irq_xen.c

Go to the documentation of this file.

/******************************************************************************
 * 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_vector fail vector=%d\n",
               __func__, vector);
}


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;
#define NAME_SIZE   15
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);
#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
#include <linux/sched.h>

/* 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 irqreturn_t
xen_resched_handler(int irq, void *dev_id)
{
    scheduler_ipi();
    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_resched_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)
{
    int irq = 0;

    if (xen_slab_ready) {
        switch (vec) {
        case IA64_TIMER_VECTOR:
            snprintf(per_cpu(xen_timer_name, cpu),
                 sizeof(per_cpu(xen_timer_name, cpu)),
                 "%s%d", action->name, cpu);
            irq = bind_virq_to_irqhandler(VIRQ_ITC, cpu,
                action->handler, action->flags,
                per_cpu(xen_timer_name, cpu), action->dev_id);
            per_cpu(xen_timer_irq, cpu) = irq;
            break;
        case IA64_IPI_RESCHEDULE:
            snprintf(per_cpu(xen_resched_name, cpu),
                 sizeof(per_cpu(xen_resched_name, cpu)),
                 "%s%d", action->name, cpu);
            irq = bind_ipi_to_irqhandler(XEN_RESCHEDULE_VECTOR, cpu,
                action->handler, action->flags,
                per_cpu(xen_resched_name, cpu), action->dev_id);
            per_cpu(xen_resched_irq, cpu) = irq;
            break;
        case IA64_IPI_VECTOR:
            snprintf(per_cpu(xen_ipi_name, cpu),
                 sizeof(per_cpu(xen_ipi_name, cpu)),
                 "%s%d", action->name, cpu);
            irq = bind_ipi_to_irqhandler(XEN_IPI_VECTOR, cpu,
                action->handler, action->flags,
                per_cpu(xen_ipi_name, cpu), action->dev_id);
            per_cpu(xen_ipi_irq, cpu) = irq;
            break;
        case IA64_CMC_VECTOR:
            snprintf(per_cpu(xen_cmc_name, cpu),
                 sizeof(per_cpu(xen_cmc_name, cpu)),
                 "%s%d", action->name, cpu);
            irq = bind_virq_to_irqhandler(VIRQ_MCA_CMC, cpu,
                        action->handler,
                        action->flags,
                        per_cpu(xen_cmc_name, cpu),
                        action->dev_id);
            per_cpu(xen_cmc_irq, cpu) = irq;
            break;
        case IA64_CMCP_VECTOR:
            snprintf(per_cpu(xen_cmcp_name, cpu),
                 sizeof(per_cpu(xen_cmcp_name, cpu)),
                 "%s%d", action->name, cpu);
            irq = bind_ipi_to_irqhandler(XEN_CMCP_VECTOR, cpu,
                        action->handler,
                        action->flags,
                        per_cpu(xen_cmcp_name, cpu),
                        action->dev_id);
            per_cpu(xen_cmcp_irq, cpu) = irq;
            break;
        case IA64_CPEP_VECTOR:
            snprintf(per_cpu(xen_cpep_name, cpu),
                 sizeof(per_cpu(xen_cpep_name, cpu)),
                 "%s%d", action->name, cpu);
            irq = bind_ipi_to_irqhandler(XEN_CPEP_VECTOR, cpu,
                        action->handler,
                        action->flags,
                        per_cpu(xen_cpep_name, cpu),
                        action->dev_id);
            per_cpu(xen_cpep_irq, cpu) = irq;
            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.
             */
            irq_set_status_flags(irq, 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.  */
        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;
        }
        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;
        }
        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;
        }
        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;
        }
        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;
        }
        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;
        }
    }
    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 __initconst = {
    .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,
};