/xen.h>
#include <xen/xen-ops.h>
#include <xen/events.h>
#include <xen/interface/xen.h>
#include <xen/interface/event_channel.h>
#include <xen/page.h>

#include "events_internal.h"

#define EVENT_WORDS_PER_PAGE (PAGE_SIZE / sizeof(event_word_t))
#define MAX_EVENT_ARRAY_PAGES (EVTCHN_FIFO_NR_CHANNELS / EVENT_WORDS_PER_PAGE)

struct evtchn_fifo_queue {
	uint32_t head[EVTCHN_FIFO_MAX_QUEUES];
};

static DEFINE_PER_CPU(struct evtchn_fifo_control_block *, cpu_control_block);
static DEFINE_PER_CPU(struct evtchn_fifo_queue, cpu_queue);
static event_word_t *event_array[MAX_EVENT_ARRAY_PAGES] __read_mostly;
static unsigned event_array_pages __read_mostly;

/*
 * sync_set_bit() and friends must be unsigned long aligned.
 */
#if BITS_PER_LONG > 32

#define BM(w) (unsigned long *)((unsigned long)w & ~0x7UL)
#define EVTCHN_FIFO_BIT(b, w) \
    (((unsigned long)w & 0x4UL) ? (EVTCHN_FIFO_ ##b + 32) : EVTCHN_FIFO_ ##b)

#else

#define BM(w) ((unsigned long *)(w))
#define EVTCHN_FIFO_BIT(b, w) EVTCHN_FIFO_ ##b

#endif

static inline event_word_t *event_word_from_port(unsigned port)
{
	unsigned i = port / EVENT_WORDS_PER_PAGE;

	return event_array[i] + port % EVENT_WORDS_PER_PAGE;
}

static unsigned evtchn_fifo_max_channels(void)
{
	return EVTCHN_FIFO_NR_CHANNELS;
}

static unsigned evtchn_fifo_nr_channels(void)
{
	return event_array_pages * EVENT_WORDS_PER_PAGE;
}

static int init_control_block(int cpu,
                              struct evtchn_fifo_control_block *control_block)
{
	struct evtchn_fifo_queue *q = &per_cpu(cpu_queue, cpu);
	struct evtchn_init_control init_control;
	unsigned int i;

	/* Reset the control block and the local HEADs. */
	clear_page(control_block);
	for (i = 0; i < EVTCHN_FIFO_MAX_QUEUES; i++)
		q->head[i] = 0;

	init_control.control_gfn = virt_to_mfn(control_block);
	init_control.offset      = 0;
	init_control.vcpu        = cpu;

	return HYPERVISOR_event_channel_op(EVTCHNOP_init_control, &init_control);
}

static void free_unused_array_pages(void)
{
	unsigned i;

	for (i = event_array_pages; i < MAX_EVENT_ARRAY_PAGES; i++) {
		if (!event_array[i])
			break;
		free_page((unsigned long)event_array[i]);
		event_array[i] = NULL;
	}
}

static void init_array_page(event_word_t *array_page)
{
	unsigned i;

	for (i = 0; i < EVENT_WORDS_PER_PAGE; i++)
		array_page[i] = 1 << EVTCHN_FIFO_MASKED;
}

static int evtchn_fifo_setup(struct irq_info *info)
{
	unsigned port = info->evtchn;
	unsigned new_array_pages;
	int ret;

	new_array_pages = port / EVENT_WORDS_PER_PAGE + 1;

	if (new_array_pages > MAX_EVENT_ARRAY_PAGES)
		return -EINVAL;

	while (event_array_pages < new_array_pages) {
		void *array_page;
		struct evtchn_expand_array expand_array;

		/* Might already have a page if we've resumed. */
		array_page = event_array[event_array_pages];
		if (!array_page) {
			array_page = (void *)__get_free_page(GFP_KERNEL);
			if (array_page == NULL) {
				ret = -ENOMEM;
				goto error;
			}
			event_array[event_array_pages] = array_page;
		}

		/* Mask all events in this page before adding it. */
		init_array_page(array_page);

		expand_array.array_gfn = virt_to_mfn(array_page);

		ret = HYPERVISOR_event_channel_op(EVTCHNOP_expand_array, &expand_array);
		if (ret < 0)
			goto error;

		event_array_pages++;
	}
	return 0;

  error:
	if (event_array_pages == 0)
		panic("xen: unable to expand event array with initial page (%d)\n", ret);
	else
		pr_err("unable to expand event array (%d)\n", ret);
	free_unused_array_pages();
	return ret;
}

static void evtchn_fifo_bind_to_cpu(struct irq_info *info, unsigned cpu)
{
	/* no-op */
}

static void evtchn_fifo_clear_pending(unsigned port)
{
	event_word_t *word = event_word_from_port(port);
	sync_clear_bit(EVTCHN_FIFO_BIT(PENDING, word), BM(word));
}

static void evtchn_fifo_set_pending(unsigned port)
{
	event_word_t *word = event_word_from_port(port);
	sync_set_bit(EVTCHN_FIFO_BIT(PENDING, word), BM(word));
}

static bool evtchn_fifo_is_pending(unsigned port)
{
	event_word_t *word = event_word_from_port(port);
	return sync_test_bit(EVTCHN_FIFO_BIT(PENDING, word), BM(word));
}

static bool evtchn_fifo_test_and_set_mask(unsigned port)
{
	event_word_t *word = event_word_from_port(port);
	return sync_test_and_set_bit(EVTCHN_FIFO_BIT(MASKED, word), BM(word));
}

static void evtchn_fifo_mask(unsigned port)
{
	event_word_t *word = event_word_from_port(port);
	sync_set_bit(EVTCHN_FIFO_BIT(MASKED, word), BM(word));
}

static bool evtchn_fifo_is_masked(unsigned port)
{
	event_word_t *word = event_word_from_port(port);
	return sync_test_bit(EVTCHN_FIFO_BIT(MASKED, word), BM(word));
}
/*
 * Clear MASKED, spinning if BUSY is set.
 */
static void clear_masked(volatile event_word_t *word)
{
	event_word_t new, old, w;

	w = *word;

	do {
		old = w & ~(1 << EVTCHN_FIFO_BUSY);
		new = old & ~(1 << EVTCHN_FIFO_MASKED);
		w = sync_cmpxchg(word, old, new);
	} while (w != old);
}

static void evtchn_fifo_unmask(unsigned port)
{
	event_word_t *word = event_word_from_port(port);

	BUG_ON(!irqs_disabled());

	clear_masked(word);
	if (evtchn_fifo_is_pending(port)) {
		struct evtchn_unmask unmask = { .port = port };
		(void)HYPERVISOR_event_channel_op(EVTCHNOP_unmask, &unmask);
	}
}

static bool evtchn_fifo_is_linked(unsigned port)
{
	event_word_t *word = event_word_from_port(port);
	return sync_test_bit(EVTCHN_FIFO_BIT(LINKED, word), BM(word));
}

static uint32_t clear_linked(volatile event_word_t *word)
{
	event_word_t new, old, w;

	w = *word;

	do {
		old = w;
		new = (w & ~((1 << EVTCHN_FIFO_LINKED)
			     | EVTCHN_FIFO_LINK_MASK));
	} while ((w = sync_cmpxchg(word, old, new)) != old);

	return w & EVTCHN_FIFO_LINK_MASK;
}

static void handle_irq_for_port(unsigned port)
{
	int irq;

	irq = get_evtchn_to_irq(port);
	if (irq != -1)
		generic_handle_irq(irq);
}

static void consume_one_event(unsigned cpu,
			      struct evtchn_fifo_control_block *control_block,
			      unsigned priority, unsigned long *ready,
			      bool drop)
{
	struct evtchn_fifo_queue *q = &per_cpu(cpu_queue, cpu);
	uint32_t head;
	unsigned port;
	event_word_t *word;

	head = q->head[priority];

	/*
	 * Reached the tail last time?  Read the new HEAD from the
	 * control block.
	 */
	if (head == 0) {
		rmb(); /* Ensure word is up-to-date before reading head. */
		head = control_block->head[priority];
	}

	port = head;
	word = event_word_from_port(port);
	head = clear_linked(word);

	/*
	 * If the link is non-zero, there are more events in the
	 * queue, otherwise the queue is empty.
	 *
	 * If the queue is empty, clear this priority from our local
	 * copy of the ready word.
	 */
	if (head == 0)
		clear_bit(priority, ready);

	if (evtchn_fifo_is_pending(port) && !evtchn_fifo_is_masked(port)) {
		if (likely(!drop))
			handle_irq_for_port(port);
	}

	q->head[priority] = head;
}

static void __evtchn_fifo_handle_events(unsigned cpu, bool drop)
{
	struct evtchn_fifo_control_block *control_block;
	unsigned long ready;
	unsigned q;

	control_block = per_cpu(cpu_control_block, cpu);

	ready = xchg(&control_block->ready, 0);

	while (ready) {
		q = find_first_bit(&ready, EVTCHN_FIFO_MAX_QUEUES);
		consume_one_event(cpu, control_block, q, &ready, drop);
		ready |= xchg(&control_block->ready, 0);
	}
}

static void evtchn_fifo_handle_events(unsigned cpu)
{
	__evtchn_fifo_handle_events(cpu, false);
}

static void evtchn_fifo_resume(void)
{
	unsigned cpu;

	for_each_possible_cpu(cpu) {
		void *control_block = per_cpu(cpu_control_block, cpu);
		int ret;

		if (!control_block)
			continue;

		/*
		 * If this CPU is offline, take the opportunity to
		 * free the control block while it is not being
		 * used.
		 */
		if (!cpu_online(cpu)) {
			free_page((unsigned long)control_block);
			per_cpu(cpu_control_block, cpu) = NULL;
			continue;
		}

		ret = init_control_block(cpu, control_block);
		if (ret < 0)
			BUG();
	}

	/*
	 * The event array starts out as empty again and is extended
	 * as normal when events are bound.  The existing pages will
	 * be reused.
	 */
	event_array_pages = 0;
}

static void evtchn_fifo_close(unsigned port, unsigned int cpu)
{
	if (cpu == NR_CPUS)
		return;

	get_online_cpus();
	if (cpu_online(cpu)) {
		if (WARN_ON(irqs_disabled()))
			goto out;

		while (evtchn_fifo_is_linked(port))
			cpu_relax();
	} else {
		__evtchn_fifo_handle_events(cpu, true);
	}

out:
	put_online_cpus();
}

static const struct evtchn_ops evtchn_ops_fifo = {
	.max_channels      = evtchn_fifo_max_channels,
	.nr_channels       = evtchn_fifo_nr_channels,
	.setup             = evtchn_fifo_setup,
	.bind_to_cpu       = evtchn_fifo_bind_to_cpu,
	.clear_pending     = evtchn_fifo_clear_pending,
	.set_pending       = evtchn_fifo_set_pending,
	.is_pending        = evtchn_fifo_is_pending,
	.test_and_set_mask = evtchn_fifo_test_and_set_mask,
	.mask              = evtchn_fifo_mask,
	.unmask            = evtchn_fifo_unmask,
	.handle_events     = evtchn_fifo_handle_events,
	.resume            = evtchn_fifo_resume,
	.close             = evtchn_fifo_close,
};

static int evtchn_fifo_alloc_control_block(unsigned cpu)
{
	void *control_block = NULL;
	int ret = -ENOMEM;

	control_block = (void *)__get_free_page(GFP_KERNEL);
	if (control_block == NULL)
		goto error;

	ret = init_control_block(cpu, control_block);
	if (ret < 0)
		goto error;

	per_cpu(cpu_control_block, cpu) = control_block;

	return 0;

  error:
	free_page((unsigned long)control_block);
	return ret;
}

static int evtchn_fifo_cpu_notification(struct notifier_block *self,
						  unsigned long action,
						  void *hcpu)
{
	int cpu = (long)hcpu;
	int ret = 0;

	switch (action) {
	case CPU_UP_PREPARE:
		if (!per_cpu(cpu_control_block, cpu))
			ret = evtchn_fifo_alloc_control_block(cpu);
		break;
	default:
		break;
	}
	return ret < 0 ? NOTIFY_BAD : NOTIFY_OK;
}

static struct notifier_block evtchn_fifo_cpu_notifier = {
	.notifier_call	= evtchn_fifo_cpu_notification,
};

int __init xen_evtchn_fifo_init(void)
{
	int cpu = get_cpu();
	int ret;

	ret = evtchn_fifo_alloc_control_block(cpu);
	if (ret < 0)
		goto out;

	pr_info("Using FIFO-based ABI\n");

	evtchn_ops = &evtchn_ops_fifo;

	register_cpu_notifier(&evtchn_fifo_cpu_notifier);
out:
	put_cpu();
	return ret;
}
Commit	Line	Data
1fe56551 DV	1	/*
	2	* Xen event channels (FIFO-based ABI)
	3	*
	4	* Copyright (C) 2013 Citrix Systems R&D ltd.
	5	*
	6	* This source code is free software; you can redistribute it and/or
	7	* modify it under the terms of the GNU General Public License as
	8	* published by the Free Software Foundation; either version 2 of the
	9	* License, or (at your option) any later version.
	10	*
	11	* Or, when distributed separately from the Linux kernel or
	12	* incorporated into other software packages, subject to the following
	13	* license:
	14	*
	15	* Permission is hereby granted, free of charge, to any person obtaining a copy
	16	* of this source file (the "Software"), to deal in the Software without
	17	* restriction, including without limitation the rights to use, copy, modify,
	18	* merge, publish, distribute, sublicense, and/or sell copies of the Software,
	19	* and to permit persons to whom the Software is furnished to do so, subject to
	20	* the following conditions:
	21	*
	22	* The above copyright notice and this permission notice shall be included in
	23	* all copies or substantial portions of the Software.
	24	*
	25	* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
	26	* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
	27	* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
	28	* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
	29	* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
	30	* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
	31	* IN THE SOFTWARE.
	32	*/
	33
	34	#define pr_fmt(fmt) "xen:" KBUILD_MODNAME ": " fmt
	35
	36	#include <linux/linkage.h>
	37	#include <linux/interrupt.h>
	38	#include <linux/irq.h>
	39	#include <linux/module.h>
	40	#include <linux/smp.h>
	41	#include <linux/percpu.h>
	42	#include <linux/cpu.h>
	43
	44	#include <asm/sync_bitops.h>
	45	#include <asm/xen/hypercall.h>
	46	#include <asm/xen/hypervisor.h>
1fe56551 DV	47
	48	#include <xen/xen.h>
	49	#include <xen/xen-ops.h>
	50	#include <xen/events.h>
	51	#include <xen/interface/xen.h>
	52	#include <xen/interface/event_channel.h>
a9fd60e2	53	#include <xen/page.h>
1fe56551 DV	54
	55	#include "events_internal.h"
	56
	57	#define EVENT_WORDS_PER_PAGE (PAGE_SIZE / sizeof(event_word_t))
	58	#define MAX_EVENT_ARRAY_PAGES (EVTCHN_FIFO_NR_CHANNELS / EVENT_WORDS_PER_PAGE)
	59
	60	struct evtchn_fifo_queue {
	61	uint32_t head[EVTCHN_FIFO_MAX_QUEUES];
	62	};
	63
	64	static DEFINE_PER_CPU(struct evtchn_fifo_control_block *, cpu_control_block);
	65	static DEFINE_PER_CPU(struct evtchn_fifo_queue, cpu_queue);
	66	static event_word_t *event_array[MAX_EVENT_ARRAY_PAGES] __read_mostly;
	67	static unsigned event_array_pages __read_mostly;
	68
05a812ac	69	/*
dcecb8fd	70	* sync_set_bit() and friends must be unsigned long aligned.
05a812ac	71	*/
dcecb8fd	72	#if BITS_PER_LONG > 32
05a812ac VM	73
	74	#define BM(w) (unsigned long *)((unsigned long)w & ~0x7UL)
	75	#define EVTCHN_FIFO_BIT(b, w) \
	76	(((unsigned long)w & 0x4UL) ? (EVTCHN_FIFO_ ##b + 32) : EVTCHN_FIFO_ ##b)
	77
	78	#else
	79
1fe56551	80	#define BM(w) ((unsigned long *)(w))
05a812ac VM	81	#define EVTCHN_FIFO_BIT(b, w) EVTCHN_FIFO_ ##b
	82
	83	#endif
1fe56551 DV	84
	85	static inline event_word_t *event_word_from_port(unsigned port)
	86	{
	87	unsigned i = port / EVENT_WORDS_PER_PAGE;
	88
	89	return event_array[i] + port % EVENT_WORDS_PER_PAGE;
	90	}
	91
	92	static unsigned evtchn_fifo_max_channels(void)
	93	{
	94	return EVTCHN_FIFO_NR_CHANNELS;
	95	}
	96
	97	static unsigned evtchn_fifo_nr_channels(void)
	98	{
	99	return event_array_pages * EVENT_WORDS_PER_PAGE;
	100	}
	101
c12784c3 DV	102	static int init_control_block(int cpu,
	103	struct evtchn_fifo_control_block *control_block)
	104	{
	105	struct evtchn_fifo_queue *q = &per_cpu(cpu_queue, cpu);
	106	struct evtchn_init_control init_control;
	107	unsigned int i;
	108
	109	/* Reset the control block and the local HEADs. */
	110	clear_page(control_block);
	111	for (i = 0; i < EVTCHN_FIFO_MAX_QUEUES; i++)
	112	q->head[i] = 0;
	113
	114	init_control.control_gfn = virt_to_mfn(control_block);
	115	init_control.offset = 0;
	116	init_control.vcpu = cpu;
	117
	118	return HYPERVISOR_event_channel_op(EVTCHNOP_init_control, &init_control);
	119	}
	120
1fe56551 DV	121	static void free_unused_array_pages(void)
	122	{
	123	unsigned i;
	124
	125	for (i = event_array_pages; i < MAX_EVENT_ARRAY_PAGES; i++) {
	126	if (!event_array[i])
	127	break;
	128	free_page((unsigned long)event_array[i]);
	129	event_array[i] = NULL;
	130	}
	131	}
	132
	133	static void init_array_page(event_word_t *array_page)
	134	{
	135	unsigned i;
	136
	137	for (i = 0; i < EVENT_WORDS_PER_PAGE; i++)
	138	array_page[i] = 1 << EVTCHN_FIFO_MASKED;
	139	}
	140
	141	static int evtchn_fifo_setup(struct irq_info *info)
	142	{
	143	unsigned port = info->evtchn;
	144	unsigned new_array_pages;
be1403b9	145	int ret;
1fe56551 DV	146
	147	new_array_pages = port / EVENT_WORDS_PER_PAGE + 1;
	148
	149	if (new_array_pages > MAX_EVENT_ARRAY_PAGES)
	150	return -EINVAL;
	151
	152	while (event_array_pages < new_array_pages) {
	153	void *array_page;
	154	struct evtchn_expand_array expand_array;
	155
	156	/* Might already have a page if we've resumed. */
	157	array_page = event_array[event_array_pages];
	158	if (!array_page) {
	159	array_page = (void *)__get_free_page(GFP_KERNEL);
be1403b9 WY	160	if (array_page == NULL) {
be1403b9 WY	161	ret = -ENOMEM;
1fe56551	162	goto error;
be1403b9	163	}
1fe56551 DV	164	event_array[event_array_pages] = array_page;
	165	}
	166
	167	/* Mask all events in this page before adding it. */
	168	init_array_page(array_page);
	169
	170	expand_array.array_gfn = virt_to_mfn(array_page);
	171
	172	ret = HYPERVISOR_event_channel_op(EVTCHNOP_expand_array, &expand_array);
	173	if (ret < 0)
	174	goto error;
	175
	176	event_array_pages++;
	177	}
	178	return 0;
	179
	180	error:
	181	if (event_array_pages == 0)
	182	panic("xen: unable to expand event array with initial page (%d)\n", ret);
	183	else
	184	pr_err("unable to expand event array (%d)\n", ret);
	185	free_unused_array_pages();
	186	return ret;
	187	}
	188
	189	static void evtchn_fifo_bind_to_cpu(struct irq_info *info, unsigned cpu)
	190	{
	191	/* no-op */
	192	}
	193
	194	static void evtchn_fifo_clear_pending(unsigned port)
	195	{
	196	event_word_t *word = event_word_from_port(port);
05a812ac	197	sync_clear_bit(EVTCHN_FIFO_BIT(PENDING, word), BM(word));
1fe56551 DV	198	}
	199
	200	static void evtchn_fifo_set_pending(unsigned port)
	201	{
	202	event_word_t *word = event_word_from_port(port);
05a812ac	203	sync_set_bit(EVTCHN_FIFO_BIT(PENDING, word), BM(word));
1fe56551 DV	204	}
	205
	206	static bool evtchn_fifo_is_pending(unsigned port)
	207	{
	208	event_word_t *word = event_word_from_port(port);
05a812ac	209	return sync_test_bit(EVTCHN_FIFO_BIT(PENDING, word), BM(word));
1fe56551 DV	210	}
	211
	212	static bool evtchn_fifo_test_and_set_mask(unsigned port)
	213	{
	214	event_word_t *word = event_word_from_port(port);
05a812ac	215	return sync_test_and_set_bit(EVTCHN_FIFO_BIT(MASKED, word), BM(word));
1fe56551 DV	216	}
	217
	218	static void evtchn_fifo_mask(unsigned port)
	219	{
	220	event_word_t *word = event_word_from_port(port);
05a812ac	221	sync_set_bit(EVTCHN_FIFO_BIT(MASKED, word), BM(word));
1fe56551 DV	222	}
1fe56551 DV	223
05a812ac VM	224	static bool evtchn_fifo_is_masked(unsigned port)
	225	{
	226	event_word_t *word = event_word_from_port(port);
	227	return sync_test_bit(EVTCHN_FIFO_BIT(MASKED, word), BM(word));
	228	}
1fe56551 DV	229	/*
	230	* Clear MASKED, spinning if BUSY is set.
	231	*/
	232	static void clear_masked(volatile event_word_t *word)
	233	{
	234	event_word_t new, old, w;
	235
	236	w = *word;
	237
	238	do {
	239	old = w & ~(1 << EVTCHN_FIFO_BUSY);
	240	new = old & ~(1 << EVTCHN_FIFO_MASKED);
	241	w = sync_cmpxchg(word, old, new);
	242	} while (w != old);
	243	}
	244
	245	static void evtchn_fifo_unmask(unsigned port)
	246	{
	247	event_word_t *word = event_word_from_port(port);
	248
	249	BUG_ON(!irqs_disabled());
	250
	251	clear_masked(word);
05a812ac	252	if (evtchn_fifo_is_pending(port)) {
1fe56551 DV	253	struct evtchn_unmask unmask = { .port = port };
	254	(void)HYPERVISOR_event_channel_op(EVTCHNOP_unmask, &unmask);
	255	}
	256	}
	257
fcdf31a7 RL	258	static bool evtchn_fifo_is_linked(unsigned port)
	259	{
	260	event_word_t *word = event_word_from_port(port);
	261	return sync_test_bit(EVTCHN_FIFO_BIT(LINKED, word), BM(word));
	262	}
	263
1fe56551 DV	264	static uint32_t clear_linked(volatile event_word_t *word)
	265	{
	266	event_word_t new, old, w;
	267
	268	w = *word;
	269
	270	do {
	271	old = w;
	272	new = (w & ~((1 << EVTCHN_FIFO_LINKED)
	273	\| EVTCHN_FIFO_LINK_MASK));
	274	} while ((w = sync_cmpxchg(word, old, new)) != old);
	275
	276	return w & EVTCHN_FIFO_LINK_MASK;
	277	}
	278
	279	static void handle_irq_for_port(unsigned port)
	280	{
	281	int irq;
1fe56551 DV	282
1fe56551 DV	283	irq = get_evtchn_to_irq(port);
589d03e9 TG	284	if (irq != -1)
589d03e9 TG	285	generic_handle_irq(irq);
1fe56551 DV	286	}
	287
	288	static void consume_one_event(unsigned cpu,
	289	struct evtchn_fifo_control_block *control_block,
fcdf31a7 RL	290	unsigned priority, unsigned long *ready,
fcdf31a7 RL	291	bool drop)
1fe56551 DV	292	{
	293	struct evtchn_fifo_queue *q = &per_cpu(cpu_queue, cpu);
	294	uint32_t head;
	295	unsigned port;
	296	event_word_t *word;
	297
	298	head = q->head[priority];
	299
	300	/*
	301	* Reached the tail last time? Read the new HEAD from the
	302	* control block.
	303	*/
	304	if (head == 0) {
	305	rmb(); /* Ensure word is up-to-date before reading head. */
	306	head = control_block->head[priority];
	307	}
	308
	309	port = head;
	310	word = event_word_from_port(port);
	311	head = clear_linked(word);
	312
	313	/*
	314	* If the link is non-zero, there are more events in the
	315	* queue, otherwise the queue is empty.
	316	*
	317	* If the queue is empty, clear this priority from our local
	318	* copy of the ready word.
	319	*/
	320	if (head == 0)
05a812ac	321	clear_bit(priority, ready);
1fe56551	322
fcdf31a7 RL	323	if (evtchn_fifo_is_pending(port) && !evtchn_fifo_is_masked(port)) {
	324	if (likely(!drop))
	325	handle_irq_for_port(port);
	326	}
1fe56551 DV	327
	328	q->head[priority] = head;
	329	}
	330
fcdf31a7	331	static void __evtchn_fifo_handle_events(unsigned cpu, bool drop)
1fe56551 DV	332	{
1fe56551 DV	333	struct evtchn_fifo_control_block *control_block;
05a812ac	334	unsigned long ready;
1fe56551 DV	335	unsigned q;
	336
	337	control_block = per_cpu(cpu_control_block, cpu);
	338
	339	ready = xchg(&control_block->ready, 0);
	340
	341	while (ready) {
e4a74312	342	q = find_first_bit(&ready, EVTCHN_FIFO_MAX_QUEUES);
fcdf31a7	343	consume_one_event(cpu, control_block, q, &ready, drop);
1fe56551 DV	344	ready \|= xchg(&control_block->ready, 0);
	345	}
	346	}
	347
fcdf31a7 RL	348	static void evtchn_fifo_handle_events(unsigned cpu)
	349	{
	350	__evtchn_fifo_handle_events(cpu, false);
	351	}
	352
1fe56551 DV	353	static void evtchn_fifo_resume(void)
	354	{
	355	unsigned cpu;
	356
	357	for_each_possible_cpu(cpu) {
	358	void *control_block = per_cpu(cpu_control_block, cpu);
1fe56551 DV	359	int ret;
	360
	361	if (!control_block)
	362	continue;
	363
	364	/*
	365	* If this CPU is offline, take the opportunity to
	366	* free the control block while it is not being
	367	* used.
	368	*/
	369	if (!cpu_online(cpu)) {
	370	free_page((unsigned long)control_block);
	371	per_cpu(cpu_control_block, cpu) = NULL;
	372	continue;
	373	}
	374
c12784c3	375	ret = init_control_block(cpu, control_block);
1fe56551 DV	376	if (ret < 0)
	377	BUG();
	378	}
	379
	380	/*
	381	* The event array starts out as empty again and is extended
	382	* as normal when events are bound. The existing pages will
	383	* be reused.
	384	*/
	385	event_array_pages = 0;
	386	}
	387
fcdf31a7 RL	388	static void evtchn_fifo_close(unsigned port, unsigned int cpu)
	389	{
	390	if (cpu == NR_CPUS)
	391	return;
	392
	393	get_online_cpus();
	394	if (cpu_online(cpu)) {
	395	if (WARN_ON(irqs_disabled()))
	396	goto out;
	397
	398	while (evtchn_fifo_is_linked(port))
	399	cpu_relax();
	400	} else {
	401	__evtchn_fifo_handle_events(cpu, true);
	402	}
	403
	404	out:
	405	put_online_cpus();
	406	}
	407
1fe56551 DV	408	static const struct evtchn_ops evtchn_ops_fifo = {
	409	.max_channels = evtchn_fifo_max_channels,
	410	.nr_channels = evtchn_fifo_nr_channels,
	411	.setup = evtchn_fifo_setup,
	412	.bind_to_cpu = evtchn_fifo_bind_to_cpu,
	413	.clear_pending = evtchn_fifo_clear_pending,
	414	.set_pending = evtchn_fifo_set_pending,
	415	.is_pending = evtchn_fifo_is_pending,
	416	.test_and_set_mask = evtchn_fifo_test_and_set_mask,
	417	.mask = evtchn_fifo_mask,
	418	.unmask = evtchn_fifo_unmask,
	419	.handle_events = evtchn_fifo_handle_events,
	420	.resume = evtchn_fifo_resume,
fcdf31a7	421	.close = evtchn_fifo_close,
1fe56551 DV	422	};
1fe56551 DV	423
c12784c3	424	static int evtchn_fifo_alloc_control_block(unsigned cpu)
1fe56551	425	{
c12784c3	426	void *control_block = NULL;
1fe56551 DV	427	int ret = -ENOMEM;
1fe56551 DV	428
c12784c3	429	control_block = (void *)__get_free_page(GFP_KERNEL);
1fe56551 DV	430	if (control_block == NULL)
	431	goto error;
	432
c12784c3	433	ret = init_control_block(cpu, control_block);
1fe56551 DV	434	if (ret < 0)
	435	goto error;
	436
c12784c3	437	per_cpu(cpu_control_block, cpu) = control_block;
1fe56551 DV	438
	439	return 0;
	440
	441	error:
c12784c3	442	free_page((unsigned long)control_block);
1fe56551 DV	443	return ret;
	444	}
	445
0db6991d	446	static int evtchn_fifo_cpu_notification(struct notifier_block *self,
1fe56551 DV	447	unsigned long action,
	448	void *hcpu)
	449	{
	450	int cpu = (long)hcpu;
	451	int ret = 0;
	452
	453	switch (action) {
	454	case CPU_UP_PREPARE:
	455	if (!per_cpu(cpu_control_block, cpu))
c12784c3	456	ret = evtchn_fifo_alloc_control_block(cpu);
1fe56551 DV	457	break;
	458	default:
	459	break;
	460	}
	461	return ret < 0 ? NOTIFY_BAD : NOTIFY_OK;
	462	}
	463
0db6991d	464	static struct notifier_block evtchn_fifo_cpu_notifier = {
1fe56551 DV	465	.notifier_call = evtchn_fifo_cpu_notification,
	466	};
	467
	468	int __init xen_evtchn_fifo_init(void)
	469	{
	470	int cpu = get_cpu();
	471	int ret;
	472
c12784c3	473	ret = evtchn_fifo_alloc_control_block(cpu);
1fe56551 DV	474	if (ret < 0)
	475	goto out;
	476
	477	pr_info("Using FIFO-based ABI\n");
	478
	479	evtchn_ops = &evtchn_ops_fifo;
	480
	481	register_cpu_notifier(&evtchn_fifo_cpu_notifier);
	482	out:
	483	put_cpu();
	484	return ret;
	485	}