irq_xen.c 11.6 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65
/******************************************************************************
 * arch/ia64/xen/irq_xen.c
 *
 * Copyright (c) 2008 Isaku Yamahata <yamahata at valinux co jp>
 *                    VA Linux Systems Japan K.K.
 *
 * This program 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.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 *
 */

#include <linux/cpu.h>

#include <xen/interface/xen.h>
#include <xen/interface/callback.h>
#include <xen/events.h>

#include <asm/xen/privop.h>

#include "irq_xen.h"

/***************************************************************************
 * pv_irq_ops
 * irq operations
 */

static int
xen_assign_irq_vector(int irq)
{
	struct physdev_irq irq_op;

	irq_op.irq = irq;
	if (HYPERVISOR_physdev_op(PHYSDEVOP_alloc_irq_vector, &irq_op))
		return -ENOSPC;

	return irq_op.vector;
}

static void
xen_free_irq_vector(int vector)
{
	struct physdev_irq irq_op;

	if (vector < IA64_FIRST_DEVICE_VECTOR ||
	    vector > IA64_LAST_DEVICE_VECTOR)
		return;

	irq_op.vector = vector;
	if (HYPERVISOR_physdev_op(PHYSDEVOP_free_irq_vector, &irq_op))
		printk(KERN_WARNING "%s: xen_free_irq_vecotr fail vector=%d\n",
		       __func__, vector);
}


66 67 68 69 70 71
static DEFINE_PER_CPU(int, xen_timer_irq) = -1;
static DEFINE_PER_CPU(int, xen_ipi_irq) = -1;
static DEFINE_PER_CPU(int, xen_resched_irq) = -1;
static DEFINE_PER_CPU(int, xen_cmc_irq) = -1;
static DEFINE_PER_CPU(int, xen_cmcp_irq) = -1;
static DEFINE_PER_CPU(int, xen_cpep_irq) = -1;
72
#define NAME_SIZE	15
73 74 75 76 77 78
static DEFINE_PER_CPU(char[NAME_SIZE], xen_timer_name);
static DEFINE_PER_CPU(char[NAME_SIZE], xen_ipi_name);
static DEFINE_PER_CPU(char[NAME_SIZE], xen_resched_name);
static DEFINE_PER_CPU(char[NAME_SIZE], xen_cmc_name);
static DEFINE_PER_CPU(char[NAME_SIZE], xen_cmcp_name);
static DEFINE_PER_CPU(char[NAME_SIZE], xen_cpep_name);
79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 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
#undef NAME_SIZE

struct saved_irq {
	unsigned int irq;
	struct irqaction *action;
};
/* 16 should be far optimistic value, since only several percpu irqs
 * are registered early.
 */
#define MAX_LATE_IRQ	16
static struct saved_irq saved_percpu_irqs[MAX_LATE_IRQ];
static unsigned short late_irq_cnt;
static unsigned short saved_irq_cnt;
static int xen_slab_ready;

#ifdef CONFIG_SMP
/* Dummy stub. Though we may check XEN_RESCHEDULE_VECTOR before __do_IRQ,
 * it ends up to issue several memory accesses upon percpu data and
 * thus adds unnecessary traffic to other paths.
 */
static irqreturn_t
xen_dummy_handler(int irq, void *dev_id)
{

	return IRQ_HANDLED;
}

static struct irqaction xen_ipi_irqaction = {
	.handler =	handle_IPI,
	.flags =	IRQF_DISABLED,
	.name =		"IPI"
};

static struct irqaction xen_resched_irqaction = {
	.handler =	xen_dummy_handler,
	.flags =	IRQF_DISABLED,
	.name =		"resched"
};

static struct irqaction xen_tlb_irqaction = {
	.handler =	xen_dummy_handler,
	.flags =	IRQF_DISABLED,
	.name =		"tlb_flush"
};
#endif

/*
 * This is xen version percpu irq registration, which needs bind
 * to xen specific evtchn sub-system. One trick here is that xen
 * evtchn binding interface depends on kmalloc because related
 * port needs to be freed at device/cpu down. So we cache the
 * registration on BSP before slab is ready and then deal them
 * at later point. For rest instances happening after slab ready,
 * we hook them to xen evtchn immediately.
 *
 * FIXME: MCA is not supported by far, and thus "nomca" boot param is
 * required.
 */
static void
__xen_register_percpu_irq(unsigned int cpu, unsigned int vec,
			struct irqaction *action, int save)
{
141
	struct irq_desc *desc;
142 143 144 145 146
	int irq = 0;

	if (xen_slab_ready) {
		switch (vec) {
		case IA64_TIMER_VECTOR:
147 148
			snprintf(per_cpu(xen_timer_name, cpu),
				 sizeof(per_cpu(xen_timer_name, cpu)),
149 150 151
				 "%s%d", action->name, cpu);
			irq = bind_virq_to_irqhandler(VIRQ_ITC, cpu,
				action->handler, action->flags,
152 153
				per_cpu(xen_timer_name, cpu), action->dev_id);
			per_cpu(xen_timer_irq, cpu) = irq;
154 155
			break;
		case IA64_IPI_RESCHEDULE:
156 157
			snprintf(per_cpu(xen_resched_name, cpu),
				 sizeof(per_cpu(xen_resched_name, cpu)),
158 159 160
				 "%s%d", action->name, cpu);
			irq = bind_ipi_to_irqhandler(XEN_RESCHEDULE_VECTOR, cpu,
				action->handler, action->flags,
161 162
				per_cpu(xen_resched_name, cpu), action->dev_id);
			per_cpu(xen_resched_irq, cpu) = irq;
163 164
			break;
		case IA64_IPI_VECTOR:
165 166
			snprintf(per_cpu(xen_ipi_name, cpu),
				 sizeof(per_cpu(xen_ipi_name, cpu)),
167 168 169
				 "%s%d", action->name, cpu);
			irq = bind_ipi_to_irqhandler(XEN_IPI_VECTOR, cpu,
				action->handler, action->flags,
170 171
				per_cpu(xen_ipi_name, cpu), action->dev_id);
			per_cpu(xen_ipi_irq, cpu) = irq;
172 173
			break;
		case IA64_CMC_VECTOR:
174 175
			snprintf(per_cpu(xen_cmc_name, cpu),
				 sizeof(per_cpu(xen_cmc_name, cpu)),
176 177
				 "%s%d", action->name, cpu);
			irq = bind_virq_to_irqhandler(VIRQ_MCA_CMC, cpu,
178 179 180 181 182
						action->handler,
						action->flags,
						per_cpu(xen_cmc_name, cpu),
						action->dev_id);
			per_cpu(xen_cmc_irq, cpu) = irq;
183 184
			break;
		case IA64_CMCP_VECTOR:
185 186
			snprintf(per_cpu(xen_cmcp_name, cpu),
				 sizeof(per_cpu(xen_cmcp_name, cpu)),
187 188
				 "%s%d", action->name, cpu);
			irq = bind_ipi_to_irqhandler(XEN_CMCP_VECTOR, cpu,
189 190 191 192 193
						action->handler,
						action->flags,
						per_cpu(xen_cmcp_name, cpu),
						action->dev_id);
			per_cpu(xen_cmcp_irq, cpu) = irq;
194 195
			break;
		case IA64_CPEP_VECTOR:
196 197
			snprintf(per_cpu(xen_cpep_name, cpu),
				 sizeof(per_cpu(xen_cpep_name, cpu)),
198 199
				 "%s%d", action->name, cpu);
			irq = bind_ipi_to_irqhandler(XEN_CPEP_VECTOR, cpu,
200 201 202 203 204
						action->handler,
						action->flags,
						per_cpu(xen_cpep_name, cpu),
						action->dev_id);
			per_cpu(xen_cpep_irq, cpu) = irq;
205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277
			break;
		case IA64_CPE_VECTOR:
		case IA64_MCA_RENDEZ_VECTOR:
		case IA64_PERFMON_VECTOR:
		case IA64_MCA_WAKEUP_VECTOR:
		case IA64_SPURIOUS_INT_VECTOR:
			/* No need to complain, these aren't supported. */
			break;
		default:
			printk(KERN_WARNING "Percpu irq %d is unsupported "
			       "by xen!\n", vec);
			break;
		}
		BUG_ON(irq < 0);

		if (irq > 0) {
			/*
			 * Mark percpu.  Without this, migrate_irqs() will
			 * mark the interrupt for migrations and trigger it
			 * on cpu hotplug.
			 */
			desc = irq_desc + irq;
			desc->status |= IRQ_PER_CPU;
		}
	}

	/* For BSP, we cache registered percpu irqs, and then re-walk
	 * them when initializing APs
	 */
	if (!cpu && save) {
		BUG_ON(saved_irq_cnt == MAX_LATE_IRQ);
		saved_percpu_irqs[saved_irq_cnt].irq = vec;
		saved_percpu_irqs[saved_irq_cnt].action = action;
		saved_irq_cnt++;
		if (!xen_slab_ready)
			late_irq_cnt++;
	}
}

static void
xen_register_percpu_irq(ia64_vector vec, struct irqaction *action)
{
	__xen_register_percpu_irq(smp_processor_id(), vec, action, 1);
}

static void
xen_bind_early_percpu_irq(void)
{
	int i;

	xen_slab_ready = 1;
	/* There's no race when accessing this cached array, since only
	 * BSP will face with such step shortly
	 */
	for (i = 0; i < late_irq_cnt; i++)
		__xen_register_percpu_irq(smp_processor_id(),
					  saved_percpu_irqs[i].irq,
					  saved_percpu_irqs[i].action, 0);
}

/* FIXME: There's no obvious point to check whether slab is ready. So
 * a hack is used here by utilizing a late time hook.
 */

#ifdef CONFIG_HOTPLUG_CPU
static int __devinit
unbind_evtchn_callback(struct notifier_block *nfb,
		       unsigned long action, void *hcpu)
{
	unsigned int cpu = (unsigned long)hcpu;

	if (action == CPU_DEAD) {
		/* Unregister evtchn.  */
278 279 280 281
		if (per_cpu(xen_cpep_irq, cpu) >= 0) {
			unbind_from_irqhandler(per_cpu(xen_cpep_irq, cpu),
					       NULL);
			per_cpu(xen_cpep_irq, cpu) = -1;
282
		}
283 284 285 286
		if (per_cpu(xen_cmcp_irq, cpu) >= 0) {
			unbind_from_irqhandler(per_cpu(xen_cmcp_irq, cpu),
					       NULL);
			per_cpu(xen_cmcp_irq, cpu) = -1;
287
		}
288 289 290
		if (per_cpu(xen_cmc_irq, cpu) >= 0) {
			unbind_from_irqhandler(per_cpu(xen_cmc_irq, cpu), NULL);
			per_cpu(xen_cmc_irq, cpu) = -1;
291
		}
292 293 294
		if (per_cpu(xen_ipi_irq, cpu) >= 0) {
			unbind_from_irqhandler(per_cpu(xen_ipi_irq, cpu), NULL);
			per_cpu(xen_ipi_irq, cpu) = -1;
295
		}
296 297 298 299
		if (per_cpu(xen_resched_irq, cpu) >= 0) {
			unbind_from_irqhandler(per_cpu(xen_resched_irq, cpu),
					       NULL);
			per_cpu(xen_resched_irq, cpu) = -1;
300
		}
301 302 303 304
		if (per_cpu(xen_timer_irq, cpu) >= 0) {
			unbind_from_irqhandler(per_cpu(xen_timer_irq, cpu),
					       NULL);
			per_cpu(xen_timer_irq, cpu) = -1;
305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438
		}
	}
	return NOTIFY_OK;
}

static struct notifier_block unbind_evtchn_notifier = {
	.notifier_call = unbind_evtchn_callback,
	.priority = 0
};
#endif

void xen_smp_intr_init_early(unsigned int cpu)
{
#ifdef CONFIG_SMP
	unsigned int i;

	for (i = 0; i < saved_irq_cnt; i++)
		__xen_register_percpu_irq(cpu, saved_percpu_irqs[i].irq,
					  saved_percpu_irqs[i].action, 0);
#endif
}

void xen_smp_intr_init(void)
{
#ifdef CONFIG_SMP
	unsigned int cpu = smp_processor_id();
	struct callback_register event = {
		.type = CALLBACKTYPE_event,
		.address = { .ip = (unsigned long)&xen_event_callback },
	};

	if (cpu == 0) {
		/* Initialization was already done for boot cpu.  */
#ifdef CONFIG_HOTPLUG_CPU
		/* Register the notifier only once.  */
		register_cpu_notifier(&unbind_evtchn_notifier);
#endif
		return;
	}

	/* This should be piggyback when setup vcpu guest context */
	BUG_ON(HYPERVISOR_callback_op(CALLBACKOP_register, &event));
#endif /* CONFIG_SMP */
}

void __init
xen_irq_init(void)
{
	struct callback_register event = {
		.type = CALLBACKTYPE_event,
		.address = { .ip = (unsigned long)&xen_event_callback },
	};

	xen_init_IRQ();
	BUG_ON(HYPERVISOR_callback_op(CALLBACKOP_register, &event));
	late_time_init = xen_bind_early_percpu_irq;
}

void
xen_platform_send_ipi(int cpu, int vector, int delivery_mode, int redirect)
{
#ifdef CONFIG_SMP
	/* TODO: we need to call vcpu_up here */
	if (unlikely(vector == ap_wakeup_vector)) {
		/* XXX
		 * This should be in __cpu_up(cpu) in ia64 smpboot.c
		 * like x86. But don't want to modify it,
		 * keep it untouched.
		 */
		xen_smp_intr_init_early(cpu);

		xen_send_ipi(cpu, vector);
		/* vcpu_prepare_and_up(cpu); */
		return;
	}
#endif

	switch (vector) {
	case IA64_IPI_VECTOR:
		xen_send_IPI_one(cpu, XEN_IPI_VECTOR);
		break;
	case IA64_IPI_RESCHEDULE:
		xen_send_IPI_one(cpu, XEN_RESCHEDULE_VECTOR);
		break;
	case IA64_CMCP_VECTOR:
		xen_send_IPI_one(cpu, XEN_CMCP_VECTOR);
		break;
	case IA64_CPEP_VECTOR:
		xen_send_IPI_one(cpu, XEN_CPEP_VECTOR);
		break;
	case IA64_TIMER_VECTOR: {
		/* this is used only once by check_sal_cache_flush()
		   at boot time */
		static int used = 0;
		if (!used) {
			xen_send_ipi(cpu, IA64_TIMER_VECTOR);
			used = 1;
			break;
		}
		/* fallthrough */
	}
	default:
		printk(KERN_WARNING "Unsupported IPI type 0x%x\n",
		       vector);
		notify_remote_via_irq(0); /* defaults to 0 irq */
		break;
	}
}

static void __init
xen_register_ipi(void)
{
#ifdef CONFIG_SMP
	register_percpu_irq(IA64_IPI_VECTOR, &xen_ipi_irqaction);
	register_percpu_irq(IA64_IPI_RESCHEDULE, &xen_resched_irqaction);
	register_percpu_irq(IA64_IPI_LOCAL_TLB_FLUSH, &xen_tlb_irqaction);
#endif
}

static void
xen_resend_irq(unsigned int vector)
{
	(void)resend_irq_on_evtchn(vector);
}

const struct pv_irq_ops xen_irq_ops __initdata = {
	.register_ipi = xen_register_ipi,

	.assign_irq_vector = xen_assign_irq_vector,
	.free_irq_vector = xen_free_irq_vector,
	.register_percpu_irq = xen_register_percpu_irq,

	.resend_irq = xen_resend_irq,
};