Initial commit; kernel source import

arch/ia64/xen/Kconfig

@@ -0,0 +1,25 @@
+#
+# This Kconfig describes xen/ia64 options
+#
+
+config XEN
+	bool "Xen hypervisor support"
+	default y
+	depends on PARAVIRT && MCKINLEY && IA64_PAGE_SIZE_16KB
+	select XEN_XENCOMM
+	select NO_IDLE_HZ
+	# followings are required to save/restore.
+	select ARCH_SUSPEND_POSSIBLE
+	select SUSPEND
+	select PM_SLEEP
+	help
+	  Enable Xen hypervisor support.  Resulting kernel runs
+	  both as a guest OS on Xen and natively on hardware.
+
+config XEN_XENCOMM
+	depends on XEN
+	bool
+
+config NO_IDLE_HZ
+	depends on XEN
+	bool

arch/ia64/xen/Makefile

@@ -0,0 +1,37 @@
+#
+# Makefile for Xen components
+#
+
+obj-y := hypercall.o xenivt.o xensetup.o xen_pv_ops.o irq_xen.o \
+	 hypervisor.o xencomm.o xcom_hcall.o grant-table.o time.o suspend.o \
+	 gate-data.o
+
+obj-$(CONFIG_IA64_GENERIC) += machvec.o
+
+# The gate DSO image is built using a special linker script.
+include $(srctree)/arch/ia64/kernel/Makefile.gate
+
+# tell compiled for xen
+CPPFLAGS_gate.lds += -D__IA64_GATE_PARAVIRTUALIZED_XEN
+AFLAGS_gate.o += -D__IA64_ASM_PARAVIRTUALIZED_XEN -D__IA64_GATE_PARAVIRTUALIZED_XEN
+
+# use same file of native.
+$(obj)/gate.o: $(src)/../kernel/gate.S FORCE
+	$(call if_changed_dep,as_o_S)
+$(obj)/gate.lds: $(src)/../kernel/gate.lds.S FORCE
+	$(call if_changed_dep,cpp_lds_S)
+
+
+AFLAGS_xenivt.o += -D__IA64_ASM_PARAVIRTUALIZED_XEN
+
+# xen multi compile
+ASM_PARAVIRT_MULTI_COMPILE_SRCS = ivt.S entry.S fsys.S
+ASM_PARAVIRT_OBJS = $(addprefix xen-,$(ASM_PARAVIRT_MULTI_COMPILE_SRCS:.S=.o))
+obj-y += $(ASM_PARAVIRT_OBJS)
+define paravirtualized_xen
+AFLAGS_$(1) += -D__IA64_ASM_PARAVIRTUALIZED_XEN
+endef
+$(foreach o,$(ASM_PARAVIRT_OBJS),$(eval $(call paravirtualized_xen,$(o))))
+
+$(obj)/xen-%.o: $(src)/../kernel/%.S FORCE
+	$(call if_changed_dep,as_o_S)

arch/ia64/xen/gate-data.S

@@ -0,0 +1,3 @@
+	.section .data..gate.xen, "aw"
+
+	.incbin "arch/ia64/xen/gate.so"

arch/ia64/xen/grant-table.c

@@ -0,0 +1,94 @@
+/******************************************************************************
+ * arch/ia64/xen/grant-table.c
+ *
+ * Copyright (c) 2006 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/module.h>
+#include <linux/vmalloc.h>
+#include <linux/slab.h>
+#include <linux/mm.h>
+
+#include <xen/interface/xen.h>
+#include <xen/interface/memory.h>
+#include <xen/grant_table.h>
+
+#include <asm/xen/hypervisor.h>
+
+/****************************************************************************
+ * grant table hack
+ * cmd: GNTTABOP_xxx
+ */
+
+int arch_gnttab_map_shared(unsigned long *frames, unsigned long nr_gframes,
+			   unsigned long max_nr_gframes,
+			   struct grant_entry **__shared)
+{
+	*__shared = __va(frames[0] << PAGE_SHIFT);
+	return 0;
+}
+
+void arch_gnttab_unmap_shared(struct grant_entry *shared,
+			      unsigned long nr_gframes)
+{
+	/* nothing */
+}
+
+static void
+gnttab_map_grant_ref_pre(struct gnttab_map_grant_ref *uop)
+{
+	uint32_t flags;
+
+	flags = uop->flags;
+
+	if (flags & GNTMAP_host_map) {
+		if (flags & GNTMAP_application_map) {
+			printk(KERN_DEBUG
+			       "GNTMAP_application_map is not supported yet: "
+			       "flags 0x%x\n", flags);
+			BUG();
+		}
+		if (flags & GNTMAP_contains_pte) {
+			printk(KERN_DEBUG
+			       "GNTMAP_contains_pte is not supported yet: "
+			       "flags 0x%x\n", flags);
+			BUG();
+		}
+	} else if (flags & GNTMAP_device_map) {
+		printk("GNTMAP_device_map is not supported yet 0x%x\n", flags);
+		BUG();	/* not yet. actually this flag is not used. */
+	} else {
+		BUG();
+	}
+}
+
+int
+HYPERVISOR_grant_table_op(unsigned int cmd, void *uop, unsigned int count)
+{
+	if (cmd == GNTTABOP_map_grant_ref) {
+		unsigned int i;
+		for (i = 0; i < count; i++) {
+			gnttab_map_grant_ref_pre(
+				(struct gnttab_map_grant_ref *)uop + i);
+		}
+	}
+	return xencomm_hypercall_grant_table_op(cmd, uop, count);
+}
+
+EXPORT_SYMBOL(HYPERVISOR_grant_table_op);

arch/ia64/xen/hypercall.S

@@ -0,0 +1,88 @@
+/*
+ * Support routines for Xen hypercalls
+ *
+ * Copyright (C) 2005 Dan Magenheimer <dan.magenheimer@hp.com>
+ * Copyright (C) 2008 Yaozu (Eddie) Dong <eddie.dong@intel.com>
+ */
+
+#include <asm/asmmacro.h>
+#include <asm/intrinsics.h>
+#include <asm/xen/privop.h>
+
+#ifdef __INTEL_COMPILER
+/*
+ * Hypercalls without parameter.
+ */
+#define __HCALL0(name,hcall)		\
+	GLOBAL_ENTRY(name);		\
+	break	hcall;			\
+	br.ret.sptk.many rp;		\
+	END(name)
+
+/*
+ * Hypercalls with 1 parameter.
+ */
+#define __HCALL1(name,hcall)		\
+	GLOBAL_ENTRY(name);		\
+	mov r8=r32;			\
+	break	hcall;			\
+	br.ret.sptk.many rp;		\
+	END(name)
+
+/*
+ * Hypercalls with 2 parameters.
+ */
+#define __HCALL2(name,hcall)		\
+	GLOBAL_ENTRY(name);		\
+	mov r8=r32;			\
+	mov r9=r33;			\
+	break	hcall;			\
+	br.ret.sptk.many rp;		\
+	END(name)
+
+__HCALL0(xen_get_psr, HYPERPRIVOP_GET_PSR)
+__HCALL0(xen_get_ivr, HYPERPRIVOP_GET_IVR)
+__HCALL0(xen_get_tpr, HYPERPRIVOP_GET_TPR)
+__HCALL0(xen_hyper_ssm_i, HYPERPRIVOP_SSM_I)
+
+__HCALL1(xen_set_tpr, HYPERPRIVOP_SET_TPR)
+__HCALL1(xen_eoi, HYPERPRIVOP_EOI)
+__HCALL1(xen_thash, HYPERPRIVOP_THASH)
+__HCALL1(xen_set_itm, HYPERPRIVOP_SET_ITM)
+__HCALL1(xen_get_rr, HYPERPRIVOP_GET_RR)
+__HCALL1(xen_fc, HYPERPRIVOP_FC)
+__HCALL1(xen_get_cpuid, HYPERPRIVOP_GET_CPUID)
+__HCALL1(xen_get_pmd, HYPERPRIVOP_GET_PMD)
+
+__HCALL2(xen_ptcga, HYPERPRIVOP_PTC_GA)
+__HCALL2(xen_set_rr, HYPERPRIVOP_SET_RR)
+__HCALL2(xen_set_kr, HYPERPRIVOP_SET_KR)
+
+GLOBAL_ENTRY(xen_set_rr0_to_rr4)
+	mov r8=r32
+	mov r9=r33
+	mov r10=r34
+	mov r11=r35
+	mov r14=r36
+	XEN_HYPER_SET_RR0_TO_RR4
+	br.ret.sptk.many rp
+	;;
+END(xen_set_rr0_to_rr4)
+#endif
+
+GLOBAL_ENTRY(xen_send_ipi)
+	mov r14=r32
+	mov r15=r33
+	mov r2=0x400
+	break 0x1000
+	;;
+	br.ret.sptk.many rp
+	;;
+END(xen_send_ipi)
+
+GLOBAL_ENTRY(__hypercall)
+	mov r2=r37
+	break 0x1000
+	br.ret.sptk.many b0
+	;;
+END(__hypercall)

arch/ia64/xen/hypervisor.c

@@ -0,0 +1,97 @@
+/******************************************************************************
+ * arch/ia64/xen/hypervisor.c
+ *
+ * Copyright (c) 2006 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/efi.h>
+#include <linux/export.h>
+#include <asm/xen/hypervisor.h>
+#include <asm/xen/privop.h>
+
+#include "irq_xen.h"
+
+struct shared_info *HYPERVISOR_shared_info __read_mostly =
+	(struct shared_info *)XSI_BASE;
+EXPORT_SYMBOL(HYPERVISOR_shared_info);
+
+DEFINE_PER_CPU(struct vcpu_info *, xen_vcpu);
+
+struct start_info *xen_start_info;
+EXPORT_SYMBOL(xen_start_info);
+
+EXPORT_SYMBOL(xen_domain_type);
+
+EXPORT_SYMBOL(__hypercall);
+
+/* Stolen from arch/x86/xen/enlighten.c */
+/*
+ * Flag to determine whether vcpu info placement is available on all
+ * VCPUs.  We assume it is to start with, and then set it to zero on
+ * the first failure.  This is because it can succeed on some VCPUs
+ * and not others, since it can involve hypervisor memory allocation,
+ * or because the guest failed to guarantee all the appropriate
+ * constraints on all VCPUs (ie buffer can't cross a page boundary).
+ *
+ * Note that any particular CPU may be using a placed vcpu structure,
+ * but we can only optimise if the all are.
+ *
+ * 0: not available, 1: available
+ */
+
+static void __init xen_vcpu_setup(int cpu)
+{
+	/*
+	 * WARNING:
+	 * before changing MAX_VIRT_CPUS,
+	 * check that shared_info fits on a page
+	 */
+	BUILD_BUG_ON(sizeof(struct shared_info) > PAGE_SIZE);
+	per_cpu(xen_vcpu, cpu) = &HYPERVISOR_shared_info->vcpu_info[cpu];
+}
+
+void __init xen_setup_vcpu_info_placement(void)
+{
+	int cpu;
+
+	for_each_possible_cpu(cpu)
+		xen_vcpu_setup(cpu);
+}
+
+void __cpuinit
+xen_cpu_init(void)
+{
+	xen_smp_intr_init();
+}
+
+/**************************************************************************
+ * opt feature
+ */
+void
+xen_ia64_enable_opt_feature(void)
+{
+	/* Enable region 7 identity map optimizations in Xen */
+	struct xen_ia64_opt_feature optf;
+
+	optf.cmd = XEN_IA64_OPTF_IDENT_MAP_REG7;
+	optf.on = XEN_IA64_OPTF_ON;
+	optf.pgprot = pgprot_val(PAGE_KERNEL);
+	optf.key = 0;	/* No key on linux. */
+	HYPERVISOR_opt_feature(&optf);
+}

arch/ia64/xen/irq_xen.c

@@ -0,0 +1,443 @@
+/******************************************************************************
+ * 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 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,
+};

arch/ia64/xen/irq_xen.h

@@ -0,0 +1,34 @@
+/******************************************************************************
+ * arch/ia64/xen/irq_xen.h
+ *
+ * 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
+ *
+ */
+
+#ifndef IRQ_XEN_H
+#define IRQ_XEN_H
+
+extern void (*late_time_init)(void);
+extern char xen_event_callback;
+void __init xen_init_IRQ(void);
+
+extern const struct pv_irq_ops xen_irq_ops __initconst;
+extern void xen_smp_intr_init(void);
+extern void xen_send_ipi(int cpu, int vec);
+
+#endif /* IRQ_XEN_H */

arch/ia64/xen/machvec.c

@@ -0,0 +1,4 @@
+#define MACHVEC_PLATFORM_NAME           xen
+#define MACHVEC_PLATFORM_HEADER         <asm/machvec_xen.h>
+#include <asm/machvec_init.h>
+

arch/ia64/xen/suspend.c

@@ -0,0 +1,59 @@
+/******************************************************************************
+ * arch/ia64/xen/suspend.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
+ *
+ * suspend/resume
+ */
+
+#include <xen/xen-ops.h>
+#include <asm/xen/hypervisor.h>
+#include "time.h"
+
+void
+xen_mm_pin_all(void)
+{
+	/* nothing */
+}
+
+void
+xen_mm_unpin_all(void)
+{
+	/* nothing */
+}
+
+void
+xen_arch_pre_suspend()
+{
+	/* nothing */
+}
+
+void
+xen_arch_post_suspend(int suspend_cancelled)
+{
+	if (suspend_cancelled)
+		return;
+
+	xen_ia64_enable_opt_feature();
+	/* add more if necessary */
+}
+
+void xen_arch_resume(void)
+{
+	xen_timer_resume_on_aps();
+}

arch/ia64/xen/time.c

@@ -0,0 +1,257 @@
+/******************************************************************************
+ * arch/ia64/xen/time.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/delay.h>
+#include <linux/kernel_stat.h>
+#include <linux/posix-timers.h>
+#include <linux/irq.h>
+#include <linux/clocksource.h>
+
+#include <asm/timex.h>
+
+#include <asm/xen/hypervisor.h>
+
+#include <xen/interface/vcpu.h>
+
+#include "../kernel/fsyscall_gtod_data.h"
+
+static DEFINE_PER_CPU(struct vcpu_runstate_info, xen_runstate);
+static DEFINE_PER_CPU(unsigned long, xen_stolen_time);
+static DEFINE_PER_CPU(unsigned long, xen_blocked_time);
+
+/* taken from i386/kernel/time-xen.c */
+static void xen_init_missing_ticks_accounting(int cpu)
+{
+	struct vcpu_register_runstate_memory_area area;
+	struct vcpu_runstate_info *runstate = &per_cpu(xen_runstate, cpu);
+	int rc;
+
+	memset(runstate, 0, sizeof(*runstate));
+
+	area.addr.v = runstate;
+	rc = HYPERVISOR_vcpu_op(VCPUOP_register_runstate_memory_area, cpu,
+				&area);
+	WARN_ON(rc && rc != -ENOSYS);
+
+	per_cpu(xen_blocked_time, cpu) = runstate->time[RUNSTATE_blocked];
+	per_cpu(xen_stolen_time, cpu) = runstate->time[RUNSTATE_runnable]
+					    + runstate->time[RUNSTATE_offline];
+}
+
+/*
+ * Runstate accounting
+ */
+/* stolen from arch/x86/xen/time.c */
+static void get_runstate_snapshot(struct vcpu_runstate_info *res)
+{
+	u64 state_time;
+	struct vcpu_runstate_info *state;
+
+	BUG_ON(preemptible());
+
+	state = &__get_cpu_var(xen_runstate);
+
+	/*
+	 * The runstate info is always updated by the hypervisor on
+	 * the current CPU, so there's no need to use anything
+	 * stronger than a compiler barrier when fetching it.
+	 */
+	do {
+		state_time = state->state_entry_time;
+		rmb();
+		*res = *state;
+		rmb();
+	} while (state->state_entry_time != state_time);
+}
+
+#define NS_PER_TICK (1000000000LL/HZ)
+
+static unsigned long
+consider_steal_time(unsigned long new_itm)
+{
+	unsigned long stolen, blocked;
+	unsigned long delta_itm = 0, stolentick = 0;
+	int cpu = smp_processor_id();
+	struct vcpu_runstate_info runstate;
+	struct task_struct *p = current;
+
+	get_runstate_snapshot(&runstate);
+
+	/*
+	 * Check for vcpu migration effect
+	 * In this case, itc value is reversed.
+	 * This causes huge stolen value.
+	 * This function just checks and reject this effect.
+	 */
+	if (!time_after_eq(runstate.time[RUNSTATE_blocked],
+			   per_cpu(xen_blocked_time, cpu)))
+		blocked = 0;
+
+	if (!time_after_eq(runstate.time[RUNSTATE_runnable] +
+			   runstate.time[RUNSTATE_offline],
+			   per_cpu(xen_stolen_time, cpu)))
+		stolen = 0;
+
+	if (!time_after(delta_itm + new_itm, ia64_get_itc()))
+		stolentick = ia64_get_itc() - new_itm;
+
+	do_div(stolentick, NS_PER_TICK);
+	stolentick++;
+
+	do_div(stolen, NS_PER_TICK);
+
+	if (stolen > stolentick)
+		stolen = stolentick;
+
+	stolentick -= stolen;
+	do_div(blocked, NS_PER_TICK);
+
+	if (blocked > stolentick)
+		blocked = stolentick;
+
+	if (stolen > 0 || blocked > 0) {
+		account_steal_ticks(stolen);
+		account_idle_ticks(blocked);
+		run_local_timers();
+
+		rcu_check_callbacks(cpu, user_mode(get_irq_regs()));
+
+		scheduler_tick();
+		run_posix_cpu_timers(p);
+		delta_itm += local_cpu_data->itm_delta * (stolen + blocked);
+
+		if (cpu == time_keeper_id)
+			xtime_update(stolen + blocked);
+
+		local_cpu_data->itm_next = delta_itm + new_itm;
+
+		per_cpu(xen_stolen_time, cpu) += NS_PER_TICK * stolen;
+		per_cpu(xen_blocked_time, cpu) += NS_PER_TICK * blocked;
+	}
+	return delta_itm;
+}
+
+static int xen_do_steal_accounting(unsigned long *new_itm)
+{
+	unsigned long delta_itm;
+	delta_itm = consider_steal_time(*new_itm);
+	*new_itm += delta_itm;
+	if (time_after(*new_itm, ia64_get_itc()) && delta_itm)
+		return 1;
+
+	return 0;
+}
+
+static void xen_itc_jitter_data_reset(void)
+{
+	u64 lcycle, ret;
+
+	do {
+		lcycle = itc_jitter_data.itc_lastcycle;
+		ret = cmpxchg(&itc_jitter_data.itc_lastcycle, lcycle, 0);
+	} while (unlikely(ret != lcycle));
+}
+
+/* based on xen_sched_clock() in arch/x86/xen/time.c. */
+/*
+ * This relies on HAVE_UNSTABLE_SCHED_CLOCK. If it can't be defined,
+ * something similar logic should be implemented here.
+ */
+/*
+ * Xen sched_clock implementation.  Returns the number of unstolen
+ * nanoseconds, which is nanoseconds the VCPU spent in RUNNING+BLOCKED
+ * states.
+ */
+static unsigned long long xen_sched_clock(void)
+{
+	struct vcpu_runstate_info runstate;
+
+	unsigned long long now;
+	unsigned long long offset;
+	unsigned long long ret;
+
+	/*
+	 * Ideally sched_clock should be called on a per-cpu basis
+	 * anyway, so preempt should already be disabled, but that's
+	 * not current practice at the moment.
+	 */
+	preempt_disable();
+
+	/*
+	 * both ia64_native_sched_clock() and xen's runstate are
+	 * based on mAR.ITC. So difference of them makes sense.
+	 */
+	now = ia64_native_sched_clock();
+
+	get_runstate_snapshot(&runstate);
+
+	WARN_ON(runstate.state != RUNSTATE_running);
+
+	offset = 0;
+	if (now > runstate.state_entry_time)
+		offset = now - runstate.state_entry_time;
+	ret = runstate.time[RUNSTATE_blocked] +
+		runstate.time[RUNSTATE_running] +
+		offset;
+
+	preempt_enable();
+
+	return ret;
+}
+
+struct pv_time_ops xen_time_ops __initdata = {
+	.init_missing_ticks_accounting	= xen_init_missing_ticks_accounting,
+	.do_steal_accounting		= xen_do_steal_accounting,
+	.clocksource_resume		= xen_itc_jitter_data_reset,
+	.sched_clock			= xen_sched_clock,
+};
+
+/* Called after suspend, to resume time.  */
+static void xen_local_tick_resume(void)
+{
+	/* Just trigger a tick.  */
+	ia64_cpu_local_tick();
+	touch_softlockup_watchdog();
+}
+
+void
+xen_timer_resume(void)
+{
+	unsigned int cpu;
+
+	xen_local_tick_resume();
+
+	for_each_online_cpu(cpu)
+		xen_init_missing_ticks_accounting(cpu);
+}
+
+static void ia64_cpu_local_tick_fn(void *unused)
+{
+	xen_local_tick_resume();
+	xen_init_missing_ticks_accounting(smp_processor_id());
+}
+
+void
+xen_timer_resume_on_aps(void)
+{
+	smp_call_function(&ia64_cpu_local_tick_fn, NULL, 1);
+}

arch/ia64/xen/time.h

@@ -0,0 +1,24 @@
+/******************************************************************************
+ * arch/ia64/xen/time.h
+ *
+ * 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
+ *
+ */
+
+extern struct pv_time_ops xen_time_ops __initdata;
+void xen_timer_resume_on_aps(void);

arch/ia64/xen/xcom_hcall.c

@@ -0,0 +1,441 @@
+/*
+ * 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, 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
+ *
+ *          Tristan Gingold <tristan.gingold@bull.net>
+ *
+ *          Copyright (c) 2007
+ *          Isaku Yamahata <yamahata at valinux co jp>
+ *                          VA Linux Systems Japan K.K.
+ *          consolidate mini and inline version.
+ */
+
+#include <linux/module.h>
+#include <xen/interface/xen.h>
+#include <xen/interface/memory.h>
+#include <xen/interface/grant_table.h>
+#include <xen/interface/callback.h>
+#include <xen/interface/vcpu.h>
+#include <asm/xen/hypervisor.h>
+#include <asm/xen/xencomm.h>
+
+/* Xencomm notes:
+ * This file defines hypercalls to be used by xencomm.  The hypercalls simply
+ * create inlines or mini descriptors for pointers and then call the raw arch
+ * hypercall xencomm_arch_hypercall_XXX
+ *
+ * If the arch wants to directly use these hypercalls, simply define macros
+ * in asm/xen/hypercall.h, eg:
+ *  #define HYPERVISOR_sched_op xencomm_hypercall_sched_op
+ *
+ * The arch may also define HYPERVISOR_xxx as a function and do more operations
+ * before/after doing the hypercall.
+ *
+ * Note: because only inline or mini descriptors are created these functions
+ * must only be called with in kernel memory parameters.
+ */
+
+int
+xencomm_hypercall_console_io(int cmd, int count, char *str)
+{
+	/* xen early printk uses console io hypercall before
+	 * xencomm initialization. In that case, we just ignore it.
+	 */
+	if (!xencomm_is_initialized())
+		return 0;
+
+	return xencomm_arch_hypercall_console_io
+		(cmd, count, xencomm_map_no_alloc(str, count));
+}
+EXPORT_SYMBOL_GPL(xencomm_hypercall_console_io);
+
+int
+xencomm_hypercall_event_channel_op(int cmd, void *op)
+{
+	struct xencomm_handle *desc;
+	desc = xencomm_map_no_alloc(op, sizeof(struct evtchn_op));
+	if (desc == NULL)
+		return -EINVAL;
+
+	return xencomm_arch_hypercall_event_channel_op(cmd, desc);
+}
+EXPORT_SYMBOL_GPL(xencomm_hypercall_event_channel_op);
+
+int
+xencomm_hypercall_xen_version(int cmd, void *arg)
+{
+	struct xencomm_handle *desc;
+	unsigned int argsize;
+
+	switch (cmd) {
+	case XENVER_version:
+		/* do not actually pass an argument */
+		return xencomm_arch_hypercall_xen_version(cmd, 0);
+	case XENVER_extraversion:
+		argsize = sizeof(struct xen_extraversion);
+		break;
+	case XENVER_compile_info:
+		argsize = sizeof(struct xen_compile_info);
+		break;
+	case XENVER_capabilities:
+		argsize = sizeof(struct xen_capabilities_info);
+		break;
+	case XENVER_changeset:
+		argsize = sizeof(struct xen_changeset_info);
+		break;
+	case XENVER_platform_parameters:
+		argsize = sizeof(struct xen_platform_parameters);
+		break;
+	case XENVER_get_features:
+		argsize = (arg == NULL) ? 0 : sizeof(struct xen_feature_info);
+		break;
+
+	default:
+		printk(KERN_DEBUG
+		       "%s: unknown version op %d\n", __func__, cmd);
+		return -ENOSYS;
+	}
+
+	desc = xencomm_map_no_alloc(arg, argsize);
+	if (desc == NULL)
+		return -EINVAL;
+
+	return xencomm_arch_hypercall_xen_version(cmd, desc);
+}
+EXPORT_SYMBOL_GPL(xencomm_hypercall_xen_version);
+
+int
+xencomm_hypercall_physdev_op(int cmd, void *op)
+{
+	unsigned int argsize;
+
+	switch (cmd) {
+	case PHYSDEVOP_apic_read:
+	case PHYSDEVOP_apic_write:
+		argsize = sizeof(struct physdev_apic);
+		break;
+	case PHYSDEVOP_alloc_irq_vector:
+	case PHYSDEVOP_free_irq_vector:
+		argsize = sizeof(struct physdev_irq);
+		break;
+	case PHYSDEVOP_irq_status_query:
+		argsize = sizeof(struct physdev_irq_status_query);
+		break;
+
+	default:
+		printk(KERN_DEBUG
+		       "%s: unknown physdev op %d\n", __func__, cmd);
+		return -ENOSYS;
+	}
+
+	return xencomm_arch_hypercall_physdev_op
+		(cmd, xencomm_map_no_alloc(op, argsize));
+}
+
+static int
+xencommize_grant_table_op(struct xencomm_mini **xc_area,
+			  unsigned int cmd, void *op, unsigned int count,
+			  struct xencomm_handle **desc)
+{
+	struct xencomm_handle *desc1;
+	unsigned int argsize;
+
+	switch (cmd) {
+	case GNTTABOP_map_grant_ref:
+		argsize = sizeof(struct gnttab_map_grant_ref);
+		break;
+	case GNTTABOP_unmap_grant_ref:
+		argsize = sizeof(struct gnttab_unmap_grant_ref);
+		break;
+	case GNTTABOP_setup_table:
+	{
+		struct gnttab_setup_table *setup = op;
+
+		argsize = sizeof(*setup);
+
+		if (count != 1)
+			return -EINVAL;
+		desc1 = __xencomm_map_no_alloc
+			(xen_guest_handle(setup->frame_list),
+			 setup->nr_frames *
+			 sizeof(*xen_guest_handle(setup->frame_list)),
+			 *xc_area);
+		if (desc1 == NULL)
+			return -EINVAL;
+		(*xc_area)++;
+		set_xen_guest_handle(setup->frame_list, (void *)desc1);
+		break;
+	}
+	case GNTTABOP_dump_table:
+		argsize = sizeof(struct gnttab_dump_table);
+		break;
+	case GNTTABOP_transfer:
+		argsize = sizeof(struct gnttab_transfer);
+		break;
+	case GNTTABOP_copy:
+		argsize = sizeof(struct gnttab_copy);
+		break;
+	case GNTTABOP_query_size:
+		argsize = sizeof(struct gnttab_query_size);
+		break;
+	default:
+		printk(KERN_DEBUG "%s: unknown hypercall grant table op %d\n",
+		       __func__, cmd);
+		BUG();
+	}
+
+	*desc = __xencomm_map_no_alloc(op, count * argsize, *xc_area);
+	if (*desc == NULL)
+		return -EINVAL;
+	(*xc_area)++;
+
+	return 0;
+}
+
+int
+xencomm_hypercall_grant_table_op(unsigned int cmd, void *op,
+				 unsigned int count)
+{
+	int rc;
+	struct xencomm_handle *desc;
+	XENCOMM_MINI_ALIGNED(xc_area, 2);
+
+	rc = xencommize_grant_table_op(&xc_area, cmd, op, count, &desc);
+	if (rc)
+		return rc;
+
+	return xencomm_arch_hypercall_grant_table_op(cmd, desc, count);
+}
+EXPORT_SYMBOL_GPL(xencomm_hypercall_grant_table_op);
+
+int
+xencomm_hypercall_sched_op(int cmd, void *arg)
+{
+	struct xencomm_handle *desc;
+	unsigned int argsize;
+
+	switch (cmd) {
+	case SCHEDOP_yield:
+	case SCHEDOP_block:
+		argsize = 0;
+		break;
+	case SCHEDOP_shutdown:
+		argsize = sizeof(struct sched_shutdown);
+		break;
+	case SCHEDOP_poll:
+	{
+		struct sched_poll *poll = arg;
+		struct xencomm_handle *ports;
+
+		argsize = sizeof(struct sched_poll);
+		ports = xencomm_map_no_alloc(xen_guest_handle(poll->ports),
+				     sizeof(*xen_guest_handle(poll->ports)));
+
+		set_xen_guest_handle(poll->ports, (void *)ports);
+		break;
+	}
+	default:
+		printk(KERN_DEBUG "%s: unknown sched op %d\n", __func__, cmd);
+		return -ENOSYS;
+	}
+
+	desc = xencomm_map_no_alloc(arg, argsize);
+	if (desc == NULL)
+		return -EINVAL;
+
+	return xencomm_arch_hypercall_sched_op(cmd, desc);
+}
+EXPORT_SYMBOL_GPL(xencomm_hypercall_sched_op);
+
+int
+xencomm_hypercall_multicall(void *call_list, int nr_calls)
+{
+	int rc;
+	int i;
+	struct multicall_entry *mce;
+	struct xencomm_handle *desc;
+	XENCOMM_MINI_ALIGNED(xc_area, nr_calls * 2);
+
+	for (i = 0; i < nr_calls; i++) {
+		mce = (struct multicall_entry *)call_list + i;
+
+		switch (mce->op) {
+		case __HYPERVISOR_update_va_mapping:
+		case __HYPERVISOR_mmu_update:
+			/* No-op on ia64.  */
+			break;
+		case __HYPERVISOR_grant_table_op:
+			rc = xencommize_grant_table_op
+				(&xc_area,
+				 mce->args[0], (void *)mce->args[1],
+				 mce->args[2], &desc);
+			if (rc)
+				return rc;
+			mce->args[1] = (unsigned long)desc;
+			break;
+		case __HYPERVISOR_memory_op:
+		default:
+			printk(KERN_DEBUG
+			       "%s: unhandled multicall op entry op %lu\n",
+			       __func__, mce->op);
+			return -ENOSYS;
+		}
+	}
+
+	desc = xencomm_map_no_alloc(call_list,
+				    nr_calls * sizeof(struct multicall_entry));
+	if (desc == NULL)
+		return -EINVAL;
+
+	return xencomm_arch_hypercall_multicall(desc, nr_calls);
+}
+EXPORT_SYMBOL_GPL(xencomm_hypercall_multicall);
+
+int
+xencomm_hypercall_callback_op(int cmd, void *arg)
+{
+	unsigned int argsize;
+	switch (cmd) {
+	case CALLBACKOP_register:
+		argsize = sizeof(struct callback_register);
+		break;
+	case CALLBACKOP_unregister:
+		argsize = sizeof(struct callback_unregister);
+		break;
+	default:
+		printk(KERN_DEBUG
+		       "%s: unknown callback op %d\n", __func__, cmd);
+		return -ENOSYS;
+	}
+
+	return xencomm_arch_hypercall_callback_op
+		(cmd, xencomm_map_no_alloc(arg, argsize));
+}
+
+static int
+xencommize_memory_reservation(struct xencomm_mini *xc_area,
+			      struct xen_memory_reservation *mop)
+{
+	struct xencomm_handle *desc;
+
+	desc = __xencomm_map_no_alloc(xen_guest_handle(mop->extent_start),
+			mop->nr_extents *
+			sizeof(*xen_guest_handle(mop->extent_start)),
+			xc_area);
+	if (desc == NULL)
+		return -EINVAL;
+
+	set_xen_guest_handle(mop->extent_start, (void *)desc);
+	return 0;
+}
+
+int
+xencomm_hypercall_memory_op(unsigned int cmd, void *arg)
+{
+	GUEST_HANDLE(xen_pfn_t) extent_start_va[2] = { {NULL}, {NULL} };
+	struct xen_memory_reservation *xmr = NULL;
+	int rc;
+	struct xencomm_handle *desc;
+	unsigned int argsize;
+	XENCOMM_MINI_ALIGNED(xc_area, 2);
+
+	switch (cmd) {
+	case XENMEM_increase_reservation:
+	case XENMEM_decrease_reservation:
+	case XENMEM_populate_physmap:
+		xmr = (struct xen_memory_reservation *)arg;
+		set_xen_guest_handle(extent_start_va[0],
+				     xen_guest_handle(xmr->extent_start));
+
+		argsize = sizeof(*xmr);
+		rc = xencommize_memory_reservation(xc_area, xmr);
+		if (rc)
+			return rc;
+		xc_area++;
+		break;
+
+	case XENMEM_maximum_ram_page:
+		argsize = 0;
+		break;
+
+	case XENMEM_add_to_physmap:
+		argsize = sizeof(struct xen_add_to_physmap);
+		break;
+
+	default:
+		printk(KERN_DEBUG "%s: unknown memory op %d\n", __func__, cmd);
+		return -ENOSYS;
+	}
+
+	desc = xencomm_map_no_alloc(arg, argsize);
+	if (desc == NULL)
+		return -EINVAL;
+
+	rc = xencomm_arch_hypercall_memory_op(cmd, desc);
+
+	switch (cmd) {
+	case XENMEM_increase_reservation:
+	case XENMEM_decrease_reservation:
+	case XENMEM_populate_physmap:
+		set_xen_guest_handle(xmr->extent_start,
+				     xen_guest_handle(extent_start_va[0]));
+		break;
+	}
+
+	return rc;
+}
+EXPORT_SYMBOL_GPL(xencomm_hypercall_memory_op);
+
+int
+xencomm_hypercall_suspend(unsigned long srec)
+{
+	struct sched_shutdown arg;
+
+	arg.reason = SHUTDOWN_suspend;
+
+	return xencomm_arch_hypercall_sched_op(
+		SCHEDOP_shutdown, xencomm_map_no_alloc(&arg, sizeof(arg)));
+}
+
+long
+xencomm_hypercall_vcpu_op(int cmd, int cpu, void *arg)
+{
+	unsigned int argsize;
+	switch (cmd) {
+	case VCPUOP_register_runstate_memory_area: {
+		struct vcpu_register_runstate_memory_area *area =
+			(struct vcpu_register_runstate_memory_area *)arg;
+		argsize = sizeof(*arg);
+		set_xen_guest_handle(area->addr.h,
+		     (void *)xencomm_map_no_alloc(area->addr.v,
+						  sizeof(area->addr.v)));
+		break;
+	}
+
+	default:
+		printk(KERN_DEBUG "%s: unknown vcpu op %d\n", __func__, cmd);
+		return -ENOSYS;
+	}
+
+	return xencomm_arch_hypercall_vcpu_op(cmd, cpu,
+					xencomm_map_no_alloc(arg, argsize));
+}
+
+long
+xencomm_hypercall_opt_feature(void *arg)
+{
+	return xencomm_arch_hypercall_opt_feature(
+		xencomm_map_no_alloc(arg,
+				     sizeof(struct xen_ia64_opt_feature)));
+}

arch/ia64/xen/xencomm.c

@@ -0,0 +1,106 @@
+/*
+ * Copyright (C) 2006 Hollis Blanchard <hollisb@us.ibm.com>, IBM Corporation
+ *
+ * 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/mm.h>
+#include <linux/err.h>
+
+static unsigned long kernel_virtual_offset;
+static int is_xencomm_initialized;
+
+/* for xen early printk. It uses console io hypercall which uses xencomm.
+ * However early printk may use it before xencomm initialization.
+ */
+int
+xencomm_is_initialized(void)
+{
+	return is_xencomm_initialized;
+}
+
+void
+xencomm_initialize(void)
+{
+	kernel_virtual_offset = KERNEL_START - ia64_tpa(KERNEL_START);
+	is_xencomm_initialized = 1;
+}
+
+/* Translate virtual address to physical address.  */
+unsigned long
+xencomm_vtop(unsigned long vaddr)
+{
+	struct page *page;
+	struct vm_area_struct *vma;
+
+	if (vaddr == 0)
+		return 0UL;
+
+	if (REGION_NUMBER(vaddr) == 5) {
+		pgd_t *pgd;
+		pud_t *pud;
+		pmd_t *pmd;
+		pte_t *ptep;
+
+		/* On ia64, TASK_SIZE refers to current.  It is not initialized
+		   during boot.
+		   Furthermore the kernel is relocatable and __pa() doesn't
+		   work on  addresses.  */
+		if (vaddr >= KERNEL_START
+		    && vaddr < (KERNEL_START + KERNEL_TR_PAGE_SIZE))
+			return vaddr - kernel_virtual_offset;
+
+		/* In kernel area -- virtually mapped.  */
+		pgd = pgd_offset_k(vaddr);
+		if (pgd_none(*pgd) || pgd_bad(*pgd))
+			return ~0UL;
+
+		pud = pud_offset(pgd, vaddr);
+		if (pud_none(*pud) || pud_bad(*pud))
+			return ~0UL;
+
+		pmd = pmd_offset(pud, vaddr);
+		if (pmd_none(*pmd) || pmd_bad(*pmd))
+			return ~0UL;
+
+		ptep = pte_offset_kernel(pmd, vaddr);
+		if (!ptep)
+			return ~0UL;
+
+		return (pte_val(*ptep) & _PFN_MASK) | (vaddr & ~PAGE_MASK);
+	}
+
+	if (vaddr > TASK_SIZE) {
+		/* percpu variables */
+		if (REGION_NUMBER(vaddr) == 7 &&
+		    REGION_OFFSET(vaddr) >= (1ULL << IA64_MAX_PHYS_BITS))
+			ia64_tpa(vaddr);
+
+		/* kernel address */
+		return __pa(vaddr);
+	}
+
+	/* XXX double-check (lack of) locking */
+	vma = find_extend_vma(current->mm, vaddr);
+	if (!vma)
+		return ~0UL;
+
+	/* We assume the page is modified.  */
+	page = follow_page(vma, vaddr, FOLL_WRITE | FOLL_TOUCH);
+	if (IS_ERR_OR_NULL(page))
+		return ~0UL;
+
+	return (page_to_pfn(page) << PAGE_SHIFT) | (vaddr & ~PAGE_MASK);
+}

arch/ia64/xen/xenivt.S

@@ -0,0 +1,52 @@
+/*
+ * arch/ia64/xen/ivt.S
+ *
+ * Copyright (C) 2005 Hewlett-Packard Co
+ *	Dan Magenheimer <dan.magenheimer@hp.com>
+ *
+ * Copyright (c) 2008 Isaku Yamahata <yamahata at valinux co jp>
+ *                    VA Linux Systems Japan K.K.
+ *                    pv_ops.
+ */
+
+#include <asm/asmmacro.h>
+#include <asm/kregs.h>
+#include <asm/pgtable.h>
+
+#include "../kernel/minstate.h"
+
+	.section .text,"ax"
+GLOBAL_ENTRY(xen_event_callback)
+	mov r31=pr		// prepare to save predicates
+	;;
+	SAVE_MIN_WITH_COVER	// uses r31; defines r2 and r3
+	;;
+	movl r3=XSI_PSR_IC
+	mov r14=1
+	;;
+	st4 [r3]=r14
+	;;
+	adds r3=8,r2		// set up second base pointer for SAVE_REST
+	srlz.i			// ensure everybody knows psr.ic is back on
+	;;
+	SAVE_REST
+	;;
+1:
+	alloc r14=ar.pfs,0,0,1,0 // must be first in an insn group
+	add out0=16,sp		// pass pointer to pt_regs as first arg
+	;;
+	br.call.sptk.many b0=xen_evtchn_do_upcall
+	;;
+	movl r20=XSI_PSR_I_ADDR
+	;;
+	ld8 r20=[r20]
+	;;
+	adds r20=-1,r20		// vcpu_info->evtchn_upcall_pending
+	;;
+	ld1 r20=[r20]
+	;;
+	cmp.ne p6,p0=r20,r0	// if there are pending events,
+	(p6) br.spnt.few 1b	// call evtchn_do_upcall again.
+	br.sptk.many xen_leave_kernel	// we know ia64_leave_kernel is
+					// paravirtualized as xen_leave_kernel
+END(xen_event_callback)

arch/ia64/xen/xensetup.S

@@ -0,0 +1,80 @@
+/*
+ * Support routines for Xen
+ *
+ * Copyright (C) 2005 Dan Magenheimer <dan.magenheimer@hp.com>
+ */
+
+#include <asm/processor.h>
+#include <asm/asmmacro.h>
+#include <asm/pgtable.h>
+#include <asm/paravirt.h>
+#include <asm/xen/privop.h>
+#include <linux/elfnote.h>
+#include <linux/init.h>
+#include <xen/interface/elfnote.h>
+
+	.section .data..read_mostly
+	.align 8
+	.global xen_domain_type
+xen_domain_type:
+	data4 XEN_NATIVE_ASM
+	.previous
+
+	__INIT
+ENTRY(startup_xen)
+	// Calculate load offset.
+	// The constant, LOAD_OFFSET, can't be used because the boot
+	// loader doesn't always load to the LMA specified by the vmlinux.lds.
+	mov r9=ip	// must be the first instruction to make sure
+			// that r9 = the physical address of startup_xen.
+			// Usually r9 = startup_xen - LOAD_OFFSET
+	movl r8=startup_xen
+	;;
+	sub r9=r9,r8	// Usually r9 = -LOAD_OFFSET.
+
+	mov r10=PARAVIRT_HYPERVISOR_TYPE_XEN
+	movl r11=_start
+	;;
+	add r11=r11,r9
+	movl r8=hypervisor_type
+	;;
+	add r8=r8,r9
+	mov b0=r11
+	;;
+	st8 [r8]=r10
+	br.cond.sptk.many b0
+	;;
+END(startup_xen)
+
+	ELFNOTE(Xen, XEN_ELFNOTE_GUEST_OS,	.asciz "linux")
+	ELFNOTE(Xen, XEN_ELFNOTE_GUEST_VERSION,	.asciz "2.6")
+	ELFNOTE(Xen, XEN_ELFNOTE_XEN_VERSION,	.asciz "xen-3.0")
+	ELFNOTE(Xen, XEN_ELFNOTE_ENTRY,		data8.ua startup_xen - LOAD_OFFSET)
+
+#define isBP	p3	// are we the Bootstrap Processor?
+
+GLOBAL_ENTRY(xen_setup_hook)
+	mov r8=XEN_PV_DOMAIN_ASM
+(isBP)	movl r9=xen_domain_type;;
+(isBP)	st4 [r9]=r8
+	movl r10=xen_ivt;;
+
+	mov cr.iva=r10
+
+	/* Set xsi base.  */
+#define FW_HYPERCALL_SET_SHARED_INFO_VA			0x600
+(isBP)	mov r2=FW_HYPERCALL_SET_SHARED_INFO_VA
+(isBP)	movl r28=XSI_BASE;;
+(isBP)	break 0x1000;;
+
+	/* setup pv_ops */
+(isBP)	mov r4=rp
+	;;
+(isBP)	br.call.sptk.many rp=xen_setup_pv_ops
+	;;
+(isBP)	mov rp=r4
+	;;
+
+	br.ret.sptk.many rp
+	;;
+END(xen_setup_hook)