i386: move hax accel files into hax/
authorClaudio Fontana <cfontana@suse.de>
Sat, 12 Dec 2020 15:55:10 +0000 (16:55 +0100)
committerEduardo Habkost <ehabkost@redhat.com>
Wed, 16 Dec 2020 19:06:53 +0000 (14:06 -0500)
Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20201212155530.23098-4-cfontana@suse.de>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
23 files changed:
MAINTAINERS
target/i386/hax-all.c [deleted file]
target/i386/hax-cpus.c [deleted file]
target/i386/hax-cpus.h [deleted file]
target/i386/hax-i386.h [deleted file]
target/i386/hax-interface.h [deleted file]
target/i386/hax-mem.c [deleted file]
target/i386/hax-posix.c [deleted file]
target/i386/hax-posix.h [deleted file]
target/i386/hax-windows.c [deleted file]
target/i386/hax-windows.h [deleted file]
target/i386/hax/hax-all.c [new file with mode: 0644]
target/i386/hax/hax-cpus.c [new file with mode: 0644]
target/i386/hax/hax-cpus.h [new file with mode: 0644]
target/i386/hax/hax-i386.h [new file with mode: 0644]
target/i386/hax/hax-interface.h [new file with mode: 0644]
target/i386/hax/hax-mem.c [new file with mode: 0644]
target/i386/hax/hax-posix.c [new file with mode: 0644]
target/i386/hax/hax-posix.h [new file with mode: 0644]
target/i386/hax/hax-windows.c [new file with mode: 0644]
target/i386/hax/hax-windows.h [new file with mode: 0644]
target/i386/hax/meson.build [new file with mode: 0644]
target/i386/meson.build

index 6f29b37834aaf784ae6b9dccfb45c4ccbf064da2..e03615d76c385a296b1585f3c1506d39ec7ee345 100644 (file)
@@ -492,7 +492,7 @@ W: https://github.com/intel/haxm/issues
 S: Maintained
 F: accel/stubs/hax-stub.c
 F: include/sysemu/hax.h
-F: target/i386/hax-*
+F: target/i386/hax/
 
 Hosts
 -----
diff --git a/target/i386/hax-all.c b/target/i386/hax-all.c
deleted file mode 100644 (file)
index fecfe8c..0000000
+++ /dev/null
@@ -1,1143 +0,0 @@
-/*
- * QEMU HAX support
- *
- * Copyright IBM, Corp. 2008
- *           Red Hat, Inc. 2008
- *
- * Authors:
- *  Anthony Liguori   <aliguori@us.ibm.com>
- *  Glauber Costa     <gcosta@redhat.com>
- *
- * Copyright (c) 2011 Intel Corporation
- *  Written by:
- *  Jiang Yunhong<yunhong.jiang@intel.com>
- *  Xin Xiaohui<xiaohui.xin@intel.com>
- *  Zhang Xiantao<xiantao.zhang@intel.com>
- *
- * This work is licensed under the terms of the GNU GPL, version 2 or later.
- * See the COPYING file in the top-level directory.
- *
- */
-
-/*
- * HAX common code for both windows and darwin
- */
-
-#include "qemu/osdep.h"
-#include "cpu.h"
-#include "exec/address-spaces.h"
-
-#include "qemu-common.h"
-#include "sysemu/accel.h"
-#include "sysemu/reset.h"
-#include "sysemu/runstate.h"
-#include "hw/boards.h"
-
-#include "hax-cpus.h"
-
-#define DEBUG_HAX 0
-
-#define DPRINTF(fmt, ...) \
-    do { \
-        if (DEBUG_HAX) { \
-            fprintf(stdout, fmt, ## __VA_ARGS__); \
-        } \
-    } while (0)
-
-/* Current version */
-const uint32_t hax_cur_version = 0x4; /* API v4: unmapping and MMIO moves */
-/* Minimum HAX kernel version */
-const uint32_t hax_min_version = 0x4; /* API v4: supports unmapping */
-
-static bool hax_allowed;
-
-struct hax_state hax_global;
-
-static void hax_vcpu_sync_state(CPUArchState *env, int modified);
-static int hax_arch_get_registers(CPUArchState *env);
-
-int hax_enabled(void)
-{
-    return hax_allowed;
-}
-
-int valid_hax_tunnel_size(uint16_t size)
-{
-    return size >= sizeof(struct hax_tunnel);
-}
-
-hax_fd hax_vcpu_get_fd(CPUArchState *env)
-{
-    struct hax_vcpu_state *vcpu = env_cpu(env)->hax_vcpu;
-    if (!vcpu) {
-        return HAX_INVALID_FD;
-    }
-    return vcpu->fd;
-}
-
-static int hax_get_capability(struct hax_state *hax)
-{
-    int ret;
-    struct hax_capabilityinfo capinfo, *cap = &capinfo;
-
-    ret = hax_capability(hax, cap);
-    if (ret) {
-        return ret;
-    }
-
-    if ((cap->wstatus & HAX_CAP_WORKSTATUS_MASK) == HAX_CAP_STATUS_NOTWORKING) {
-        if (cap->winfo & HAX_CAP_FAILREASON_VT) {
-            DPRINTF
-                ("VTX feature is not enabled, HAX driver will not work.\n");
-        } else if (cap->winfo & HAX_CAP_FAILREASON_NX) {
-            DPRINTF
-                ("NX feature is not enabled, HAX driver will not work.\n");
-        }
-        return -ENXIO;
-
-    }
-
-    if (!(cap->winfo & HAX_CAP_UG)) {
-        fprintf(stderr, "UG mode is not supported by the hardware.\n");
-        return -ENOTSUP;
-    }
-
-    hax->supports_64bit_ramblock = !!(cap->winfo & HAX_CAP_64BIT_RAMBLOCK);
-
-    if (cap->wstatus & HAX_CAP_MEMQUOTA) {
-        if (cap->mem_quota < hax->mem_quota) {
-            fprintf(stderr, "The VM memory needed exceeds the driver limit.\n");
-            return -ENOSPC;
-        }
-    }
-    return 0;
-}
-
-static int hax_version_support(struct hax_state *hax)
-{
-    int ret;
-    struct hax_module_version version;
-
-    ret = hax_mod_version(hax, &version);
-    if (ret < 0) {
-        return 0;
-    }
-
-    if (hax_min_version > version.cur_version) {
-        fprintf(stderr, "Incompatible HAX module version %d,",
-                version.cur_version);
-        fprintf(stderr, "requires minimum version %d\n", hax_min_version);
-        return 0;
-    }
-    if (hax_cur_version < version.compat_version) {
-        fprintf(stderr, "Incompatible QEMU HAX API version %x,",
-                hax_cur_version);
-        fprintf(stderr, "requires minimum HAX API version %x\n",
-                version.compat_version);
-        return 0;
-    }
-
-    return 1;
-}
-
-int hax_vcpu_create(int id)
-{
-    struct hax_vcpu_state *vcpu = NULL;
-    int ret;
-
-    if (!hax_global.vm) {
-        fprintf(stderr, "vcpu %x created failed, vm is null\n", id);
-        return -1;
-    }
-
-    if (hax_global.vm->vcpus[id]) {
-        fprintf(stderr, "vcpu %x allocated already\n", id);
-        return 0;
-    }
-
-    vcpu = g_new0(struct hax_vcpu_state, 1);
-
-    ret = hax_host_create_vcpu(hax_global.vm->fd, id);
-    if (ret) {
-        fprintf(stderr, "Failed to create vcpu %x\n", id);
-        goto error;
-    }
-
-    vcpu->vcpu_id = id;
-    vcpu->fd = hax_host_open_vcpu(hax_global.vm->id, id);
-    if (hax_invalid_fd(vcpu->fd)) {
-        fprintf(stderr, "Failed to open the vcpu\n");
-        ret = -ENODEV;
-        goto error;
-    }
-
-    hax_global.vm->vcpus[id] = vcpu;
-
-    ret = hax_host_setup_vcpu_channel(vcpu);
-    if (ret) {
-        fprintf(stderr, "Invalid hax tunnel size\n");
-        ret = -EINVAL;
-        goto error;
-    }
-    return 0;
-
-  error:
-    /* vcpu and tunnel will be closed automatically */
-    if (vcpu && !hax_invalid_fd(vcpu->fd)) {
-        hax_close_fd(vcpu->fd);
-    }
-
-    hax_global.vm->vcpus[id] = NULL;
-    g_free(vcpu);
-    return -1;
-}
-
-int hax_vcpu_destroy(CPUState *cpu)
-{
-    struct hax_vcpu_state *vcpu = cpu->hax_vcpu;
-
-    if (!hax_global.vm) {
-        fprintf(stderr, "vcpu %x destroy failed, vm is null\n", vcpu->vcpu_id);
-        return -1;
-    }
-
-    if (!vcpu) {
-        return 0;
-    }
-
-    /*
-     * 1. The hax_tunnel is also destroyed when vcpu is destroyed
-     * 2. close fd will cause hax module vcpu be cleaned
-     */
-    hax_close_fd(vcpu->fd);
-    hax_global.vm->vcpus[vcpu->vcpu_id] = NULL;
-    g_free(vcpu);
-    return 0;
-}
-
-int hax_init_vcpu(CPUState *cpu)
-{
-    int ret;
-
-    ret = hax_vcpu_create(cpu->cpu_index);
-    if (ret < 0) {
-        fprintf(stderr, "Failed to create HAX vcpu\n");
-        exit(-1);
-    }
-
-    cpu->hax_vcpu = hax_global.vm->vcpus[cpu->cpu_index];
-    cpu->vcpu_dirty = true;
-    qemu_register_reset(hax_reset_vcpu_state, (CPUArchState *) (cpu->env_ptr));
-
-    return ret;
-}
-
-struct hax_vm *hax_vm_create(struct hax_state *hax, int max_cpus)
-{
-    struct hax_vm *vm;
-    int vm_id = 0, ret, i;
-
-    if (hax_invalid_fd(hax->fd)) {
-        return NULL;
-    }
-
-    if (hax->vm) {
-        return hax->vm;
-    }
-
-    if (max_cpus > HAX_MAX_VCPU) {
-        fprintf(stderr, "Maximum VCPU number QEMU supported is %d\n", HAX_MAX_VCPU);
-        return NULL;
-    }
-
-    vm = g_new0(struct hax_vm, 1);
-
-    ret = hax_host_create_vm(hax, &vm_id);
-    if (ret) {
-        fprintf(stderr, "Failed to create vm %x\n", ret);
-        goto error;
-    }
-    vm->id = vm_id;
-    vm->fd = hax_host_open_vm(hax, vm_id);
-    if (hax_invalid_fd(vm->fd)) {
-        fprintf(stderr, "Failed to open vm %d\n", vm_id);
-        goto error;
-    }
-
-    vm->numvcpus = max_cpus;
-    vm->vcpus = g_new0(struct hax_vcpu_state *, vm->numvcpus);
-    for (i = 0; i < vm->numvcpus; i++) {
-        vm->vcpus[i] = NULL;
-    }
-
-    hax->vm = vm;
-    return vm;
-
-  error:
-    g_free(vm);
-    hax->vm = NULL;
-    return NULL;
-}
-
-int hax_vm_destroy(struct hax_vm *vm)
-{
-    int i;
-
-    for (i = 0; i < vm->numvcpus; i++)
-        if (vm->vcpus[i]) {
-            fprintf(stderr, "VCPU should be cleaned before vm clean\n");
-            return -1;
-        }
-    hax_close_fd(vm->fd);
-    vm->numvcpus = 0;
-    g_free(vm->vcpus);
-    g_free(vm);
-    hax_global.vm = NULL;
-    return 0;
-}
-
-static int hax_init(ram_addr_t ram_size, int max_cpus)
-{
-    struct hax_state *hax = NULL;
-    struct hax_qemu_version qversion;
-    int ret;
-
-    hax = &hax_global;
-
-    memset(hax, 0, sizeof(struct hax_state));
-    hax->mem_quota = ram_size;
-
-    hax->fd = hax_mod_open();
-    if (hax_invalid_fd(hax->fd)) {
-        hax->fd = 0;
-        ret = -ENODEV;
-        goto error;
-    }
-
-    ret = hax_get_capability(hax);
-
-    if (ret) {
-        if (ret != -ENOSPC) {
-            ret = -EINVAL;
-        }
-        goto error;
-    }
-
-    if (!hax_version_support(hax)) {
-        ret = -EINVAL;
-        goto error;
-    }
-
-    hax->vm = hax_vm_create(hax, max_cpus);
-    if (!hax->vm) {
-        fprintf(stderr, "Failed to create HAX VM\n");
-        ret = -EINVAL;
-        goto error;
-    }
-
-    hax_memory_init();
-
-    qversion.cur_version = hax_cur_version;
-    qversion.min_version = hax_min_version;
-    hax_notify_qemu_version(hax->vm->fd, &qversion);
-
-    return ret;
-  error:
-    if (hax->vm) {
-        hax_vm_destroy(hax->vm);
-    }
-    if (hax->fd) {
-        hax_mod_close(hax);
-    }
-
-    return ret;
-}
-
-static int hax_accel_init(MachineState *ms)
-{
-    int ret = hax_init(ms->ram_size, (int)ms->smp.max_cpus);
-
-    if (ret && (ret != -ENOSPC)) {
-        fprintf(stderr, "No accelerator found.\n");
-    } else {
-        fprintf(stdout, "HAX is %s and emulator runs in %s mode.\n",
-                !ret ? "working" : "not working",
-                !ret ? "fast virt" : "emulation");
-    }
-    if (ret == 0) {
-        cpus_register_accel(&hax_cpus);
-    }
-    return ret;
-}
-
-static int hax_handle_fastmmio(CPUArchState *env, struct hax_fastmmio *hft)
-{
-    if (hft->direction < 2) {
-        cpu_physical_memory_rw(hft->gpa, &hft->value, hft->size,
-                               hft->direction);
-    } else {
-        /*
-         * HAX API v4 supports transferring data between two MMIO addresses,
-         * hft->gpa and hft->gpa2 (instructions such as MOVS require this):
-         *  hft->direction == 2: gpa ==> gpa2
-         */
-        uint64_t value;
-        cpu_physical_memory_read(hft->gpa, &value, hft->size);
-        cpu_physical_memory_write(hft->gpa2, &value, hft->size);
-    }
-
-    return 0;
-}
-
-static int hax_handle_io(CPUArchState *env, uint32_t df, uint16_t port,
-                         int direction, int size, int count, void *buffer)
-{
-    uint8_t *ptr;
-    int i;
-    MemTxAttrs attrs = { 0 };
-
-    if (!df) {
-        ptr = (uint8_t *) buffer;
-    } else {
-        ptr = buffer + size * count - size;
-    }
-    for (i = 0; i < count; i++) {
-        address_space_rw(&address_space_io, port, attrs,
-                         ptr, size, direction == HAX_EXIT_IO_OUT);
-        if (!df) {
-            ptr += size;
-        } else {
-            ptr -= size;
-        }
-    }
-
-    return 0;
-}
-
-static int hax_vcpu_interrupt(CPUArchState *env)
-{
-    CPUState *cpu = env_cpu(env);
-    struct hax_vcpu_state *vcpu = cpu->hax_vcpu;
-    struct hax_tunnel *ht = vcpu->tunnel;
-
-    /*
-     * Try to inject an interrupt if the guest can accept it
-     * Unlike KVM, HAX kernel check for the eflags, instead of qemu
-     */
-    if (ht->ready_for_interrupt_injection &&
-        (cpu->interrupt_request & CPU_INTERRUPT_HARD)) {
-        int irq;
-
-        irq = cpu_get_pic_interrupt(env);
-        if (irq >= 0) {
-            hax_inject_interrupt(env, irq);
-            cpu->interrupt_request &= ~CPU_INTERRUPT_HARD;
-        }
-    }
-
-    /* If we have an interrupt but the guest is not ready to receive an
-     * interrupt, request an interrupt window exit.  This will
-     * cause a return to userspace as soon as the guest is ready to
-     * receive interrupts. */
-    if ((cpu->interrupt_request & CPU_INTERRUPT_HARD)) {
-        ht->request_interrupt_window = 1;
-    } else {
-        ht->request_interrupt_window = 0;
-    }
-    return 0;
-}
-
-void hax_raise_event(CPUState *cpu)
-{
-    struct hax_vcpu_state *vcpu = cpu->hax_vcpu;
-
-    if (!vcpu) {
-        return;
-    }
-    vcpu->tunnel->user_event_pending = 1;
-}
-
-/*
- * Ask hax kernel module to run the CPU for us till:
- * 1. Guest crash or shutdown
- * 2. Need QEMU's emulation like guest execute MMIO instruction
- * 3. Guest execute HLT
- * 4. QEMU have Signal/event pending
- * 5. An unknown VMX exit happens
- */
-static int hax_vcpu_hax_exec(CPUArchState *env)
-{
-    int ret = 0;
-    CPUState *cpu = env_cpu(env);
-    X86CPU *x86_cpu = X86_CPU(cpu);
-    struct hax_vcpu_state *vcpu = cpu->hax_vcpu;
-    struct hax_tunnel *ht = vcpu->tunnel;
-
-    if (!hax_enabled()) {
-        DPRINTF("Trying to vcpu execute at eip:" TARGET_FMT_lx "\n", env->eip);
-        return 0;
-    }
-
-    if (cpu->interrupt_request & CPU_INTERRUPT_POLL) {
-        cpu->interrupt_request &= ~CPU_INTERRUPT_POLL;
-        apic_poll_irq(x86_cpu->apic_state);
-    }
-
-    /* After a vcpu is halted (either because it is an AP and has just been
-     * reset, or because it has executed the HLT instruction), it will not be
-     * run (hax_vcpu_run()) until it is unhalted. The next few if blocks check
-     * for events that may change the halted state of this vcpu:
-     *  a) Maskable interrupt, when RFLAGS.IF is 1;
-     *     Note: env->eflags may not reflect the current RFLAGS state, because
-     *           it is not updated after each hax_vcpu_run(). We cannot afford
-     *           to fail to recognize any unhalt-by-maskable-interrupt event
-     *           (in which case the vcpu will halt forever), and yet we cannot
-     *           afford the overhead of hax_vcpu_sync_state(). The current
-     *           solution is to err on the side of caution and have the HLT
-     *           handler (see case HAX_EXIT_HLT below) unconditionally set the
-     *           IF_MASK bit in env->eflags, which, in effect, disables the
-     *           RFLAGS.IF check.
-     *  b) NMI;
-     *  c) INIT signal;
-     *  d) SIPI signal.
-     */
-    if (((cpu->interrupt_request & CPU_INTERRUPT_HARD) &&
-         (env->eflags & IF_MASK)) ||
-        (cpu->interrupt_request & CPU_INTERRUPT_NMI)) {
-        cpu->halted = 0;
-    }
-
-    if (cpu->interrupt_request & CPU_INTERRUPT_INIT) {
-        DPRINTF("\nhax_vcpu_hax_exec: handling INIT for %d\n",
-                cpu->cpu_index);
-        do_cpu_init(x86_cpu);
-        hax_vcpu_sync_state(env, 1);
-    }
-
-    if (cpu->interrupt_request & CPU_INTERRUPT_SIPI) {
-        DPRINTF("hax_vcpu_hax_exec: handling SIPI for %d\n",
-                cpu->cpu_index);
-        hax_vcpu_sync_state(env, 0);
-        do_cpu_sipi(x86_cpu);
-        hax_vcpu_sync_state(env, 1);
-    }
-
-    if (cpu->halted) {
-        /* If this vcpu is halted, we must not ask HAXM to run it. Instead, we
-         * break out of hax_smp_cpu_exec() as if this vcpu had executed HLT.
-         * That way, this vcpu thread will be trapped in qemu_wait_io_event(),
-         * until the vcpu is unhalted.
-         */
-        cpu->exception_index = EXCP_HLT;
-        return 0;
-    }
-
-    do {
-        int hax_ret;
-
-        if (cpu->exit_request) {
-            ret = 1;
-            break;
-        }
-
-        hax_vcpu_interrupt(env);
-
-        qemu_mutex_unlock_iothread();
-        cpu_exec_start(cpu);
-        hax_ret = hax_vcpu_run(vcpu);
-        cpu_exec_end(cpu);
-        qemu_mutex_lock_iothread();
-
-        /* Simply continue the vcpu_run if system call interrupted */
-        if (hax_ret == -EINTR || hax_ret == -EAGAIN) {
-            DPRINTF("io window interrupted\n");
-            continue;
-        }
-
-        if (hax_ret < 0) {
-            fprintf(stderr, "vcpu run failed for vcpu  %x\n", vcpu->vcpu_id);
-            abort();
-        }
-        switch (ht->_exit_status) {
-        case HAX_EXIT_IO:
-            ret = hax_handle_io(env, ht->pio._df, ht->pio._port,
-                            ht->pio._direction,
-                            ht->pio._size, ht->pio._count, vcpu->iobuf);
-            break;
-        case HAX_EXIT_FAST_MMIO:
-            ret = hax_handle_fastmmio(env, (struct hax_fastmmio *) vcpu->iobuf);
-            break;
-        /* Guest state changed, currently only for shutdown */
-        case HAX_EXIT_STATECHANGE:
-            fprintf(stdout, "VCPU shutdown request\n");
-            qemu_system_shutdown_request(SHUTDOWN_CAUSE_GUEST_SHUTDOWN);
-            hax_vcpu_sync_state(env, 0);
-            ret = 1;
-            break;
-        case HAX_EXIT_UNKNOWN_VMEXIT:
-            fprintf(stderr, "Unknown VMX exit %x from guest\n",
-                    ht->_exit_reason);
-            qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
-            hax_vcpu_sync_state(env, 0);
-            cpu_dump_state(cpu, stderr, 0);
-            ret = -1;
-            break;
-        case HAX_EXIT_HLT:
-            if (!(cpu->interrupt_request & CPU_INTERRUPT_HARD) &&
-                !(cpu->interrupt_request & CPU_INTERRUPT_NMI)) {
-                /* hlt instruction with interrupt disabled is shutdown */
-                env->eflags |= IF_MASK;
-                cpu->halted = 1;
-                cpu->exception_index = EXCP_HLT;
-                ret = 1;
-            }
-            break;
-        /* these situations will continue to hax module */
-        case HAX_EXIT_INTERRUPT:
-        case HAX_EXIT_PAUSED:
-            break;
-        case HAX_EXIT_MMIO:
-            /* Should not happen on UG system */
-            fprintf(stderr, "HAX: unsupported MMIO emulation\n");
-            ret = -1;
-            break;
-        case HAX_EXIT_REAL:
-            /* Should not happen on UG system */
-            fprintf(stderr, "HAX: unimplemented real mode emulation\n");
-            ret = -1;
-            break;
-        default:
-            fprintf(stderr, "Unknown exit %x from HAX\n", ht->_exit_status);
-            qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
-            hax_vcpu_sync_state(env, 0);
-            cpu_dump_state(cpu, stderr, 0);
-            ret = 1;
-            break;
-        }
-    } while (!ret);
-
-    if (cpu->exit_request) {
-        cpu->exit_request = 0;
-        cpu->exception_index = EXCP_INTERRUPT;
-    }
-    return ret < 0;
-}
-
-static void do_hax_cpu_synchronize_state(CPUState *cpu, run_on_cpu_data arg)
-{
-    CPUArchState *env = cpu->env_ptr;
-
-    hax_arch_get_registers(env);
-    cpu->vcpu_dirty = true;
-}
-
-void hax_cpu_synchronize_state(CPUState *cpu)
-{
-    if (!cpu->vcpu_dirty) {
-        run_on_cpu(cpu, do_hax_cpu_synchronize_state, RUN_ON_CPU_NULL);
-    }
-}
-
-static void do_hax_cpu_synchronize_post_reset(CPUState *cpu,
-                                              run_on_cpu_data arg)
-{
-    CPUArchState *env = cpu->env_ptr;
-
-    hax_vcpu_sync_state(env, 1);
-    cpu->vcpu_dirty = false;
-}
-
-void hax_cpu_synchronize_post_reset(CPUState *cpu)
-{
-    run_on_cpu(cpu, do_hax_cpu_synchronize_post_reset, RUN_ON_CPU_NULL);
-}
-
-static void do_hax_cpu_synchronize_post_init(CPUState *cpu, run_on_cpu_data arg)
-{
-    CPUArchState *env = cpu->env_ptr;
-
-    hax_vcpu_sync_state(env, 1);
-    cpu->vcpu_dirty = false;
-}
-
-void hax_cpu_synchronize_post_init(CPUState *cpu)
-{
-    run_on_cpu(cpu, do_hax_cpu_synchronize_post_init, RUN_ON_CPU_NULL);
-}
-
-static void do_hax_cpu_synchronize_pre_loadvm(CPUState *cpu, run_on_cpu_data arg)
-{
-    cpu->vcpu_dirty = true;
-}
-
-void hax_cpu_synchronize_pre_loadvm(CPUState *cpu)
-{
-    run_on_cpu(cpu, do_hax_cpu_synchronize_pre_loadvm, RUN_ON_CPU_NULL);
-}
-
-int hax_smp_cpu_exec(CPUState *cpu)
-{
-    CPUArchState *env = (CPUArchState *) (cpu->env_ptr);
-    int fatal;
-    int ret;
-
-    while (1) {
-        if (cpu->exception_index >= EXCP_INTERRUPT) {
-            ret = cpu->exception_index;
-            cpu->exception_index = -1;
-            break;
-        }
-
-        fatal = hax_vcpu_hax_exec(env);
-
-        if (fatal) {
-            fprintf(stderr, "Unsupported HAX vcpu return\n");
-            abort();
-        }
-    }
-
-    return ret;
-}
-
-static void set_v8086_seg(struct segment_desc_t *lhs, const SegmentCache *rhs)
-{
-    memset(lhs, 0, sizeof(struct segment_desc_t));
-    lhs->selector = rhs->selector;
-    lhs->base = rhs->base;
-    lhs->limit = rhs->limit;
-    lhs->type = 3;
-    lhs->present = 1;
-    lhs->dpl = 3;
-    lhs->operand_size = 0;
-    lhs->desc = 1;
-    lhs->long_mode = 0;
-    lhs->granularity = 0;
-    lhs->available = 0;
-}
-
-static void get_seg(SegmentCache *lhs, const struct segment_desc_t *rhs)
-{
-    lhs->selector = rhs->selector;
-    lhs->base = rhs->base;
-    lhs->limit = rhs->limit;
-    lhs->flags = (rhs->type << DESC_TYPE_SHIFT)
-        | (rhs->present * DESC_P_MASK)
-        | (rhs->dpl << DESC_DPL_SHIFT)
-        | (rhs->operand_size << DESC_B_SHIFT)
-        | (rhs->desc * DESC_S_MASK)
-        | (rhs->long_mode << DESC_L_SHIFT)
-        | (rhs->granularity * DESC_G_MASK) | (rhs->available * DESC_AVL_MASK);
-}
-
-static void set_seg(struct segment_desc_t *lhs, const SegmentCache *rhs)
-{
-    unsigned flags = rhs->flags;
-
-    memset(lhs, 0, sizeof(struct segment_desc_t));
-    lhs->selector = rhs->selector;
-    lhs->base = rhs->base;
-    lhs->limit = rhs->limit;
-    lhs->type = (flags >> DESC_TYPE_SHIFT) & 15;
-    lhs->present = (flags & DESC_P_MASK) != 0;
-    lhs->dpl = rhs->selector & 3;
-    lhs->operand_size = (flags >> DESC_B_SHIFT) & 1;
-    lhs->desc = (flags & DESC_S_MASK) != 0;
-    lhs->long_mode = (flags >> DESC_L_SHIFT) & 1;
-    lhs->granularity = (flags & DESC_G_MASK) != 0;
-    lhs->available = (flags & DESC_AVL_MASK) != 0;
-}
-
-static void hax_getput_reg(uint64_t *hax_reg, target_ulong *qemu_reg, int set)
-{
-    target_ulong reg = *hax_reg;
-
-    if (set) {
-        *hax_reg = *qemu_reg;
-    } else {
-        *qemu_reg = reg;
-    }
-}
-
-/* The sregs has been synced with HAX kernel already before this call */
-static int hax_get_segments(CPUArchState *env, struct vcpu_state_t *sregs)
-{
-    get_seg(&env->segs[R_CS], &sregs->_cs);
-    get_seg(&env->segs[R_DS], &sregs->_ds);
-    get_seg(&env->segs[R_ES], &sregs->_es);
-    get_seg(&env->segs[R_FS], &sregs->_fs);
-    get_seg(&env->segs[R_GS], &sregs->_gs);
-    get_seg(&env->segs[R_SS], &sregs->_ss);
-
-    get_seg(&env->tr, &sregs->_tr);
-    get_seg(&env->ldt, &sregs->_ldt);
-    env->idt.limit = sregs->_idt.limit;
-    env->idt.base = sregs->_idt.base;
-    env->gdt.limit = sregs->_gdt.limit;
-    env->gdt.base = sregs->_gdt.base;
-    return 0;
-}
-
-static int hax_set_segments(CPUArchState *env, struct vcpu_state_t *sregs)
-{
-    if ((env->eflags & VM_MASK)) {
-        set_v8086_seg(&sregs->_cs, &env->segs[R_CS]);
-        set_v8086_seg(&sregs->_ds, &env->segs[R_DS]);
-        set_v8086_seg(&sregs->_es, &env->segs[R_ES]);
-        set_v8086_seg(&sregs->_fs, &env->segs[R_FS]);
-        set_v8086_seg(&sregs->_gs, &env->segs[R_GS]);
-        set_v8086_seg(&sregs->_ss, &env->segs[R_SS]);
-    } else {
-        set_seg(&sregs->_cs, &env->segs[R_CS]);
-        set_seg(&sregs->_ds, &env->segs[R_DS]);
-        set_seg(&sregs->_es, &env->segs[R_ES]);
-        set_seg(&sregs->_fs, &env->segs[R_FS]);
-        set_seg(&sregs->_gs, &env->segs[R_GS]);
-        set_seg(&sregs->_ss, &env->segs[R_SS]);
-
-        if (env->cr[0] & CR0_PE_MASK) {
-            /* force ss cpl to cs cpl */
-            sregs->_ss.selector = (sregs->_ss.selector & ~3) |
-                                  (sregs->_cs.selector & 3);
-            sregs->_ss.dpl = sregs->_ss.selector & 3;
-        }
-    }
-
-    set_seg(&sregs->_tr, &env->tr);
-    set_seg(&sregs->_ldt, &env->ldt);
-    sregs->_idt.limit = env->idt.limit;
-    sregs->_idt.base = env->idt.base;
-    sregs->_gdt.limit = env->gdt.limit;
-    sregs->_gdt.base = env->gdt.base;
-    return 0;
-}
-
-static int hax_sync_vcpu_register(CPUArchState *env, int set)
-{
-    struct vcpu_state_t regs;
-    int ret;
-    memset(&regs, 0, sizeof(struct vcpu_state_t));
-
-    if (!set) {
-        ret = hax_sync_vcpu_state(env, &regs, 0);
-        if (ret < 0) {
-            return -1;
-        }
-    }
-
-    /* generic register */
-    hax_getput_reg(&regs._rax, &env->regs[R_EAX], set);
-    hax_getput_reg(&regs._rbx, &env->regs[R_EBX], set);
-    hax_getput_reg(&regs._rcx, &env->regs[R_ECX], set);
-    hax_getput_reg(&regs._rdx, &env->regs[R_EDX], set);
-    hax_getput_reg(&regs._rsi, &env->regs[R_ESI], set);
-    hax_getput_reg(&regs._rdi, &env->regs[R_EDI], set);
-    hax_getput_reg(&regs._rsp, &env->regs[R_ESP], set);
-    hax_getput_reg(&regs._rbp, &env->regs[R_EBP], set);
-#ifdef TARGET_X86_64
-    hax_getput_reg(&regs._r8, &env->regs[8], set);
-    hax_getput_reg(&regs._r9, &env->regs[9], set);
-    hax_getput_reg(&regs._r10, &env->regs[10], set);
-    hax_getput_reg(&regs._r11, &env->regs[11], set);
-    hax_getput_reg(&regs._r12, &env->regs[12], set);
-    hax_getput_reg(&regs._r13, &env->regs[13], set);
-    hax_getput_reg(&regs._r14, &env->regs[14], set);
-    hax_getput_reg(&regs._r15, &env->regs[15], set);
-#endif
-    hax_getput_reg(&regs._rflags, &env->eflags, set);
-    hax_getput_reg(&regs._rip, &env->eip, set);
-
-    if (set) {
-        regs._cr0 = env->cr[0];
-        regs._cr2 = env->cr[2];
-        regs._cr3 = env->cr[3];
-        regs._cr4 = env->cr[4];
-        hax_set_segments(env, &regs);
-    } else {
-        env->cr[0] = regs._cr0;
-        env->cr[2] = regs._cr2;
-        env->cr[3] = regs._cr3;
-        env->cr[4] = regs._cr4;
-        hax_get_segments(env, &regs);
-    }
-
-    if (set) {
-        ret = hax_sync_vcpu_state(env, &regs, 1);
-        if (ret < 0) {
-            return -1;
-        }
-    }
-    return 0;
-}
-
-static void hax_msr_entry_set(struct vmx_msr *item, uint32_t index,
-                              uint64_t value)
-{
-    item->entry = index;
-    item->value = value;
-}
-
-static int hax_get_msrs(CPUArchState *env)
-{
-    struct hax_msr_data md;
-    struct vmx_msr *msrs = md.entries;
-    int ret, i, n;
-
-    n = 0;
-    msrs[n++].entry = MSR_IA32_SYSENTER_CS;
-    msrs[n++].entry = MSR_IA32_SYSENTER_ESP;
-    msrs[n++].entry = MSR_IA32_SYSENTER_EIP;
-    msrs[n++].entry = MSR_IA32_TSC;
-#ifdef TARGET_X86_64
-    msrs[n++].entry = MSR_EFER;
-    msrs[n++].entry = MSR_STAR;
-    msrs[n++].entry = MSR_LSTAR;
-    msrs[n++].entry = MSR_CSTAR;
-    msrs[n++].entry = MSR_FMASK;
-    msrs[n++].entry = MSR_KERNELGSBASE;
-#endif
-    md.nr_msr = n;
-    ret = hax_sync_msr(env, &md, 0);
-    if (ret < 0) {
-        return ret;
-    }
-
-    for (i = 0; i < md.done; i++) {
-        switch (msrs[i].entry) {
-        case MSR_IA32_SYSENTER_CS:
-            env->sysenter_cs = msrs[i].value;
-            break;
-        case MSR_IA32_SYSENTER_ESP:
-            env->sysenter_esp = msrs[i].value;
-            break;
-        case MSR_IA32_SYSENTER_EIP:
-            env->sysenter_eip = msrs[i].value;
-            break;
-        case MSR_IA32_TSC:
-            env->tsc = msrs[i].value;
-            break;
-#ifdef TARGET_X86_64
-        case MSR_EFER:
-            env->efer = msrs[i].value;
-            break;
-        case MSR_STAR:
-            env->star = msrs[i].value;
-            break;
-        case MSR_LSTAR:
-            env->lstar = msrs[i].value;
-            break;
-        case MSR_CSTAR:
-            env->cstar = msrs[i].value;
-            break;
-        case MSR_FMASK:
-            env->fmask = msrs[i].value;
-            break;
-        case MSR_KERNELGSBASE:
-            env->kernelgsbase = msrs[i].value;
-            break;
-#endif
-        }
-    }
-
-    return 0;
-}
-
-static int hax_set_msrs(CPUArchState *env)
-{
-    struct hax_msr_data md;
-    struct vmx_msr *msrs;
-    msrs = md.entries;
-    int n = 0;
-
-    memset(&md, 0, sizeof(struct hax_msr_data));
-    hax_msr_entry_set(&msrs[n++], MSR_IA32_SYSENTER_CS, env->sysenter_cs);
-    hax_msr_entry_set(&msrs[n++], MSR_IA32_SYSENTER_ESP, env->sysenter_esp);
-    hax_msr_entry_set(&msrs[n++], MSR_IA32_SYSENTER_EIP, env->sysenter_eip);
-    hax_msr_entry_set(&msrs[n++], MSR_IA32_TSC, env->tsc);
-#ifdef TARGET_X86_64
-    hax_msr_entry_set(&msrs[n++], MSR_EFER, env->efer);
-    hax_msr_entry_set(&msrs[n++], MSR_STAR, env->star);
-    hax_msr_entry_set(&msrs[n++], MSR_LSTAR, env->lstar);
-    hax_msr_entry_set(&msrs[n++], MSR_CSTAR, env->cstar);
-    hax_msr_entry_set(&msrs[n++], MSR_FMASK, env->fmask);
-    hax_msr_entry_set(&msrs[n++], MSR_KERNELGSBASE, env->kernelgsbase);
-#endif
-    md.nr_msr = n;
-    md.done = 0;
-
-    return hax_sync_msr(env, &md, 1);
-}
-
-static int hax_get_fpu(CPUArchState *env)
-{
-    struct fx_layout fpu;
-    int i, ret;
-
-    ret = hax_sync_fpu(env, &fpu, 0);
-    if (ret < 0) {
-        return ret;
-    }
-
-    env->fpstt = (fpu.fsw >> 11) & 7;
-    env->fpus = fpu.fsw;
-    env->fpuc = fpu.fcw;
-    for (i = 0; i < 8; ++i) {
-        env->fptags[i] = !((fpu.ftw >> i) & 1);
-    }
-    memcpy(env->fpregs, fpu.st_mm, sizeof(env->fpregs));
-
-    for (i = 0; i < 8; i++) {
-        env->xmm_regs[i].ZMM_Q(0) = ldq_p(&fpu.mmx_1[i][0]);
-        env->xmm_regs[i].ZMM_Q(1) = ldq_p(&fpu.mmx_1[i][8]);
-        if (CPU_NB_REGS > 8) {
-            env->xmm_regs[i + 8].ZMM_Q(0) = ldq_p(&fpu.mmx_2[i][0]);
-            env->xmm_regs[i + 8].ZMM_Q(1) = ldq_p(&fpu.mmx_2[i][8]);
-        }
-    }
-    env->mxcsr = fpu.mxcsr;
-
-    return 0;
-}
-
-static int hax_set_fpu(CPUArchState *env)
-{
-    struct fx_layout fpu;
-    int i;
-
-    memset(&fpu, 0, sizeof(fpu));
-    fpu.fsw = env->fpus & ~(7 << 11);
-    fpu.fsw |= (env->fpstt & 7) << 11;
-    fpu.fcw = env->fpuc;
-
-    for (i = 0; i < 8; ++i) {
-        fpu.ftw |= (!env->fptags[i]) << i;
-    }
-
-    memcpy(fpu.st_mm, env->fpregs, sizeof(env->fpregs));
-    for (i = 0; i < 8; i++) {
-        stq_p(&fpu.mmx_1[i][0], env->xmm_regs[i].ZMM_Q(0));
-        stq_p(&fpu.mmx_1[i][8], env->xmm_regs[i].ZMM_Q(1));
-        if (CPU_NB_REGS > 8) {
-            stq_p(&fpu.mmx_2[i][0], env->xmm_regs[i + 8].ZMM_Q(0));
-            stq_p(&fpu.mmx_2[i][8], env->xmm_regs[i + 8].ZMM_Q(1));
-        }
-    }
-
-    fpu.mxcsr = env->mxcsr;
-
-    return hax_sync_fpu(env, &fpu, 1);
-}
-
-static int hax_arch_get_registers(CPUArchState *env)
-{
-    int ret;
-
-    ret = hax_sync_vcpu_register(env, 0);
-    if (ret < 0) {
-        return ret;
-    }
-
-    ret = hax_get_fpu(env);
-    if (ret < 0) {
-        return ret;
-    }
-
-    ret = hax_get_msrs(env);
-    if (ret < 0) {
-        return ret;
-    }
-
-    x86_update_hflags(env);
-    return 0;
-}
-
-static int hax_arch_set_registers(CPUArchState *env)
-{
-    int ret;
-    ret = hax_sync_vcpu_register(env, 1);
-
-    if (ret < 0) {
-        fprintf(stderr, "Failed to sync vcpu reg\n");
-        return ret;
-    }
-    ret = hax_set_fpu(env);
-    if (ret < 0) {
-        fprintf(stderr, "FPU failed\n");
-        return ret;
-    }
-    ret = hax_set_msrs(env);
-    if (ret < 0) {
-        fprintf(stderr, "MSR failed\n");
-        return ret;
-    }
-
-    return 0;
-}
-
-static void hax_vcpu_sync_state(CPUArchState *env, int modified)
-{
-    if (hax_enabled()) {
-        if (modified) {
-            hax_arch_set_registers(env);
-        } else {
-            hax_arch_get_registers(env);
-        }
-    }
-}
-
-/*
- * much simpler than kvm, at least in first stage because:
- * We don't need consider the device pass-through, we don't need
- * consider the framebuffer, and we may even remove the bios at all
- */
-int hax_sync_vcpus(void)
-{
-    if (hax_enabled()) {
-        CPUState *cpu;
-
-        cpu = first_cpu;
-        if (!cpu) {
-            return 0;
-        }
-
-        for (; cpu != NULL; cpu = CPU_NEXT(cpu)) {
-            int ret;
-
-            ret = hax_arch_set_registers(cpu->env_ptr);
-            if (ret < 0) {
-                return ret;
-            }
-        }
-    }
-
-    return 0;
-}
-
-void hax_reset_vcpu_state(void *opaque)
-{
-    CPUState *cpu;
-    for (cpu = first_cpu; cpu != NULL; cpu = CPU_NEXT(cpu)) {
-        cpu->hax_vcpu->tunnel->user_event_pending = 0;
-        cpu->hax_vcpu->tunnel->ready_for_interrupt_injection = 0;
-    }
-}
-
-static void hax_accel_class_init(ObjectClass *oc, void *data)
-{
-    AccelClass *ac = ACCEL_CLASS(oc);
-    ac->name = "HAX";
-    ac->init_machine = hax_accel_init;
-    ac->allowed = &hax_allowed;
-}
-
-static const TypeInfo hax_accel_type = {
-    .name = ACCEL_CLASS_NAME("hax"),
-    .parent = TYPE_ACCEL,
-    .class_init = hax_accel_class_init,
-};
-
-static void hax_type_init(void)
-{
-    type_register_static(&hax_accel_type);
-}
-
-type_init(hax_type_init);
diff --git a/target/i386/hax-cpus.c b/target/i386/hax-cpus.c
deleted file mode 100644 (file)
index f72c85b..0000000
+++ /dev/null
@@ -1,85 +0,0 @@
-/*
- * QEMU HAX support
- *
- * Copyright IBM, Corp. 2008
- *           Red Hat, Inc. 2008
- *
- * Authors:
- *  Anthony Liguori   <aliguori@us.ibm.com>
- *  Glauber Costa     <gcosta@redhat.com>
- *
- * Copyright (c) 2011 Intel Corporation
- *  Written by:
- *  Jiang Yunhong<yunhong.jiang@intel.com>
- *  Xin Xiaohui<xiaohui.xin@intel.com>
- *  Zhang Xiantao<xiantao.zhang@intel.com>
- *
- * This work is licensed under the terms of the GNU GPL, version 2 or later.
- * See the COPYING file in the top-level directory.
- *
- */
-
-#include "qemu/osdep.h"
-#include "qemu/error-report.h"
-#include "qemu/main-loop.h"
-#include "sysemu/runstate.h"
-#include "sysemu/cpus.h"
-#include "qemu/guest-random.h"
-
-#include "hax-cpus.h"
-
-static void *hax_cpu_thread_fn(void *arg)
-{
-    CPUState *cpu = arg;
-    int r;
-
-    rcu_register_thread();
-    qemu_mutex_lock_iothread();
-    qemu_thread_get_self(cpu->thread);
-
-    cpu->thread_id = qemu_get_thread_id();
-    current_cpu = cpu;
-    hax_init_vcpu(cpu);
-    cpu_thread_signal_created(cpu);
-    qemu_guest_random_seed_thread_part2(cpu->random_seed);
-
-    do {
-        if (cpu_can_run(cpu)) {
-            r = hax_smp_cpu_exec(cpu);
-            if (r == EXCP_DEBUG) {
-                cpu_handle_guest_debug(cpu);
-            }
-        }
-
-        qemu_wait_io_event(cpu);
-    } while (!cpu->unplug || cpu_can_run(cpu));
-    rcu_unregister_thread();
-    return NULL;
-}
-
-static void hax_start_vcpu_thread(CPUState *cpu)
-{
-    char thread_name[VCPU_THREAD_NAME_SIZE];
-
-    cpu->thread = g_malloc0(sizeof(QemuThread));
-    cpu->halt_cond = g_malloc0(sizeof(QemuCond));
-    qemu_cond_init(cpu->halt_cond);
-
-    snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/HAX",
-             cpu->cpu_index);
-    qemu_thread_create(cpu->thread, thread_name, hax_cpu_thread_fn,
-                       cpu, QEMU_THREAD_JOINABLE);
-#ifdef _WIN32
-    cpu->hThread = qemu_thread_get_handle(cpu->thread);
-#endif
-}
-
-const CpusAccel hax_cpus = {
-    .create_vcpu_thread = hax_start_vcpu_thread,
-    .kick_vcpu_thread = hax_kick_vcpu_thread,
-
-    .synchronize_post_reset = hax_cpu_synchronize_post_reset,
-    .synchronize_post_init = hax_cpu_synchronize_post_init,
-    .synchronize_state = hax_cpu_synchronize_state,
-    .synchronize_pre_loadvm = hax_cpu_synchronize_pre_loadvm,
-};
diff --git a/target/i386/hax-cpus.h b/target/i386/hax-cpus.h
deleted file mode 100644 (file)
index ee8ab7a..0000000
+++ /dev/null
@@ -1,33 +0,0 @@
-/*
- * Accelerator CPUS Interface
- *
- * Copyright 2020 SUSE LLC
- *
- * This work is licensed under the terms of the GNU GPL, version 2 or later.
- * See the COPYING file in the top-level directory.
- */
-
-#ifndef HAX_CPUS_H
-#define HAX_CPUS_H
-
-#include "sysemu/cpus.h"
-
-extern const CpusAccel hax_cpus;
-
-#include "hax-interface.h"
-#include "hax-i386.h"
-
-int hax_init_vcpu(CPUState *cpu);
-int hax_smp_cpu_exec(CPUState *cpu);
-int hax_populate_ram(uint64_t va, uint64_t size);
-
-void hax_cpu_synchronize_state(CPUState *cpu);
-void hax_cpu_synchronize_post_reset(CPUState *cpu);
-void hax_cpu_synchronize_post_init(CPUState *cpu);
-void hax_cpu_synchronize_pre_loadvm(CPUState *cpu);
-
-int hax_vcpu_destroy(CPUState *cpu);
-void hax_raise_event(CPUState *cpu);
-void hax_reset_vcpu_state(void *opaque);
-
-#endif /* HAX_CPUS_H */
diff --git a/target/i386/hax-i386.h b/target/i386/hax-i386.h
deleted file mode 100644 (file)
index 48c4abe..0000000
+++ /dev/null
@@ -1,96 +0,0 @@
-/*
- * QEMU HAXM support
- *
- * Copyright (c) 2011 Intel Corporation
- *  Written by:
- *  Jiang Yunhong<yunhong.jiang@intel.com>
- *
- * This work is licensed under the terms of the GNU GPL, version 2 or later.
- * See the COPYING file in the top-level directory.
- *
- */
-
-#ifndef HAX_I386_H
-#define HAX_I386_H
-
-#include "cpu.h"
-#include "sysemu/hax.h"
-
-#ifdef CONFIG_POSIX
-typedef int hax_fd;
-#endif
-
-#ifdef CONFIG_WIN32
-typedef HANDLE hax_fd;
-#endif
-
-extern struct hax_state hax_global;
-struct hax_vcpu_state {
-    hax_fd fd;
-    int vcpu_id;
-    struct hax_tunnel *tunnel;
-    unsigned char *iobuf;
-};
-
-struct hax_state {
-    hax_fd fd; /* the global hax device interface */
-    uint32_t version;
-    struct hax_vm *vm;
-    uint64_t mem_quota;
-    bool supports_64bit_ramblock;
-};
-
-#define HAX_MAX_VCPU 0x10
-
-struct hax_vm {
-    hax_fd fd;
-    int id;
-    int numvcpus;
-    struct hax_vcpu_state **vcpus;
-};
-
-#ifdef NEED_CPU_H
-/* Functions exported to host specific mode */
-hax_fd hax_vcpu_get_fd(CPUArchState *env);
-int valid_hax_tunnel_size(uint16_t size);
-
-/* Host specific functions */
-int hax_mod_version(struct hax_state *hax, struct hax_module_version *version);
-int hax_inject_interrupt(CPUArchState *env, int vector);
-struct hax_vm *hax_vm_create(struct hax_state *hax, int max_cpus);
-int hax_vcpu_run(struct hax_vcpu_state *vcpu);
-int hax_vcpu_create(int id);
-void hax_kick_vcpu_thread(CPUState *cpu);
-
-int hax_sync_vcpu_state(CPUArchState *env, struct vcpu_state_t *state,
-                        int set);
-int hax_sync_msr(CPUArchState *env, struct hax_msr_data *msrs, int set);
-int hax_sync_fpu(CPUArchState *env, struct fx_layout *fl, int set);
-#endif
-
-int hax_vm_destroy(struct hax_vm *vm);
-int hax_capability(struct hax_state *hax, struct hax_capabilityinfo *cap);
-int hax_notify_qemu_version(hax_fd vm_fd, struct hax_qemu_version *qversion);
-int hax_set_ram(uint64_t start_pa, uint32_t size, uint64_t host_va, int flags);
-
-/* Common host function */
-int hax_host_create_vm(struct hax_state *hax, int *vm_id);
-hax_fd hax_host_open_vm(struct hax_state *hax, int vm_id);
-int hax_host_create_vcpu(hax_fd vm_fd, int vcpuid);
-hax_fd hax_host_open_vcpu(int vmid, int vcpuid);
-int hax_host_setup_vcpu_channel(struct hax_vcpu_state *vcpu);
-hax_fd hax_mod_open(void);
-void hax_memory_init(void);
-
-
-#ifdef CONFIG_POSIX
-#include "target/i386/hax-posix.h"
-#endif
-
-#ifdef CONFIG_WIN32
-#include "target/i386/hax-windows.h"
-#endif
-
-#include "target/i386/hax-interface.h"
-
-#endif
diff --git a/target/i386/hax-interface.h b/target/i386/hax-interface.h
deleted file mode 100644 (file)
index 537ae08..0000000
+++ /dev/null
@@ -1,369 +0,0 @@
-/*
- * QEMU HAXM support
- *
- * Copyright (c) 2011 Intel Corporation
- *  Written by:
- *  Jiang Yunhong<yunhong.jiang@intel.com>
- *  Xin Xiaohui<xiaohui.xin@intel.com>
- *  Zhang Xiantao<xiantao.zhang@intel.com>
- *
- * This work is licensed under the terms of the GNU GPL, version 2 or later.
- * See the COPYING file in the top-level directory.
- *
- */
-
-/* Interface with HAX kernel module */
-
-#ifndef HAX_INTERFACE_H
-#define HAX_INTERFACE_H
-
-/* fx_layout has 3 formats table 3-56, 512bytes */
-struct fx_layout {
-    uint16_t fcw;
-    uint16_t fsw;
-    uint8_t ftw;
-    uint8_t res1;
-    uint16_t fop;
-    union {
-        struct {
-            uint32_t fip;
-            uint16_t fcs;
-            uint16_t res2;
-        };
-        uint64_t fpu_ip;
-    };
-    union {
-        struct {
-            uint32_t fdp;
-            uint16_t fds;
-            uint16_t res3;
-        };
-        uint64_t fpu_dp;
-    };
-    uint32_t mxcsr;
-    uint32_t mxcsr_mask;
-    uint8_t st_mm[8][16];
-    uint8_t mmx_1[8][16];
-    uint8_t mmx_2[8][16];
-    uint8_t pad[96];
-} __attribute__ ((aligned(8)));
-
-struct vmx_msr {
-    uint64_t entry;
-    uint64_t value;
-} __attribute__ ((__packed__));
-
-/*
- * Fixed array is not good, but it makes Mac support a bit easier by avoiding
- * memory map or copyin staff.
- */
-#define HAX_MAX_MSR_ARRAY 0x20
-struct hax_msr_data {
-    uint16_t nr_msr;
-    uint16_t done;
-    uint16_t pad[2];
-    struct vmx_msr entries[HAX_MAX_MSR_ARRAY];
-} __attribute__ ((__packed__));
-
-union interruptibility_state_t {
-    uint32_t raw;
-    struct {
-        uint32_t sti_blocking:1;
-        uint32_t movss_blocking:1;
-        uint32_t smi_blocking:1;
-        uint32_t nmi_blocking:1;
-        uint32_t reserved:28;
-    };
-    uint64_t pad;
-};
-
-typedef union interruptibility_state_t interruptibility_state_t;
-
-/* Segment descriptor */
-struct segment_desc_t {
-    uint16_t selector;
-    uint16_t _dummy;
-    uint32_t limit;
-    uint64_t base;
-    union {
-        struct {
-            uint32_t type:4;
-            uint32_t desc:1;
-            uint32_t dpl:2;
-            uint32_t present:1;
-            uint32_t:4;
-            uint32_t available:1;
-            uint32_t long_mode:1;
-            uint32_t operand_size:1;
-            uint32_t granularity:1;
-            uint32_t null:1;
-            uint32_t:15;
-        };
-        uint32_t ar;
-    };
-    uint32_t ipad;
-};
-
-typedef struct segment_desc_t segment_desc_t;
-
-struct vcpu_state_t {
-    union {
-        uint64_t _regs[16];
-        struct {
-            union {
-                struct {
-                    uint8_t _al, _ah;
-                };
-                uint16_t _ax;
-                uint32_t _eax;
-                uint64_t _rax;
-            };
-            union {
-                struct {
-                    uint8_t _cl, _ch;
-                };
-                uint16_t _cx;
-                uint32_t _ecx;
-                uint64_t _rcx;
-            };
-            union {
-                struct {
-                    uint8_t _dl, _dh;
-                };
-                uint16_t _dx;
-                uint32_t _edx;
-                uint64_t _rdx;
-            };
-            union {
-                struct {
-                    uint8_t _bl, _bh;
-                };
-                uint16_t _bx;
-                uint32_t _ebx;
-                uint64_t _rbx;
-            };
-            union {
-                uint16_t _sp;
-                uint32_t _esp;
-                uint64_t _rsp;
-            };
-            union {
-                uint16_t _bp;
-                uint32_t _ebp;
-                uint64_t _rbp;
-            };
-            union {
-                uint16_t _si;
-                uint32_t _esi;
-                uint64_t _rsi;
-            };
-            union {
-                uint16_t _di;
-                uint32_t _edi;
-                uint64_t _rdi;
-            };
-
-            uint64_t _r8;
-            uint64_t _r9;
-            uint64_t _r10;
-            uint64_t _r11;
-            uint64_t _r12;
-            uint64_t _r13;
-            uint64_t _r14;
-            uint64_t _r15;
-        };
-    };
-
-    union {
-        uint32_t _eip;
-        uint64_t _rip;
-    };
-
-    union {
-        uint32_t _eflags;
-        uint64_t _rflags;
-    };
-
-    segment_desc_t _cs;
-    segment_desc_t _ss;
-    segment_desc_t _ds;
-    segment_desc_t _es;
-    segment_desc_t _fs;
-    segment_desc_t _gs;
-    segment_desc_t _ldt;
-    segment_desc_t _tr;
-
-    segment_desc_t _gdt;
-    segment_desc_t _idt;
-
-    uint64_t _cr0;
-    uint64_t _cr2;
-    uint64_t _cr3;
-    uint64_t _cr4;
-
-    uint64_t _dr0;
-    uint64_t _dr1;
-    uint64_t _dr2;
-    uint64_t _dr3;
-    uint64_t _dr6;
-    uint64_t _dr7;
-    uint64_t _pde;
-
-    uint32_t _efer;
-
-    uint32_t _sysenter_cs;
-    uint64_t _sysenter_eip;
-    uint64_t _sysenter_esp;
-
-    uint32_t _activity_state;
-    uint32_t pad;
-    interruptibility_state_t _interruptibility_state;
-};
-
-/* HAX exit status */
-enum exit_status {
-    /* IO port request */
-    HAX_EXIT_IO = 1,
-    /* MMIO instruction emulation */
-    HAX_EXIT_MMIO,
-    /* QEMU emulation mode request, currently means guest enter non-PG mode */
-    HAX_EXIT_REAL,
-    /*
-     * Interrupt window open, qemu can inject interrupt now
-     * Also used when signal pending since at that time qemu usually need
-     * check interrupt
-     */
-    HAX_EXIT_INTERRUPT,
-    /* Unknown vmexit, mostly trigger reboot */
-    HAX_EXIT_UNKNOWN_VMEXIT,
-    /* HALT from guest */
-    HAX_EXIT_HLT,
-    /* Reboot request, like because of tripple fault in guest */
-    HAX_EXIT_STATECHANGE,
-    /* the vcpu is now only paused when destroy, so simply return to hax */
-    HAX_EXIT_PAUSED,
-    HAX_EXIT_FAST_MMIO,
-};
-
-/*
- * The interface definition:
- * 1. vcpu_run execute will return 0 on success, otherwise mean failed
- * 2. exit_status return the exit reason, as stated in enum exit_status
- * 3. exit_reason is the vmx exit reason
- */
-struct hax_tunnel {
-    uint32_t _exit_reason;
-    uint32_t _exit_flag;
-    uint32_t _exit_status;
-    uint32_t user_event_pending;
-    int ready_for_interrupt_injection;
-    int request_interrupt_window;
-    union {
-        struct {
-            /* 0: read, 1: write */
-#define HAX_EXIT_IO_IN  1
-#define HAX_EXIT_IO_OUT 0
-            uint8_t _direction;
-            uint8_t _df;
-            uint16_t _size;
-            uint16_t _port;
-            uint16_t _count;
-            uint8_t _flags;
-            uint8_t _pad0;
-            uint16_t _pad1;
-            uint32_t _pad2;
-            uint64_t _vaddr;
-        } pio;
-        struct {
-            uint64_t gla;
-        } mmio;
-        struct {
-        } state;
-    };
-} __attribute__ ((__packed__));
-
-struct hax_module_version {
-    uint32_t compat_version;
-    uint32_t cur_version;
-} __attribute__ ((__packed__));
-
-/* This interface is support only after API version 2 */
-struct hax_qemu_version {
-    /* Current API version in QEMU */
-    uint32_t cur_version;
-    /* The minimum API version supported by QEMU */
-    uint32_t min_version;
-} __attribute__ ((__packed__));
-
-/* The mac specfic interface to qemu, mostly is ioctl related */
-struct hax_tunnel_info {
-    uint64_t va;
-    uint64_t io_va;
-    uint16_t size;
-    uint16_t pad[3];
-} __attribute__ ((__packed__));
-
-struct hax_alloc_ram_info {
-    uint32_t size;
-    uint32_t pad;
-    uint64_t va;
-} __attribute__ ((__packed__));
-
-struct hax_ramblock_info {
-    uint64_t start_va;
-    uint64_t size;
-    uint64_t reserved;
-} __attribute__ ((__packed__));
-
-#define HAX_RAM_INFO_ROM     0x01 /* Read-Only */
-#define HAX_RAM_INFO_INVALID 0x80 /* Unmapped, usually used for MMIO */
-struct hax_set_ram_info {
-    uint64_t pa_start;
-    uint32_t size;
-    uint8_t flags;
-    uint8_t pad[3];
-    uint64_t va;
-} __attribute__ ((__packed__));
-
-#define HAX_CAP_STATUS_WORKING     0x1
-#define HAX_CAP_STATUS_NOTWORKING  0x0
-#define HAX_CAP_WORKSTATUS_MASK    0x1
-
-#define HAX_CAP_FAILREASON_VT      0x1
-#define HAX_CAP_FAILREASON_NX      0x2
-
-#define HAX_CAP_MEMQUOTA           0x2
-#define HAX_CAP_UG                 0x4
-#define HAX_CAP_64BIT_RAMBLOCK     0x8
-
-struct hax_capabilityinfo {
-    /* bit 0: 1 - working
-     *        0 - not working, possibly because NT/NX disabled
-     * bit 1: 1 - memory limitation working
-     *        0 - no memory limitation
-     */
-    uint16_t wstatus;
-    /* valid when not working
-     * bit 0: VT not enabeld
-     * bit 1: NX not enabled*/
-    uint16_t winfo;
-    uint32_t pad;
-    uint64_t mem_quota;
-} __attribute__ ((__packed__));
-
-struct hax_fastmmio {
-    uint64_t gpa;
-    union {
-        uint64_t value;
-        uint64_t gpa2;  /* since HAX API v4 */
-    };
-    uint8_t size;
-    uint8_t direction;
-    uint16_t reg_index;
-    uint32_t pad0;
-    uint64_t _cr0;
-    uint64_t _cr2;
-    uint64_t _cr3;
-    uint64_t _cr4;
-} __attribute__ ((__packed__));
-#endif
diff --git a/target/i386/hax-mem.c b/target/i386/hax-mem.c
deleted file mode 100644 (file)
index 71e637c..0000000
+++ /dev/null
@@ -1,321 +0,0 @@
-/*
- * HAX memory mapping operations
- *
- * Copyright (c) 2015-16 Intel Corporation
- * Copyright 2016 Google, Inc.
- *
- * This work is licensed under the terms of the GNU GPL, version 2.  See
- * the COPYING file in the top-level directory.
- */
-
-#include "qemu/osdep.h"
-#include "cpu.h"
-#include "exec/address-spaces.h"
-#include "qemu/error-report.h"
-
-#include "hax-cpus.h"
-#include "qemu/queue.h"
-
-#define DEBUG_HAX_MEM 0
-
-#define DPRINTF(fmt, ...) \
-    do { \
-        if (DEBUG_HAX_MEM) { \
-            fprintf(stdout, fmt, ## __VA_ARGS__); \
-        } \
-    } while (0)
-
-/**
- * HAXMapping: describes a pending guest physical memory mapping
- *
- * @start_pa: a guest physical address marking the start of the region; must be
- *            page-aligned
- * @size: a guest physical address marking the end of the region; must be
- *          page-aligned
- * @host_va: the host virtual address of the start of the mapping
- * @flags: mapping parameters e.g. HAX_RAM_INFO_ROM or HAX_RAM_INFO_INVALID
- * @entry: additional fields for linking #HAXMapping instances together
- */
-typedef struct HAXMapping {
-    uint64_t start_pa;
-    uint32_t size;
-    uint64_t host_va;
-    int flags;
-    QTAILQ_ENTRY(HAXMapping) entry;
-} HAXMapping;
-
-/*
- * A doubly-linked list (actually a tail queue) of the pending page mappings
- * for the ongoing memory transaction.
- *
- * It is used to optimize the number of page mapping updates done through the
- * kernel module. For example, it's effective when a driver is digging an MMIO
- * hole inside an existing memory mapping. It will get a deletion of the whole
- * region, then the addition of the 2 remaining RAM areas around the hole and
- * finally the memory transaction commit. During the commit, it will effectively
- * send to the kernel only the removal of the pages from the MMIO hole after
- * having computed locally the result of the deletion and additions.
- */
-static QTAILQ_HEAD(, HAXMapping) mappings =
-    QTAILQ_HEAD_INITIALIZER(mappings);
-
-/**
- * hax_mapping_dump_list: dumps @mappings to stdout (for debugging)
- */
-static void hax_mapping_dump_list(void)
-{
-    HAXMapping *entry;
-
-    DPRINTF("%s updates:\n", __func__);
-    QTAILQ_FOREACH(entry, &mappings, entry) {
-        DPRINTF("\t%c 0x%016" PRIx64 "->0x%016" PRIx64 " VA 0x%016" PRIx64
-                "%s\n", entry->flags & HAX_RAM_INFO_INVALID ? '-' : '+',
-                entry->start_pa, entry->start_pa + entry->size, entry->host_va,
-                entry->flags & HAX_RAM_INFO_ROM ? " ROM" : "");
-    }
-}
-
-static void hax_insert_mapping_before(HAXMapping *next, uint64_t start_pa,
-                                      uint32_t size, uint64_t host_va,
-                                      uint8_t flags)
-{
-    HAXMapping *entry;
-
-    entry = g_malloc0(sizeof(*entry));
-    entry->start_pa = start_pa;
-    entry->size = size;
-    entry->host_va = host_va;
-    entry->flags = flags;
-    if (!next) {
-        QTAILQ_INSERT_TAIL(&mappings, entry, entry);
-    } else {
-        QTAILQ_INSERT_BEFORE(next, entry, entry);
-    }
-}
-
-static bool hax_mapping_is_opposite(HAXMapping *entry, uint64_t host_va,
-                                    uint8_t flags)
-{
-    /* removed then added without change for the read-only flag */
-    bool nop_flags = (entry->flags ^ flags) == HAX_RAM_INFO_INVALID;
-
-    return (entry->host_va == host_va) && nop_flags;
-}
-
-static void hax_update_mapping(uint64_t start_pa, uint32_t size,
-                               uint64_t host_va, uint8_t flags)
-{
-    uint64_t end_pa = start_pa + size;
-    HAXMapping *entry, *next;
-
-    QTAILQ_FOREACH_SAFE(entry, &mappings, entry, next) {
-        uint32_t chunk_sz;
-        if (start_pa >= entry->start_pa + entry->size) {
-            continue;
-        }
-        if (start_pa < entry->start_pa) {
-            chunk_sz = end_pa <= entry->start_pa ? size
-                                                 : entry->start_pa - start_pa;
-            hax_insert_mapping_before(entry, start_pa, chunk_sz,
-                                      host_va, flags);
-            start_pa += chunk_sz;
-            host_va += chunk_sz;
-            size -= chunk_sz;
-        } else if (start_pa > entry->start_pa) {
-            /* split the existing chunk at start_pa */
-            chunk_sz = start_pa - entry->start_pa;
-            hax_insert_mapping_before(entry, entry->start_pa, chunk_sz,
-                                      entry->host_va, entry->flags);
-            entry->start_pa += chunk_sz;
-            entry->host_va += chunk_sz;
-            entry->size -= chunk_sz;
-        }
-        /* now start_pa == entry->start_pa */
-        chunk_sz = MIN(size, entry->size);
-        if (chunk_sz) {
-            bool nop = hax_mapping_is_opposite(entry, host_va, flags);
-            bool partial = chunk_sz < entry->size;
-            if (partial) {
-                /* remove the beginning of the existing chunk */
-                entry->start_pa += chunk_sz;
-                entry->host_va += chunk_sz;
-                entry->size -= chunk_sz;
-                if (!nop) {
-                    hax_insert_mapping_before(entry, start_pa, chunk_sz,
-                                              host_va, flags);
-                }
-            } else { /* affects the full mapping entry */
-                if (nop) { /* no change to this mapping, remove it */
-                    QTAILQ_REMOVE(&mappings, entry, entry);
-                    g_free(entry);
-                } else { /* update mapping properties */
-                    entry->host_va = host_va;
-                    entry->flags = flags;
-                }
-            }
-            start_pa += chunk_sz;
-            host_va += chunk_sz;
-            size -= chunk_sz;
-        }
-        if (!size) { /* we are done */
-            break;
-        }
-    }
-    if (size) { /* add the leftover */
-        hax_insert_mapping_before(NULL, start_pa, size, host_va, flags);
-    }
-}
-
-static void hax_process_section(MemoryRegionSection *section, uint8_t flags)
-{
-    MemoryRegion *mr = section->mr;
-    hwaddr start_pa = section->offset_within_address_space;
-    ram_addr_t size = int128_get64(section->size);
-    unsigned int delta;
-    uint64_t host_va;
-    uint32_t max_mapping_size;
-
-    /* We only care about RAM and ROM regions */
-    if (!memory_region_is_ram(mr)) {
-        if (memory_region_is_romd(mr)) {
-            /* HAXM kernel module does not support ROMD yet  */
-            warn_report("Ignoring ROMD region 0x%016" PRIx64 "->0x%016" PRIx64,
-                        start_pa, start_pa + size);
-        }
-        return;
-    }
-
-    /* Adjust start_pa and size so that they are page-aligned. (Cf
-     * kvm_set_phys_mem() in kvm-all.c).
-     */
-    delta = qemu_real_host_page_size - (start_pa & ~qemu_real_host_page_mask);
-    delta &= ~qemu_real_host_page_mask;
-    if (delta > size) {
-        return;
-    }
-    start_pa += delta;
-    size -= delta;
-    size &= qemu_real_host_page_mask;
-    if (!size || (start_pa & ~qemu_real_host_page_mask)) {
-        return;
-    }
-
-    host_va = (uintptr_t)memory_region_get_ram_ptr(mr)
-            + section->offset_within_region + delta;
-    if (memory_region_is_rom(section->mr)) {
-        flags |= HAX_RAM_INFO_ROM;
-    }
-
-    /*
-     * The kernel module interface uses 32-bit sizes:
-     * https://github.com/intel/haxm/blob/master/API.md#hax_vm_ioctl_set_ram
-     *
-     * If the mapping size is longer than 32 bits, we can't process it in one
-     * call into the kernel. Instead, we split the mapping into smaller ones,
-     * and call hax_update_mapping() on each.
-     */
-    max_mapping_size = UINT32_MAX & qemu_real_host_page_mask;
-    while (size > max_mapping_size) {
-        hax_update_mapping(start_pa, max_mapping_size, host_va, flags);
-        start_pa += max_mapping_size;
-        size -= max_mapping_size;
-        host_va += max_mapping_size;
-    }
-    /* Now size <= max_mapping_size */
-    hax_update_mapping(start_pa, (uint32_t)size, host_va, flags);
-}
-
-static void hax_region_add(MemoryListener *listener,
-                           MemoryRegionSection *section)
-{
-    memory_region_ref(section->mr);
-    hax_process_section(section, 0);
-}
-
-static void hax_region_del(MemoryListener *listener,
-                           MemoryRegionSection *section)
-{
-    hax_process_section(section, HAX_RAM_INFO_INVALID);
-    memory_region_unref(section->mr);
-}
-
-static void hax_transaction_begin(MemoryListener *listener)
-{
-    g_assert(QTAILQ_EMPTY(&mappings));
-}
-
-static void hax_transaction_commit(MemoryListener *listener)
-{
-    if (!QTAILQ_EMPTY(&mappings)) {
-        HAXMapping *entry, *next;
-
-        if (DEBUG_HAX_MEM) {
-            hax_mapping_dump_list();
-        }
-        QTAILQ_FOREACH_SAFE(entry, &mappings, entry, next) {
-            if (entry->flags & HAX_RAM_INFO_INVALID) {
-                /* for unmapping, put the values expected by the kernel */
-                entry->flags = HAX_RAM_INFO_INVALID;
-                entry->host_va = 0;
-            }
-            if (hax_set_ram(entry->start_pa, entry->size,
-                            entry->host_va, entry->flags)) {
-                fprintf(stderr, "%s: Failed mapping @0x%016" PRIx64 "+0x%"
-                        PRIx32 " flags %02x\n", __func__, entry->start_pa,
-                        entry->size, entry->flags);
-            }
-            QTAILQ_REMOVE(&mappings, entry, entry);
-            g_free(entry);
-        }
-    }
-}
-
-/* currently we fake the dirty bitmap sync, always dirty */
-static void hax_log_sync(MemoryListener *listener,
-                         MemoryRegionSection *section)
-{
-    MemoryRegion *mr = section->mr;
-
-    if (!memory_region_is_ram(mr)) {
-        /* Skip MMIO regions */
-        return;
-    }
-
-    memory_region_set_dirty(mr, 0, int128_get64(section->size));
-}
-
-static MemoryListener hax_memory_listener = {
-    .begin = hax_transaction_begin,
-    .commit = hax_transaction_commit,
-    .region_add = hax_region_add,
-    .region_del = hax_region_del,
-    .log_sync = hax_log_sync,
-    .priority = 10,
-};
-
-static void hax_ram_block_added(RAMBlockNotifier *n, void *host, size_t size)
-{
-    /*
-     * We must register each RAM block with the HAXM kernel module, or
-     * hax_set_ram() will fail for any mapping into the RAM block:
-     * https://github.com/intel/haxm/blob/master/API.md#hax_vm_ioctl_alloc_ram
-     *
-     * Old versions of the HAXM kernel module (< 6.2.0) used to preallocate all
-     * host physical pages for the RAM block as part of this registration
-     * process, hence the name hax_populate_ram().
-     */
-    if (hax_populate_ram((uint64_t)(uintptr_t)host, size) < 0) {
-        fprintf(stderr, "HAX failed to populate RAM\n");
-        abort();
-    }
-}
-
-static struct RAMBlockNotifier hax_ram_notifier = {
-    .ram_block_added = hax_ram_block_added,
-};
-
-void hax_memory_init(void)
-{
-    ram_block_notifier_add(&hax_ram_notifier);
-    memory_listener_register(&hax_memory_listener, &address_space_memory);
-}
diff --git a/target/i386/hax-posix.c b/target/i386/hax-posix.c
deleted file mode 100644 (file)
index 735a749..0000000
+++ /dev/null
@@ -1,305 +0,0 @@
-/*
- * QEMU HAXM support
- *
- * Copyright (c) 2011 Intel Corporation
- *  Written by:
- *  Jiang Yunhong<yunhong.jiang@intel.com>
- *
- * This work is licensed under the terms of the GNU GPL, version 2 or later.
- * See the COPYING file in the top-level directory.
- *
- */
-
-/* HAX module interface - darwin version */
-#include "qemu/osdep.h"
-#include <sys/ioctl.h>
-
-#include "sysemu/cpus.h"
-#include "hax-cpus.h"
-
-hax_fd hax_mod_open(void)
-{
-    int fd = open("/dev/HAX", O_RDWR);
-    if (fd == -1) {
-        fprintf(stderr, "Failed to open the hax module\n");
-    }
-
-    qemu_set_cloexec(fd);
-
-    return fd;
-}
-
-int hax_populate_ram(uint64_t va, uint64_t size)
-{
-    int ret;
-
-    if (!hax_global.vm || !hax_global.vm->fd) {
-        fprintf(stderr, "Allocate memory before vm create?\n");
-        return -EINVAL;
-    }
-
-    if (hax_global.supports_64bit_ramblock) {
-        struct hax_ramblock_info ramblock = {
-            .start_va = va,
-            .size = size,
-            .reserved = 0
-        };
-
-        ret = ioctl(hax_global.vm->fd, HAX_VM_IOCTL_ADD_RAMBLOCK, &ramblock);
-    } else {
-        struct hax_alloc_ram_info info = {
-            .size = (uint32_t)size,
-            .pad = 0,
-            .va = va
-        };
-
-        ret = ioctl(hax_global.vm->fd, HAX_VM_IOCTL_ALLOC_RAM, &info);
-    }
-    if (ret < 0) {
-        fprintf(stderr, "Failed to register RAM block: ret=%d, va=0x%" PRIx64
-                ", size=0x%" PRIx64 ", method=%s\n", ret, va, size,
-                hax_global.supports_64bit_ramblock ? "new" : "legacy");
-        return ret;
-    }
-    return 0;
-}
-
-int hax_set_ram(uint64_t start_pa, uint32_t size, uint64_t host_va, int flags)
-{
-    struct hax_set_ram_info info;
-    int ret;
-
-    info.pa_start = start_pa;
-    info.size = size;
-    info.va = host_va;
-    info.flags = (uint8_t) flags;
-
-    ret = ioctl(hax_global.vm->fd, HAX_VM_IOCTL_SET_RAM, &info);
-    if (ret < 0) {
-        return -errno;
-    }
-    return 0;
-}
-
-int hax_capability(struct hax_state *hax, struct hax_capabilityinfo *cap)
-{
-    int ret;
-
-    ret = ioctl(hax->fd, HAX_IOCTL_CAPABILITY, cap);
-    if (ret == -1) {
-        fprintf(stderr, "Failed to get HAX capability\n");
-        return -errno;
-    }
-
-    return 0;
-}
-
-int hax_mod_version(struct hax_state *hax, struct hax_module_version *version)
-{
-    int ret;
-
-    ret = ioctl(hax->fd, HAX_IOCTL_VERSION, version);
-    if (ret == -1) {
-        fprintf(stderr, "Failed to get HAX version\n");
-        return -errno;
-    }
-
-    return 0;
-}
-
-static char *hax_vm_devfs_string(int vm_id)
-{
-    return g_strdup_printf("/dev/hax_vm/vm%02d", vm_id);
-}
-
-static char *hax_vcpu_devfs_string(int vm_id, int vcpu_id)
-{
-    return g_strdup_printf("/dev/hax_vm%02d/vcpu%02d", vm_id, vcpu_id);
-}
-
-int hax_host_create_vm(struct hax_state *hax, int *vmid)
-{
-    int ret;
-    int vm_id = 0;
-
-    if (hax_invalid_fd(hax->fd)) {
-        return -EINVAL;
-    }
-
-    if (hax->vm) {
-        return 0;
-    }
-
-    ret = ioctl(hax->fd, HAX_IOCTL_CREATE_VM, &vm_id);
-    *vmid = vm_id;
-    return ret;
-}
-
-hax_fd hax_host_open_vm(struct hax_state *hax, int vm_id)
-{
-    hax_fd fd;
-    char *vm_name = NULL;
-
-    vm_name = hax_vm_devfs_string(vm_id);
-    if (!vm_name) {
-        return -1;
-    }
-
-    fd = open(vm_name, O_RDWR);
-    g_free(vm_name);
-
-    qemu_set_cloexec(fd);
-
-    return fd;
-}
-
-int hax_notify_qemu_version(hax_fd vm_fd, struct hax_qemu_version *qversion)
-{
-    int ret;
-
-    if (hax_invalid_fd(vm_fd)) {
-        return -EINVAL;
-    }
-
-    ret = ioctl(vm_fd, HAX_VM_IOCTL_NOTIFY_QEMU_VERSION, qversion);
-
-    if (ret < 0) {
-        fprintf(stderr, "Failed to notify qemu API version\n");
-        return ret;
-    }
-    return 0;
-}
-
-/* Simply assume the size should be bigger than the hax_tunnel,
- * since the hax_tunnel can be extended later with compatibility considered
- */
-int hax_host_create_vcpu(hax_fd vm_fd, int vcpuid)
-{
-    int ret;
-
-    ret = ioctl(vm_fd, HAX_VM_IOCTL_VCPU_CREATE, &vcpuid);
-    if (ret < 0) {
-        fprintf(stderr, "Failed to create vcpu %x\n", vcpuid);
-    }
-
-    return ret;
-}
-
-hax_fd hax_host_open_vcpu(int vmid, int vcpuid)
-{
-    char *devfs_path = NULL;
-    hax_fd fd;
-
-    devfs_path = hax_vcpu_devfs_string(vmid, vcpuid);
-    if (!devfs_path) {
-        fprintf(stderr, "Failed to get the devfs\n");
-        return -EINVAL;
-    }
-
-    fd = open(devfs_path, O_RDWR);
-    g_free(devfs_path);
-    if (fd < 0) {
-        fprintf(stderr, "Failed to open the vcpu devfs\n");
-    }
-    qemu_set_cloexec(fd);
-    return fd;
-}
-
-int hax_host_setup_vcpu_channel(struct hax_vcpu_state *vcpu)
-{
-    int ret;
-    struct hax_tunnel_info info;
-
-    ret = ioctl(vcpu->fd, HAX_VCPU_IOCTL_SETUP_TUNNEL, &info);
-    if (ret) {
-        fprintf(stderr, "Failed to setup the hax tunnel\n");
-        return ret;
-    }
-
-    if (!valid_hax_tunnel_size(info.size)) {
-        fprintf(stderr, "Invalid hax tunnel size %x\n", info.size);
-        ret = -EINVAL;
-        return ret;
-    }
-
-    vcpu->tunnel = (struct hax_tunnel *) (intptr_t) (info.va);
-    vcpu->iobuf = (unsigned char *) (intptr_t) (info.io_va);
-    return 0;
-}
-
-int hax_vcpu_run(struct hax_vcpu_state *vcpu)
-{
-    return ioctl(vcpu->fd, HAX_VCPU_IOCTL_RUN, NULL);
-}
-
-int hax_sync_fpu(CPUArchState *env, struct fx_layout *fl, int set)
-{
-    int ret, fd;
-
-    fd = hax_vcpu_get_fd(env);
-    if (fd <= 0) {
-        return -1;
-    }
-
-    if (set) {
-        ret = ioctl(fd, HAX_VCPU_IOCTL_SET_FPU, fl);
-    } else {
-        ret = ioctl(fd, HAX_VCPU_IOCTL_GET_FPU, fl);
-    }
-    return ret;
-}
-
-int hax_sync_msr(CPUArchState *env, struct hax_msr_data *msrs, int set)
-{
-    int ret, fd;
-
-    fd = hax_vcpu_get_fd(env);
-    if (fd <= 0) {
-        return -1;
-    }
-    if (set) {
-        ret = ioctl(fd, HAX_VCPU_IOCTL_SET_MSRS, msrs);
-    } else {
-        ret = ioctl(fd, HAX_VCPU_IOCTL_GET_MSRS, msrs);
-    }
-    return ret;
-}
-
-int hax_sync_vcpu_state(CPUArchState *env, struct vcpu_state_t *state, int set)
-{
-    int ret, fd;
-
-    fd = hax_vcpu_get_fd(env);
-    if (fd <= 0) {
-        return -1;
-    }
-
-    if (set) {
-        ret = ioctl(fd, HAX_VCPU_SET_REGS, state);
-    } else {
-        ret = ioctl(fd, HAX_VCPU_GET_REGS, state);
-    }
-    return ret;
-}
-
-int hax_inject_interrupt(CPUArchState *env, int vector)
-{
-    int fd;
-
-    fd = hax_vcpu_get_fd(env);
-    if (fd <= 0) {
-        return -1;
-    }
-
-    return ioctl(fd, HAX_VCPU_IOCTL_INTERRUPT, &vector);
-}
-
-void hax_kick_vcpu_thread(CPUState *cpu)
-{
-    /*
-     * FIXME: race condition with the exit_request check in
-     * hax_vcpu_hax_exec
-     */
-    cpu->exit_request = 1;
-    cpus_kick_thread(cpu);
-}
diff --git a/target/i386/hax-posix.h b/target/i386/hax-posix.h
deleted file mode 100644 (file)
index fb7c644..0000000
+++ /dev/null
@@ -1,61 +0,0 @@
-/*
- * QEMU HAXM support
- *
- * Copyright (c) 2011 Intel Corporation
- *  Written by:
- *  Jiang Yunhong<yunhong.jiang@intel.com>
- *  Xin Xiaohui<xiaohui.xin@intel.com>
- *  Zhang Xiantao<xiantao.zhang@intel.com>
- *
- * This work is licensed under the terms of the GNU GPL, version 2 or later.
- * See the COPYING file in the top-level directory.
- *
- */
-
-#ifndef TARGET_I386_HAX_POSIX_H
-#define TARGET_I386_HAX_POSIX_H
-
-#include <sys/ioctl.h>
-
-#define HAX_INVALID_FD  (-1)
-static inline int hax_invalid_fd(hax_fd fd)
-{
-    return fd <= 0;
-}
-
-static inline void hax_mod_close(struct hax_state *hax)
-{
-    close(hax->fd);
-}
-
-static inline void hax_close_fd(hax_fd fd)
-{
-    close(fd);
-}
-
-/* HAX model level ioctl */
-#define HAX_IOCTL_VERSION _IOWR(0, 0x20, struct hax_module_version)
-#define HAX_IOCTL_CREATE_VM _IOWR(0, 0x21, uint32_t)
-#define HAX_IOCTL_DESTROY_VM _IOW(0, 0x22, uint32_t)
-#define HAX_IOCTL_CAPABILITY _IOR(0, 0x23, struct hax_capabilityinfo)
-
-#define HAX_VM_IOCTL_VCPU_CREATE _IOWR(0, 0x80, uint32_t)
-#define HAX_VM_IOCTL_ALLOC_RAM _IOWR(0, 0x81, struct hax_alloc_ram_info)
-#define HAX_VM_IOCTL_SET_RAM _IOWR(0, 0x82, struct hax_set_ram_info)
-#define HAX_VM_IOCTL_VCPU_DESTROY _IOW(0, 0x83, uint32_t)
-#define HAX_VM_IOCTL_NOTIFY_QEMU_VERSION _IOW(0, 0x84, struct hax_qemu_version)
-#define HAX_VM_IOCTL_ADD_RAMBLOCK _IOW(0, 0x85, struct hax_ramblock_info)
-
-#define HAX_VCPU_IOCTL_RUN  _IO(0, 0xc0)
-#define HAX_VCPU_IOCTL_SET_MSRS _IOWR(0, 0xc1, struct hax_msr_data)
-#define HAX_VCPU_IOCTL_GET_MSRS _IOWR(0, 0xc2, struct hax_msr_data)
-
-#define HAX_VCPU_IOCTL_SET_FPU  _IOW(0, 0xc3, struct fx_layout)
-#define HAX_VCPU_IOCTL_GET_FPU  _IOR(0, 0xc4, struct fx_layout)
-
-#define HAX_VCPU_IOCTL_SETUP_TUNNEL _IOWR(0, 0xc5, struct hax_tunnel_info)
-#define HAX_VCPU_IOCTL_INTERRUPT _IOWR(0, 0xc6, uint32_t)
-#define HAX_VCPU_SET_REGS       _IOWR(0, 0xc7, struct vcpu_state_t)
-#define HAX_VCPU_GET_REGS       _IOWR(0, 0xc8, struct vcpu_state_t)
-
-#endif /* TARGET_I386_HAX_POSIX_H */
diff --git a/target/i386/hax-windows.c b/target/i386/hax-windows.c
deleted file mode 100644 (file)
index 6c82dfb..0000000
+++ /dev/null
@@ -1,485 +0,0 @@
-/*
- * QEMU HAXM support
- *
- * Copyright (c) 2011 Intel Corporation
- *  Written by:
- *  Jiang Yunhong<yunhong.jiang@intel.com>
- *
- * This work is licensed under the terms of the GNU GPL, version 2 or later.
- * See the COPYING file in the top-level directory.
- *
- */
-
-#include "qemu/osdep.h"
-#include "cpu.h"
-#include "hax-cpus.h"
-
-/*
- * return 0 when success, -1 when driver not loaded,
- * other negative value for other failure
- */
-static int hax_open_device(hax_fd *fd)
-{
-    uint32_t errNum = 0;
-    HANDLE hDevice;
-
-    if (!fd) {
-        return -2;
-    }
-
-    hDevice = CreateFile("\\\\.\\HAX",
-                         GENERIC_READ | GENERIC_WRITE,
-                         0, NULL, CREATE_ALWAYS, FILE_ATTRIBUTE_NORMAL, NULL);
-
-    if (hDevice == INVALID_HANDLE_VALUE) {
-        fprintf(stderr, "Failed to open the HAX device!\n");
-        errNum = GetLastError();
-        if (errNum == ERROR_FILE_NOT_FOUND) {
-            return -1;
-        }
-        return -2;
-    }
-    *fd = hDevice;
-    return 0;
-}
-
-/* hax_fd hax_mod_open */
- hax_fd hax_mod_open(void)
-{
-    int ret;
-    hax_fd fd = NULL;
-
-    ret = hax_open_device(&fd);
-    if (ret != 0) {
-        fprintf(stderr, "Open HAX device failed\n");
-    }
-
-    return fd;
-}
-
-int hax_populate_ram(uint64_t va, uint64_t size)
-{
-    int ret;
-    HANDLE hDeviceVM;
-    DWORD dSize = 0;
-
-    if (!hax_global.vm || !hax_global.vm->fd) {
-        fprintf(stderr, "Allocate memory before vm create?\n");
-        return -EINVAL;
-    }
-
-    hDeviceVM = hax_global.vm->fd;
-    if (hax_global.supports_64bit_ramblock) {
-        struct hax_ramblock_info ramblock = {
-            .start_va = va,
-            .size = size,
-            .reserved = 0
-        };
-
-        ret = DeviceIoControl(hDeviceVM,
-                              HAX_VM_IOCTL_ADD_RAMBLOCK,
-                              &ramblock, sizeof(ramblock), NULL, 0, &dSize,
-                              (LPOVERLAPPED) NULL);
-    } else {
-        struct hax_alloc_ram_info info = {
-            .size = (uint32_t) size,
-            .pad = 0,
-            .va = va
-        };
-
-        ret = DeviceIoControl(hDeviceVM,
-                              HAX_VM_IOCTL_ALLOC_RAM,
-                              &info, sizeof(info), NULL, 0, &dSize,
-                              (LPOVERLAPPED) NULL);
-    }
-
-    if (!ret) {
-        fprintf(stderr, "Failed to register RAM block: va=0x%" PRIx64
-                ", size=0x%" PRIx64 ", method=%s\n", va, size,
-                hax_global.supports_64bit_ramblock ? "new" : "legacy");
-        return ret;
-    }
-
-    return 0;
-}
-
-int hax_set_ram(uint64_t start_pa, uint32_t size, uint64_t host_va, int flags)
-{
-    struct hax_set_ram_info info;
-    HANDLE hDeviceVM = hax_global.vm->fd;
-    DWORD dSize = 0;
-    int ret;
-
-    info.pa_start = start_pa;
-    info.size = size;
-    info.va = host_va;
-    info.flags = (uint8_t) flags;
-
-    ret = DeviceIoControl(hDeviceVM, HAX_VM_IOCTL_SET_RAM,
-                          &info, sizeof(info), NULL, 0, &dSize,
-                          (LPOVERLAPPED) NULL);
-
-    if (!ret) {
-        return -EFAULT;
-    } else {
-        return 0;
-    }
-}
-
-int hax_capability(struct hax_state *hax, struct hax_capabilityinfo *cap)
-{
-    int ret;
-    HANDLE hDevice = hax->fd;        /* handle to hax module */
-    DWORD dSize = 0;
-    DWORD err = 0;
-
-    if (hax_invalid_fd(hDevice)) {
-        fprintf(stderr, "Invalid fd for hax device!\n");
-        return -ENODEV;
-    }
-
-    ret = DeviceIoControl(hDevice, HAX_IOCTL_CAPABILITY, NULL, 0, cap,
-                          sizeof(*cap), &dSize, (LPOVERLAPPED) NULL);
-
-    if (!ret) {
-        err = GetLastError();
-        if (err == ERROR_INSUFFICIENT_BUFFER || err == ERROR_MORE_DATA) {
-            fprintf(stderr, "hax capability is too long to hold.\n");
-        }
-        fprintf(stderr, "Failed to get Hax capability:%luu\n", err);
-        return -EFAULT;
-    } else {
-        return 0;
-    }
-}
-
-int hax_mod_version(struct hax_state *hax, struct hax_module_version *version)
-{
-    int ret;
-    HANDLE hDevice = hax->fd; /* handle to hax module */
-    DWORD dSize = 0;
-    DWORD err = 0;
-
-    if (hax_invalid_fd(hDevice)) {
-        fprintf(stderr, "Invalid fd for hax device!\n");
-        return -ENODEV;
-    }
-
-    ret = DeviceIoControl(hDevice,
-                          HAX_IOCTL_VERSION,
-                          NULL, 0,
-                          version, sizeof(*version), &dSize,
-                          (LPOVERLAPPED) NULL);
-
-    if (!ret) {
-        err = GetLastError();
-        if (err == ERROR_INSUFFICIENT_BUFFER || err == ERROR_MORE_DATA) {
-            fprintf(stderr, "hax module verion is too long to hold.\n");
-        }
-        fprintf(stderr, "Failed to get Hax module version:%lu\n", err);
-        return -EFAULT;
-    } else {
-        return 0;
-    }
-}
-
-static char *hax_vm_devfs_string(int vm_id)
-{
-    return g_strdup_printf("\\\\.\\hax_vm%02d", vm_id);
-}
-
-static char *hax_vcpu_devfs_string(int vm_id, int vcpu_id)
-{
-    return g_strdup_printf("\\\\.\\hax_vm%02d_vcpu%02d", vm_id, vcpu_id);
-}
-
-int hax_host_create_vm(struct hax_state *hax, int *vmid)
-{
-    int ret;
-    int vm_id = 0;
-    DWORD dSize = 0;
-
-    if (hax_invalid_fd(hax->fd)) {
-        return -EINVAL;
-    }
-
-    if (hax->vm) {
-        return 0;
-    }
-
-    ret = DeviceIoControl(hax->fd,
-                          HAX_IOCTL_CREATE_VM,
-                          NULL, 0, &vm_id, sizeof(vm_id), &dSize,
-                          (LPOVERLAPPED) NULL);
-    if (!ret) {
-        fprintf(stderr, "Failed to create VM. Error code: %lu\n",
-                GetLastError());
-        return -1;
-    }
-    *vmid = vm_id;
-    return 0;
-}
-
-hax_fd hax_host_open_vm(struct hax_state *hax, int vm_id)
-{
-    char *vm_name = NULL;
-    hax_fd hDeviceVM;
-
-    vm_name = hax_vm_devfs_string(vm_id);
-    if (!vm_name) {
-        fprintf(stderr, "Failed to open VM. VM name is null\n");
-        return INVALID_HANDLE_VALUE;
-    }
-
-    hDeviceVM = CreateFile(vm_name,
-                           GENERIC_READ | GENERIC_WRITE,
-                           0, NULL, CREATE_ALWAYS, FILE_ATTRIBUTE_NORMAL, NULL);
-    if (hDeviceVM == INVALID_HANDLE_VALUE) {
-        fprintf(stderr, "Open the vm device error:%s, ec:%lu\n",
-                vm_name, GetLastError());
-    }
-
-    g_free(vm_name);
-    return hDeviceVM;
-}
-
-int hax_notify_qemu_version(hax_fd vm_fd, struct hax_qemu_version *qversion)
-{
-    int ret;
-    DWORD dSize = 0;
-    if (hax_invalid_fd(vm_fd)) {
-        return -EINVAL;
-    }
-    ret = DeviceIoControl(vm_fd,
-                          HAX_VM_IOCTL_NOTIFY_QEMU_VERSION,
-                          qversion, sizeof(struct hax_qemu_version),
-                          NULL, 0, &dSize, (LPOVERLAPPED) NULL);
-    if (!ret) {
-        fprintf(stderr, "Failed to notify qemu API version\n");
-        return -1;
-    }
-    return 0;
-}
-
-int hax_host_create_vcpu(hax_fd vm_fd, int vcpuid)
-{
-    int ret;
-    DWORD dSize = 0;
-
-    ret = DeviceIoControl(vm_fd,
-                          HAX_VM_IOCTL_VCPU_CREATE,
-                          &vcpuid, sizeof(vcpuid), NULL, 0, &dSize,
-                          (LPOVERLAPPED) NULL);
-    if (!ret) {
-        fprintf(stderr, "Failed to create vcpu %x\n", vcpuid);
-        return -1;
-    }
-
-    return 0;
-}
-
-hax_fd hax_host_open_vcpu(int vmid, int vcpuid)
-{
-    char *devfs_path = NULL;
-    hax_fd hDeviceVCPU;
-
-    devfs_path = hax_vcpu_devfs_string(vmid, vcpuid);
-    if (!devfs_path) {
-        fprintf(stderr, "Failed to get the devfs\n");
-        return INVALID_HANDLE_VALUE;
-    }
-
-    hDeviceVCPU = CreateFile(devfs_path,
-                             GENERIC_READ | GENERIC_WRITE,
-                             0, NULL, CREATE_ALWAYS, FILE_ATTRIBUTE_NORMAL,
-                             NULL);
-
-    if (hDeviceVCPU == INVALID_HANDLE_VALUE) {
-        fprintf(stderr, "Failed to open the vcpu devfs\n");
-    }
-    g_free(devfs_path);
-    return hDeviceVCPU;
-}
-
-int hax_host_setup_vcpu_channel(struct hax_vcpu_state *vcpu)
-{
-    hax_fd hDeviceVCPU = vcpu->fd;
-    int ret;
-    struct hax_tunnel_info info;
-    DWORD dSize = 0;
-
-    ret = DeviceIoControl(hDeviceVCPU,
-                          HAX_VCPU_IOCTL_SETUP_TUNNEL,
-                          NULL, 0, &info, sizeof(info), &dSize,
-                          (LPOVERLAPPED) NULL);
-    if (!ret) {
-        fprintf(stderr, "Failed to setup the hax tunnel\n");
-        return -1;
-    }
-
-    if (!valid_hax_tunnel_size(info.size)) {
-        fprintf(stderr, "Invalid hax tunnel size %x\n", info.size);
-        ret = -EINVAL;
-        return ret;
-    }
-    vcpu->tunnel = (struct hax_tunnel *) (intptr_t) (info.va);
-    vcpu->iobuf = (unsigned char *) (intptr_t) (info.io_va);
-    return 0;
-}
-
-int hax_vcpu_run(struct hax_vcpu_state *vcpu)
-{
-    int ret;
-    HANDLE hDeviceVCPU = vcpu->fd;
-    DWORD dSize = 0;
-
-    ret = DeviceIoControl(hDeviceVCPU,
-                          HAX_VCPU_IOCTL_RUN,
-                          NULL, 0, NULL, 0, &dSize, (LPOVERLAPPED) NULL);
-    if (!ret) {
-        return -EFAULT;
-    } else {
-        return 0;
-    }
-}
-
-int hax_sync_fpu(CPUArchState *env, struct fx_layout *fl, int set)
-{
-    int ret;
-    hax_fd fd;
-    HANDLE hDeviceVCPU;
-    DWORD dSize = 0;
-
-    fd = hax_vcpu_get_fd(env);
-    if (hax_invalid_fd(fd)) {
-        return -1;
-    }
-
-    hDeviceVCPU = fd;
-
-    if (set) {
-        ret = DeviceIoControl(hDeviceVCPU,
-                              HAX_VCPU_IOCTL_SET_FPU,
-                              fl, sizeof(*fl), NULL, 0, &dSize,
-                              (LPOVERLAPPED) NULL);
-    } else {
-        ret = DeviceIoControl(hDeviceVCPU,
-                              HAX_VCPU_IOCTL_GET_FPU,
-                              NULL, 0, fl, sizeof(*fl), &dSize,
-                              (LPOVERLAPPED) NULL);
-    }
-    if (!ret) {
-        return -EFAULT;
-    } else {
-        return 0;
-    }
-}
-
-int hax_sync_msr(CPUArchState *env, struct hax_msr_data *msrs, int set)
-{
-    int ret;
-    hax_fd fd;
-    HANDLE hDeviceVCPU;
-    DWORD dSize = 0;
-
-    fd = hax_vcpu_get_fd(env);
-    if (hax_invalid_fd(fd)) {
-        return -1;
-    }
-    hDeviceVCPU = fd;
-
-    if (set) {
-        ret = DeviceIoControl(hDeviceVCPU,
-                              HAX_VCPU_IOCTL_SET_MSRS,
-                              msrs, sizeof(*msrs),
-                              msrs, sizeof(*msrs), &dSize, (LPOVERLAPPED) NULL);
-    } else {
-        ret = DeviceIoControl(hDeviceVCPU,
-                              HAX_VCPU_IOCTL_GET_MSRS,
-                              msrs, sizeof(*msrs),
-                              msrs, sizeof(*msrs), &dSize, (LPOVERLAPPED) NULL);
-    }
-    if (!ret) {
-        return -EFAULT;
-    } else {
-        return 0;
-    }
-}
-
-int hax_sync_vcpu_state(CPUArchState *env, struct vcpu_state_t *state, int set)
-{
-    int ret;
-    hax_fd fd;
-    HANDLE hDeviceVCPU;
-    DWORD dSize;
-
-    fd = hax_vcpu_get_fd(env);
-    if (hax_invalid_fd(fd)) {
-        return -1;
-    }
-
-    hDeviceVCPU = fd;
-
-    if (set) {
-        ret = DeviceIoControl(hDeviceVCPU,
-                              HAX_VCPU_SET_REGS,
-                              state, sizeof(*state),
-                              NULL, 0, &dSize, (LPOVERLAPPED) NULL);
-    } else {
-        ret = DeviceIoControl(hDeviceVCPU,
-                              HAX_VCPU_GET_REGS,
-                              NULL, 0,
-                              state, sizeof(*state), &dSize,
-                              (LPOVERLAPPED) NULL);
-    }
-    if (!ret) {
-        return -EFAULT;
-    } else {
-        return 0;
-    }
-}
-
-int hax_inject_interrupt(CPUArchState *env, int vector)
-{
-    int ret;
-    hax_fd fd;
-    HANDLE hDeviceVCPU;
-    DWORD dSize;
-
-    fd = hax_vcpu_get_fd(env);
-    if (hax_invalid_fd(fd)) {
-        return -1;
-    }
-
-    hDeviceVCPU = fd;
-
-    ret = DeviceIoControl(hDeviceVCPU,
-                          HAX_VCPU_IOCTL_INTERRUPT,
-                          &vector, sizeof(vector), NULL, 0, &dSize,
-                          (LPOVERLAPPED) NULL);
-    if (!ret) {
-        return -EFAULT;
-    } else {
-        return 0;
-    }
-}
-
-static void CALLBACK dummy_apc_func(ULONG_PTR unused)
-{
-}
-
-void hax_kick_vcpu_thread(CPUState *cpu)
-{
-    /*
-     * FIXME: race condition with the exit_request check in
-     * hax_vcpu_hax_exec
-     */
-    cpu->exit_request = 1;
-    if (!qemu_cpu_is_self(cpu)) {
-        if (!QueueUserAPC(dummy_apc_func, cpu->hThread, 0)) {
-            fprintf(stderr, "%s: QueueUserAPC failed with error %lu\n",
-                    __func__, GetLastError());
-            exit(1);
-        }
-    }
-}
diff --git a/target/i386/hax-windows.h b/target/i386/hax-windows.h
deleted file mode 100644 (file)
index a5ce12d..0000000
+++ /dev/null
@@ -1,88 +0,0 @@
-/*
- * QEMU HAXM support
- *
- * Copyright IBM, Corp. 2008
- *
- * Authors:
- *  Anthony Liguori   <aliguori@us.ibm.com>
- *
- * Copyright (c) 2011 Intel Corporation
- *  Written by:
- *  Jiang Yunhong<yunhong.jiang@intel.com>
- *  Xin Xiaohui<xiaohui.xin@intel.com>
- *  Zhang Xiantao<xiantao.zhang@intel.com>
- *
- * This work is licensed under the terms of the GNU GPL, version 2 or later.
- * See the COPYING file in the top-level directory.
- *
- */
-
-#ifndef TARGET_I386_HAX_WINDOWS_H
-#define TARGET_I386_HAX_WINDOWS_H
-
-#include <winioctl.h>
-#include <windef.h>
-
-#include "hax-cpus.h"
-
-#define HAX_INVALID_FD INVALID_HANDLE_VALUE
-
-static inline void hax_mod_close(struct hax_state *hax)
-{
-    CloseHandle(hax->fd);
-}
-
-static inline void hax_close_fd(hax_fd fd)
-{
-    CloseHandle(fd);
-}
-
-static inline int hax_invalid_fd(hax_fd fd)
-{
-    return (fd == INVALID_HANDLE_VALUE);
-}
-
-#define HAX_DEVICE_TYPE 0x4000
-
-#define HAX_IOCTL_VERSION       CTL_CODE(HAX_DEVICE_TYPE, 0x900, \
-                                         METHOD_BUFFERED, FILE_ANY_ACCESS)
-#define HAX_IOCTL_CREATE_VM     CTL_CODE(HAX_DEVICE_TYPE, 0x901, \
-                                         METHOD_BUFFERED, FILE_ANY_ACCESS)
-#define HAX_IOCTL_CAPABILITY    CTL_CODE(HAX_DEVICE_TYPE, 0x910, \
-                                         METHOD_BUFFERED, FILE_ANY_ACCESS)
-
-#define HAX_VM_IOCTL_VCPU_CREATE   CTL_CODE(HAX_DEVICE_TYPE, 0x902, \
-                                            METHOD_BUFFERED, FILE_ANY_ACCESS)
-#define HAX_VM_IOCTL_ALLOC_RAM     CTL_CODE(HAX_DEVICE_TYPE, 0x903, \
-                                            METHOD_BUFFERED, FILE_ANY_ACCESS)
-#define HAX_VM_IOCTL_SET_RAM       CTL_CODE(HAX_DEVICE_TYPE, 0x904, \
-                                            METHOD_BUFFERED, FILE_ANY_ACCESS)
-#define HAX_VM_IOCTL_VCPU_DESTROY  CTL_CODE(HAX_DEVICE_TYPE, 0x905, \
-                                            METHOD_BUFFERED, FILE_ANY_ACCESS)
-#define HAX_VM_IOCTL_ADD_RAMBLOCK  CTL_CODE(HAX_DEVICE_TYPE, 0x913, \
-                                            METHOD_BUFFERED, FILE_ANY_ACCESS)
-
-#define HAX_VCPU_IOCTL_RUN      CTL_CODE(HAX_DEVICE_TYPE, 0x906, \
-                                         METHOD_BUFFERED, FILE_ANY_ACCESS)
-#define HAX_VCPU_IOCTL_SET_MSRS CTL_CODE(HAX_DEVICE_TYPE, 0x907, \
-                                         METHOD_BUFFERED, FILE_ANY_ACCESS)
-#define HAX_VCPU_IOCTL_GET_MSRS CTL_CODE(HAX_DEVICE_TYPE, 0x908, \
-                                         METHOD_BUFFERED, FILE_ANY_ACCESS)
-#define HAX_VCPU_IOCTL_SET_FPU  CTL_CODE(HAX_DEVICE_TYPE, 0x909, \
-                                         METHOD_BUFFERED, FILE_ANY_ACCESS)
-#define HAX_VCPU_IOCTL_GET_FPU  CTL_CODE(HAX_DEVICE_TYPE, 0x90a, \
-                                         METHOD_BUFFERED, FILE_ANY_ACCESS)
-
-#define HAX_VCPU_IOCTL_SETUP_TUNNEL  CTL_CODE(HAX_DEVICE_TYPE, 0x90b, \
-                                              METHOD_BUFFERED, FILE_ANY_ACCESS)
-#define HAX_VCPU_IOCTL_INTERRUPT     CTL_CODE(HAX_DEVICE_TYPE, 0x90c, \
-                                              METHOD_BUFFERED, FILE_ANY_ACCESS)
-#define HAX_VCPU_SET_REGS            CTL_CODE(HAX_DEVICE_TYPE, 0x90d, \
-                                              METHOD_BUFFERED, FILE_ANY_ACCESS)
-#define HAX_VCPU_GET_REGS            CTL_CODE(HAX_DEVICE_TYPE, 0x90e, \
-                                              METHOD_BUFFERED, FILE_ANY_ACCESS)
-
-#define HAX_VM_IOCTL_NOTIFY_QEMU_VERSION CTL_CODE(HAX_DEVICE_TYPE, 0x910, \
-                                                  METHOD_BUFFERED,        \
-                                                  FILE_ANY_ACCESS)
-#endif /* TARGET_I386_HAX_WINDOWS_H */
diff --git a/target/i386/hax/hax-all.c b/target/i386/hax/hax-all.c
new file mode 100644 (file)
index 0000000..fecfe8c
--- /dev/null
@@ -0,0 +1,1143 @@
+/*
+ * QEMU HAX support
+ *
+ * Copyright IBM, Corp. 2008
+ *           Red Hat, Inc. 2008
+ *
+ * Authors:
+ *  Anthony Liguori   <aliguori@us.ibm.com>
+ *  Glauber Costa     <gcosta@redhat.com>
+ *
+ * Copyright (c) 2011 Intel Corporation
+ *  Written by:
+ *  Jiang Yunhong<yunhong.jiang@intel.com>
+ *  Xin Xiaohui<xiaohui.xin@intel.com>
+ *  Zhang Xiantao<xiantao.zhang@intel.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+/*
+ * HAX common code for both windows and darwin
+ */
+
+#include "qemu/osdep.h"
+#include "cpu.h"
+#include "exec/address-spaces.h"
+
+#include "qemu-common.h"
+#include "sysemu/accel.h"
+#include "sysemu/reset.h"
+#include "sysemu/runstate.h"
+#include "hw/boards.h"
+
+#include "hax-cpus.h"
+
+#define DEBUG_HAX 0
+
+#define DPRINTF(fmt, ...) \
+    do { \
+        if (DEBUG_HAX) { \
+            fprintf(stdout, fmt, ## __VA_ARGS__); \
+        } \
+    } while (0)
+
+/* Current version */
+const uint32_t hax_cur_version = 0x4; /* API v4: unmapping and MMIO moves */
+/* Minimum HAX kernel version */
+const uint32_t hax_min_version = 0x4; /* API v4: supports unmapping */
+
+static bool hax_allowed;
+
+struct hax_state hax_global;
+
+static void hax_vcpu_sync_state(CPUArchState *env, int modified);
+static int hax_arch_get_registers(CPUArchState *env);
+
+int hax_enabled(void)
+{
+    return hax_allowed;
+}
+
+int valid_hax_tunnel_size(uint16_t size)
+{
+    return size >= sizeof(struct hax_tunnel);
+}
+
+hax_fd hax_vcpu_get_fd(CPUArchState *env)
+{
+    struct hax_vcpu_state *vcpu = env_cpu(env)->hax_vcpu;
+    if (!vcpu) {
+        return HAX_INVALID_FD;
+    }
+    return vcpu->fd;
+}
+
+static int hax_get_capability(struct hax_state *hax)
+{
+    int ret;
+    struct hax_capabilityinfo capinfo, *cap = &capinfo;
+
+    ret = hax_capability(hax, cap);
+    if (ret) {
+        return ret;
+    }
+
+    if ((cap->wstatus & HAX_CAP_WORKSTATUS_MASK) == HAX_CAP_STATUS_NOTWORKING) {
+        if (cap->winfo & HAX_CAP_FAILREASON_VT) {
+            DPRINTF
+                ("VTX feature is not enabled, HAX driver will not work.\n");
+        } else if (cap->winfo & HAX_CAP_FAILREASON_NX) {
+            DPRINTF
+                ("NX feature is not enabled, HAX driver will not work.\n");
+        }
+        return -ENXIO;
+
+    }
+
+    if (!(cap->winfo & HAX_CAP_UG)) {
+        fprintf(stderr, "UG mode is not supported by the hardware.\n");
+        return -ENOTSUP;
+    }
+
+    hax->supports_64bit_ramblock = !!(cap->winfo & HAX_CAP_64BIT_RAMBLOCK);
+
+    if (cap->wstatus & HAX_CAP_MEMQUOTA) {
+        if (cap->mem_quota < hax->mem_quota) {
+            fprintf(stderr, "The VM memory needed exceeds the driver limit.\n");
+            return -ENOSPC;
+        }
+    }
+    return 0;
+}
+
+static int hax_version_support(struct hax_state *hax)
+{
+    int ret;
+    struct hax_module_version version;
+
+    ret = hax_mod_version(hax, &version);
+    if (ret < 0) {
+        return 0;
+    }
+
+    if (hax_min_version > version.cur_version) {
+        fprintf(stderr, "Incompatible HAX module version %d,",
+                version.cur_version);
+        fprintf(stderr, "requires minimum version %d\n", hax_min_version);
+        return 0;
+    }
+    if (hax_cur_version < version.compat_version) {
+        fprintf(stderr, "Incompatible QEMU HAX API version %x,",
+                hax_cur_version);
+        fprintf(stderr, "requires minimum HAX API version %x\n",
+                version.compat_version);
+        return 0;
+    }
+
+    return 1;
+}
+
+int hax_vcpu_create(int id)
+{
+    struct hax_vcpu_state *vcpu = NULL;
+    int ret;
+
+    if (!hax_global.vm) {
+        fprintf(stderr, "vcpu %x created failed, vm is null\n", id);
+        return -1;
+    }
+
+    if (hax_global.vm->vcpus[id]) {
+        fprintf(stderr, "vcpu %x allocated already\n", id);
+        return 0;
+    }
+
+    vcpu = g_new0(struct hax_vcpu_state, 1);
+
+    ret = hax_host_create_vcpu(hax_global.vm->fd, id);
+    if (ret) {
+        fprintf(stderr, "Failed to create vcpu %x\n", id);
+        goto error;
+    }
+
+    vcpu->vcpu_id = id;
+    vcpu->fd = hax_host_open_vcpu(hax_global.vm->id, id);
+    if (hax_invalid_fd(vcpu->fd)) {
+        fprintf(stderr, "Failed to open the vcpu\n");
+        ret = -ENODEV;
+        goto error;
+    }
+
+    hax_global.vm->vcpus[id] = vcpu;
+
+    ret = hax_host_setup_vcpu_channel(vcpu);
+    if (ret) {
+        fprintf(stderr, "Invalid hax tunnel size\n");
+        ret = -EINVAL;
+        goto error;
+    }
+    return 0;
+
+  error:
+    /* vcpu and tunnel will be closed automatically */
+    if (vcpu && !hax_invalid_fd(vcpu->fd)) {
+        hax_close_fd(vcpu->fd);
+    }
+
+    hax_global.vm->vcpus[id] = NULL;
+    g_free(vcpu);
+    return -1;
+}
+
+int hax_vcpu_destroy(CPUState *cpu)
+{
+    struct hax_vcpu_state *vcpu = cpu->hax_vcpu;
+
+    if (!hax_global.vm) {
+        fprintf(stderr, "vcpu %x destroy failed, vm is null\n", vcpu->vcpu_id);
+        return -1;
+    }
+
+    if (!vcpu) {
+        return 0;
+    }
+
+    /*
+     * 1. The hax_tunnel is also destroyed when vcpu is destroyed
+     * 2. close fd will cause hax module vcpu be cleaned
+     */
+    hax_close_fd(vcpu->fd);
+    hax_global.vm->vcpus[vcpu->vcpu_id] = NULL;
+    g_free(vcpu);
+    return 0;
+}
+
+int hax_init_vcpu(CPUState *cpu)
+{
+    int ret;
+
+    ret = hax_vcpu_create(cpu->cpu_index);
+    if (ret < 0) {
+        fprintf(stderr, "Failed to create HAX vcpu\n");
+        exit(-1);
+    }
+
+    cpu->hax_vcpu = hax_global.vm->vcpus[cpu->cpu_index];
+    cpu->vcpu_dirty = true;
+    qemu_register_reset(hax_reset_vcpu_state, (CPUArchState *) (cpu->env_ptr));
+
+    return ret;
+}
+
+struct hax_vm *hax_vm_create(struct hax_state *hax, int max_cpus)
+{
+    struct hax_vm *vm;
+    int vm_id = 0, ret, i;
+
+    if (hax_invalid_fd(hax->fd)) {
+        return NULL;
+    }
+
+    if (hax->vm) {
+        return hax->vm;
+    }
+
+    if (max_cpus > HAX_MAX_VCPU) {
+        fprintf(stderr, "Maximum VCPU number QEMU supported is %d\n", HAX_MAX_VCPU);
+        return NULL;
+    }
+
+    vm = g_new0(struct hax_vm, 1);
+
+    ret = hax_host_create_vm(hax, &vm_id);
+    if (ret) {
+        fprintf(stderr, "Failed to create vm %x\n", ret);
+        goto error;
+    }
+    vm->id = vm_id;
+    vm->fd = hax_host_open_vm(hax, vm_id);
+    if (hax_invalid_fd(vm->fd)) {
+        fprintf(stderr, "Failed to open vm %d\n", vm_id);
+        goto error;
+    }
+
+    vm->numvcpus = max_cpus;
+    vm->vcpus = g_new0(struct hax_vcpu_state *, vm->numvcpus);
+    for (i = 0; i < vm->numvcpus; i++) {
+        vm->vcpus[i] = NULL;
+    }
+
+    hax->vm = vm;
+    return vm;
+
+  error:
+    g_free(vm);
+    hax->vm = NULL;
+    return NULL;
+}
+
+int hax_vm_destroy(struct hax_vm *vm)
+{
+    int i;
+
+    for (i = 0; i < vm->numvcpus; i++)
+        if (vm->vcpus[i]) {
+            fprintf(stderr, "VCPU should be cleaned before vm clean\n");
+            return -1;
+        }
+    hax_close_fd(vm->fd);
+    vm->numvcpus = 0;
+    g_free(vm->vcpus);
+    g_free(vm);
+    hax_global.vm = NULL;
+    return 0;
+}
+
+static int hax_init(ram_addr_t ram_size, int max_cpus)
+{
+    struct hax_state *hax = NULL;
+    struct hax_qemu_version qversion;
+    int ret;
+
+    hax = &hax_global;
+
+    memset(hax, 0, sizeof(struct hax_state));
+    hax->mem_quota = ram_size;
+
+    hax->fd = hax_mod_open();
+    if (hax_invalid_fd(hax->fd)) {
+        hax->fd = 0;
+        ret = -ENODEV;
+        goto error;
+    }
+
+    ret = hax_get_capability(hax);
+
+    if (ret) {
+        if (ret != -ENOSPC) {
+            ret = -EINVAL;
+        }
+        goto error;
+    }
+
+    if (!hax_version_support(hax)) {
+        ret = -EINVAL;
+        goto error;
+    }
+
+    hax->vm = hax_vm_create(hax, max_cpus);
+    if (!hax->vm) {
+        fprintf(stderr, "Failed to create HAX VM\n");
+        ret = -EINVAL;
+        goto error;
+    }
+
+    hax_memory_init();
+
+    qversion.cur_version = hax_cur_version;
+    qversion.min_version = hax_min_version;
+    hax_notify_qemu_version(hax->vm->fd, &qversion);
+
+    return ret;
+  error:
+    if (hax->vm) {
+        hax_vm_destroy(hax->vm);
+    }
+    if (hax->fd) {
+        hax_mod_close(hax);
+    }
+
+    return ret;
+}
+
+static int hax_accel_init(MachineState *ms)
+{
+    int ret = hax_init(ms->ram_size, (int)ms->smp.max_cpus);
+
+    if (ret && (ret != -ENOSPC)) {
+        fprintf(stderr, "No accelerator found.\n");
+    } else {
+        fprintf(stdout, "HAX is %s and emulator runs in %s mode.\n",
+                !ret ? "working" : "not working",
+                !ret ? "fast virt" : "emulation");
+    }
+    if (ret == 0) {
+        cpus_register_accel(&hax_cpus);
+    }
+    return ret;
+}
+
+static int hax_handle_fastmmio(CPUArchState *env, struct hax_fastmmio *hft)
+{
+    if (hft->direction < 2) {
+        cpu_physical_memory_rw(hft->gpa, &hft->value, hft->size,
+                               hft->direction);
+    } else {
+        /*
+         * HAX API v4 supports transferring data between two MMIO addresses,
+         * hft->gpa and hft->gpa2 (instructions such as MOVS require this):
+         *  hft->direction == 2: gpa ==> gpa2
+         */
+        uint64_t value;
+        cpu_physical_memory_read(hft->gpa, &value, hft->size);
+        cpu_physical_memory_write(hft->gpa2, &value, hft->size);
+    }
+
+    return 0;
+}
+
+static int hax_handle_io(CPUArchState *env, uint32_t df, uint16_t port,
+                         int direction, int size, int count, void *buffer)
+{
+    uint8_t *ptr;
+    int i;
+    MemTxAttrs attrs = { 0 };
+
+    if (!df) {
+        ptr = (uint8_t *) buffer;
+    } else {
+        ptr = buffer + size * count - size;
+    }
+    for (i = 0; i < count; i++) {
+        address_space_rw(&address_space_io, port, attrs,
+                         ptr, size, direction == HAX_EXIT_IO_OUT);
+        if (!df) {
+            ptr += size;
+        } else {
+            ptr -= size;
+        }
+    }
+
+    return 0;
+}
+
+static int hax_vcpu_interrupt(CPUArchState *env)
+{
+    CPUState *cpu = env_cpu(env);
+    struct hax_vcpu_state *vcpu = cpu->hax_vcpu;
+    struct hax_tunnel *ht = vcpu->tunnel;
+
+    /*
+     * Try to inject an interrupt if the guest can accept it
+     * Unlike KVM, HAX kernel check for the eflags, instead of qemu
+     */
+    if (ht->ready_for_interrupt_injection &&
+        (cpu->interrupt_request & CPU_INTERRUPT_HARD)) {
+        int irq;
+
+        irq = cpu_get_pic_interrupt(env);
+        if (irq >= 0) {
+            hax_inject_interrupt(env, irq);
+            cpu->interrupt_request &= ~CPU_INTERRUPT_HARD;
+        }
+    }
+
+    /* If we have an interrupt but the guest is not ready to receive an
+     * interrupt, request an interrupt window exit.  This will
+     * cause a return to userspace as soon as the guest is ready to
+     * receive interrupts. */
+    if ((cpu->interrupt_request & CPU_INTERRUPT_HARD)) {
+        ht->request_interrupt_window = 1;
+    } else {
+        ht->request_interrupt_window = 0;
+    }
+    return 0;
+}
+
+void hax_raise_event(CPUState *cpu)
+{
+    struct hax_vcpu_state *vcpu = cpu->hax_vcpu;
+
+    if (!vcpu) {
+        return;
+    }
+    vcpu->tunnel->user_event_pending = 1;
+}
+
+/*
+ * Ask hax kernel module to run the CPU for us till:
+ * 1. Guest crash or shutdown
+ * 2. Need QEMU's emulation like guest execute MMIO instruction
+ * 3. Guest execute HLT
+ * 4. QEMU have Signal/event pending
+ * 5. An unknown VMX exit happens
+ */
+static int hax_vcpu_hax_exec(CPUArchState *env)
+{
+    int ret = 0;
+    CPUState *cpu = env_cpu(env);
+    X86CPU *x86_cpu = X86_CPU(cpu);
+    struct hax_vcpu_state *vcpu = cpu->hax_vcpu;
+    struct hax_tunnel *ht = vcpu->tunnel;
+
+    if (!hax_enabled()) {
+        DPRINTF("Trying to vcpu execute at eip:" TARGET_FMT_lx "\n", env->eip);
+        return 0;
+    }
+
+    if (cpu->interrupt_request & CPU_INTERRUPT_POLL) {
+        cpu->interrupt_request &= ~CPU_INTERRUPT_POLL;
+        apic_poll_irq(x86_cpu->apic_state);
+    }
+
+    /* After a vcpu is halted (either because it is an AP and has just been
+     * reset, or because it has executed the HLT instruction), it will not be
+     * run (hax_vcpu_run()) until it is unhalted. The next few if blocks check
+     * for events that may change the halted state of this vcpu:
+     *  a) Maskable interrupt, when RFLAGS.IF is 1;
+     *     Note: env->eflags may not reflect the current RFLAGS state, because
+     *           it is not updated after each hax_vcpu_run(). We cannot afford
+     *           to fail to recognize any unhalt-by-maskable-interrupt event
+     *           (in which case the vcpu will halt forever), and yet we cannot
+     *           afford the overhead of hax_vcpu_sync_state(). The current
+     *           solution is to err on the side of caution and have the HLT
+     *           handler (see case HAX_EXIT_HLT below) unconditionally set the
+     *           IF_MASK bit in env->eflags, which, in effect, disables the
+     *           RFLAGS.IF check.
+     *  b) NMI;
+     *  c) INIT signal;
+     *  d) SIPI signal.
+     */
+    if (((cpu->interrupt_request & CPU_INTERRUPT_HARD) &&
+         (env->eflags & IF_MASK)) ||
+        (cpu->interrupt_request & CPU_INTERRUPT_NMI)) {
+        cpu->halted = 0;
+    }
+
+    if (cpu->interrupt_request & CPU_INTERRUPT_INIT) {
+        DPRINTF("\nhax_vcpu_hax_exec: handling INIT for %d\n",
+                cpu->cpu_index);
+        do_cpu_init(x86_cpu);
+        hax_vcpu_sync_state(env, 1);
+    }
+
+    if (cpu->interrupt_request & CPU_INTERRUPT_SIPI) {
+        DPRINTF("hax_vcpu_hax_exec: handling SIPI for %d\n",
+                cpu->cpu_index);
+        hax_vcpu_sync_state(env, 0);
+        do_cpu_sipi(x86_cpu);
+        hax_vcpu_sync_state(env, 1);
+    }
+
+    if (cpu->halted) {
+        /* If this vcpu is halted, we must not ask HAXM to run it. Instead, we
+         * break out of hax_smp_cpu_exec() as if this vcpu had executed HLT.
+         * That way, this vcpu thread will be trapped in qemu_wait_io_event(),
+         * until the vcpu is unhalted.
+         */
+        cpu->exception_index = EXCP_HLT;
+        return 0;
+    }
+
+    do {
+        int hax_ret;
+
+        if (cpu->exit_request) {
+            ret = 1;
+            break;
+        }
+
+        hax_vcpu_interrupt(env);
+
+        qemu_mutex_unlock_iothread();
+        cpu_exec_start(cpu);
+        hax_ret = hax_vcpu_run(vcpu);
+        cpu_exec_end(cpu);
+        qemu_mutex_lock_iothread();
+
+        /* Simply continue the vcpu_run if system call interrupted */
+        if (hax_ret == -EINTR || hax_ret == -EAGAIN) {
+            DPRINTF("io window interrupted\n");
+            continue;
+        }
+
+        if (hax_ret < 0) {
+            fprintf(stderr, "vcpu run failed for vcpu  %x\n", vcpu->vcpu_id);
+            abort();
+        }
+        switch (ht->_exit_status) {
+        case HAX_EXIT_IO:
+            ret = hax_handle_io(env, ht->pio._df, ht->pio._port,
+                            ht->pio._direction,
+                            ht->pio._size, ht->pio._count, vcpu->iobuf);
+            break;
+        case HAX_EXIT_FAST_MMIO:
+            ret = hax_handle_fastmmio(env, (struct hax_fastmmio *) vcpu->iobuf);
+            break;
+        /* Guest state changed, currently only for shutdown */
+        case HAX_EXIT_STATECHANGE:
+            fprintf(stdout, "VCPU shutdown request\n");
+            qemu_system_shutdown_request(SHUTDOWN_CAUSE_GUEST_SHUTDOWN);
+            hax_vcpu_sync_state(env, 0);
+            ret = 1;
+            break;
+        case HAX_EXIT_UNKNOWN_VMEXIT:
+            fprintf(stderr, "Unknown VMX exit %x from guest\n",
+                    ht->_exit_reason);
+            qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
+            hax_vcpu_sync_state(env, 0);
+            cpu_dump_state(cpu, stderr, 0);
+            ret = -1;
+            break;
+        case HAX_EXIT_HLT:
+            if (!(cpu->interrupt_request & CPU_INTERRUPT_HARD) &&
+                !(cpu->interrupt_request & CPU_INTERRUPT_NMI)) {
+                /* hlt instruction with interrupt disabled is shutdown */
+                env->eflags |= IF_MASK;
+                cpu->halted = 1;
+                cpu->exception_index = EXCP_HLT;
+                ret = 1;
+            }
+            break;
+        /* these situations will continue to hax module */
+        case HAX_EXIT_INTERRUPT:
+        case HAX_EXIT_PAUSED:
+            break;
+        case HAX_EXIT_MMIO:
+            /* Should not happen on UG system */
+            fprintf(stderr, "HAX: unsupported MMIO emulation\n");
+            ret = -1;
+            break;
+        case HAX_EXIT_REAL:
+            /* Should not happen on UG system */
+            fprintf(stderr, "HAX: unimplemented real mode emulation\n");
+            ret = -1;
+            break;
+        default:
+            fprintf(stderr, "Unknown exit %x from HAX\n", ht->_exit_status);
+            qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
+            hax_vcpu_sync_state(env, 0);
+            cpu_dump_state(cpu, stderr, 0);
+            ret = 1;
+            break;
+        }
+    } while (!ret);
+
+    if (cpu->exit_request) {
+        cpu->exit_request = 0;
+        cpu->exception_index = EXCP_INTERRUPT;
+    }
+    return ret < 0;
+}
+
+static void do_hax_cpu_synchronize_state(CPUState *cpu, run_on_cpu_data arg)
+{
+    CPUArchState *env = cpu->env_ptr;
+
+    hax_arch_get_registers(env);
+    cpu->vcpu_dirty = true;
+}
+
+void hax_cpu_synchronize_state(CPUState *cpu)
+{
+    if (!cpu->vcpu_dirty) {
+        run_on_cpu(cpu, do_hax_cpu_synchronize_state, RUN_ON_CPU_NULL);
+    }
+}
+
+static void do_hax_cpu_synchronize_post_reset(CPUState *cpu,
+                                              run_on_cpu_data arg)
+{
+    CPUArchState *env = cpu->env_ptr;
+
+    hax_vcpu_sync_state(env, 1);
+    cpu->vcpu_dirty = false;
+}
+
+void hax_cpu_synchronize_post_reset(CPUState *cpu)
+{
+    run_on_cpu(cpu, do_hax_cpu_synchronize_post_reset, RUN_ON_CPU_NULL);
+}
+
+static void do_hax_cpu_synchronize_post_init(CPUState *cpu, run_on_cpu_data arg)
+{
+    CPUArchState *env = cpu->env_ptr;
+
+    hax_vcpu_sync_state(env, 1);
+    cpu->vcpu_dirty = false;
+}
+
+void hax_cpu_synchronize_post_init(CPUState *cpu)
+{
+    run_on_cpu(cpu, do_hax_cpu_synchronize_post_init, RUN_ON_CPU_NULL);
+}
+
+static void do_hax_cpu_synchronize_pre_loadvm(CPUState *cpu, run_on_cpu_data arg)
+{
+    cpu->vcpu_dirty = true;
+}
+
+void hax_cpu_synchronize_pre_loadvm(CPUState *cpu)
+{
+    run_on_cpu(cpu, do_hax_cpu_synchronize_pre_loadvm, RUN_ON_CPU_NULL);
+}
+
+int hax_smp_cpu_exec(CPUState *cpu)
+{
+    CPUArchState *env = (CPUArchState *) (cpu->env_ptr);
+    int fatal;
+    int ret;
+
+    while (1) {
+        if (cpu->exception_index >= EXCP_INTERRUPT) {
+            ret = cpu->exception_index;
+            cpu->exception_index = -1;
+            break;
+        }
+
+        fatal = hax_vcpu_hax_exec(env);
+
+        if (fatal) {
+            fprintf(stderr, "Unsupported HAX vcpu return\n");
+            abort();
+        }
+    }
+
+    return ret;
+}
+
+static void set_v8086_seg(struct segment_desc_t *lhs, const SegmentCache *rhs)
+{
+    memset(lhs, 0, sizeof(struct segment_desc_t));
+    lhs->selector = rhs->selector;
+    lhs->base = rhs->base;
+    lhs->limit = rhs->limit;
+    lhs->type = 3;
+    lhs->present = 1;
+    lhs->dpl = 3;
+    lhs->operand_size = 0;
+    lhs->desc = 1;
+    lhs->long_mode = 0;
+    lhs->granularity = 0;
+    lhs->available = 0;
+}
+
+static void get_seg(SegmentCache *lhs, const struct segment_desc_t *rhs)
+{
+    lhs->selector = rhs->selector;
+    lhs->base = rhs->base;
+    lhs->limit = rhs->limit;
+    lhs->flags = (rhs->type << DESC_TYPE_SHIFT)
+        | (rhs->present * DESC_P_MASK)
+        | (rhs->dpl << DESC_DPL_SHIFT)
+        | (rhs->operand_size << DESC_B_SHIFT)
+        | (rhs->desc * DESC_S_MASK)
+        | (rhs->long_mode << DESC_L_SHIFT)
+        | (rhs->granularity * DESC_G_MASK) | (rhs->available * DESC_AVL_MASK);
+}
+
+static void set_seg(struct segment_desc_t *lhs, const SegmentCache *rhs)
+{
+    unsigned flags = rhs->flags;
+
+    memset(lhs, 0, sizeof(struct segment_desc_t));
+    lhs->selector = rhs->selector;
+    lhs->base = rhs->base;
+    lhs->limit = rhs->limit;
+    lhs->type = (flags >> DESC_TYPE_SHIFT) & 15;
+    lhs->present = (flags & DESC_P_MASK) != 0;
+    lhs->dpl = rhs->selector & 3;
+    lhs->operand_size = (flags >> DESC_B_SHIFT) & 1;
+    lhs->desc = (flags & DESC_S_MASK) != 0;
+    lhs->long_mode = (flags >> DESC_L_SHIFT) & 1;
+    lhs->granularity = (flags & DESC_G_MASK) != 0;
+    lhs->available = (flags & DESC_AVL_MASK) != 0;
+}
+
+static void hax_getput_reg(uint64_t *hax_reg, target_ulong *qemu_reg, int set)
+{
+    target_ulong reg = *hax_reg;
+
+    if (set) {
+        *hax_reg = *qemu_reg;
+    } else {
+        *qemu_reg = reg;
+    }
+}
+
+/* The sregs has been synced with HAX kernel already before this call */
+static int hax_get_segments(CPUArchState *env, struct vcpu_state_t *sregs)
+{
+    get_seg(&env->segs[R_CS], &sregs->_cs);
+    get_seg(&env->segs[R_DS], &sregs->_ds);
+    get_seg(&env->segs[R_ES], &sregs->_es);
+    get_seg(&env->segs[R_FS], &sregs->_fs);
+    get_seg(&env->segs[R_GS], &sregs->_gs);
+    get_seg(&env->segs[R_SS], &sregs->_ss);
+
+    get_seg(&env->tr, &sregs->_tr);
+    get_seg(&env->ldt, &sregs->_ldt);
+    env->idt.limit = sregs->_idt.limit;
+    env->idt.base = sregs->_idt.base;
+    env->gdt.limit = sregs->_gdt.limit;
+    env->gdt.base = sregs->_gdt.base;
+    return 0;
+}
+
+static int hax_set_segments(CPUArchState *env, struct vcpu_state_t *sregs)
+{
+    if ((env->eflags & VM_MASK)) {
+        set_v8086_seg(&sregs->_cs, &env->segs[R_CS]);
+        set_v8086_seg(&sregs->_ds, &env->segs[R_DS]);
+        set_v8086_seg(&sregs->_es, &env->segs[R_ES]);
+        set_v8086_seg(&sregs->_fs, &env->segs[R_FS]);
+        set_v8086_seg(&sregs->_gs, &env->segs[R_GS]);
+        set_v8086_seg(&sregs->_ss, &env->segs[R_SS]);
+    } else {
+        set_seg(&sregs->_cs, &env->segs[R_CS]);
+        set_seg(&sregs->_ds, &env->segs[R_DS]);
+        set_seg(&sregs->_es, &env->segs[R_ES]);
+        set_seg(&sregs->_fs, &env->segs[R_FS]);
+        set_seg(&sregs->_gs, &env->segs[R_GS]);
+        set_seg(&sregs->_ss, &env->segs[R_SS]);
+
+        if (env->cr[0] & CR0_PE_MASK) {
+            /* force ss cpl to cs cpl */
+            sregs->_ss.selector = (sregs->_ss.selector & ~3) |
+                                  (sregs->_cs.selector & 3);
+            sregs->_ss.dpl = sregs->_ss.selector & 3;
+        }
+    }
+
+    set_seg(&sregs->_tr, &env->tr);
+    set_seg(&sregs->_ldt, &env->ldt);
+    sregs->_idt.limit = env->idt.limit;
+    sregs->_idt.base = env->idt.base;
+    sregs->_gdt.limit = env->gdt.limit;
+    sregs->_gdt.base = env->gdt.base;
+    return 0;
+}
+
+static int hax_sync_vcpu_register(CPUArchState *env, int set)
+{
+    struct vcpu_state_t regs;
+    int ret;
+    memset(&regs, 0, sizeof(struct vcpu_state_t));
+
+    if (!set) {
+        ret = hax_sync_vcpu_state(env, &regs, 0);
+        if (ret < 0) {
+            return -1;
+        }
+    }
+
+    /* generic register */
+    hax_getput_reg(&regs._rax, &env->regs[R_EAX], set);
+    hax_getput_reg(&regs._rbx, &env->regs[R_EBX], set);
+    hax_getput_reg(&regs._rcx, &env->regs[R_ECX], set);
+    hax_getput_reg(&regs._rdx, &env->regs[R_EDX], set);
+    hax_getput_reg(&regs._rsi, &env->regs[R_ESI], set);
+    hax_getput_reg(&regs._rdi, &env->regs[R_EDI], set);
+    hax_getput_reg(&regs._rsp, &env->regs[R_ESP], set);
+    hax_getput_reg(&regs._rbp, &env->regs[R_EBP], set);
+#ifdef TARGET_X86_64
+    hax_getput_reg(&regs._r8, &env->regs[8], set);
+    hax_getput_reg(&regs._r9, &env->regs[9], set);
+    hax_getput_reg(&regs._r10, &env->regs[10], set);
+    hax_getput_reg(&regs._r11, &env->regs[11], set);
+    hax_getput_reg(&regs._r12, &env->regs[12], set);
+    hax_getput_reg(&regs._r13, &env->regs[13], set);
+    hax_getput_reg(&regs._r14, &env->regs[14], set);
+    hax_getput_reg(&regs._r15, &env->regs[15], set);
+#endif
+    hax_getput_reg(&regs._rflags, &env->eflags, set);
+    hax_getput_reg(&regs._rip, &env->eip, set);
+
+    if (set) {
+        regs._cr0 = env->cr[0];
+        regs._cr2 = env->cr[2];
+        regs._cr3 = env->cr[3];
+        regs._cr4 = env->cr[4];
+        hax_set_segments(env, &regs);
+    } else {
+        env->cr[0] = regs._cr0;
+        env->cr[2] = regs._cr2;
+        env->cr[3] = regs._cr3;
+        env->cr[4] = regs._cr4;
+        hax_get_segments(env, &regs);
+    }
+
+    if (set) {
+        ret = hax_sync_vcpu_state(env, &regs, 1);
+        if (ret < 0) {
+            return -1;
+        }
+    }
+    return 0;
+}
+
+static void hax_msr_entry_set(struct vmx_msr *item, uint32_t index,
+                              uint64_t value)
+{
+    item->entry = index;
+    item->value = value;
+}
+
+static int hax_get_msrs(CPUArchState *env)
+{
+    struct hax_msr_data md;
+    struct vmx_msr *msrs = md.entries;
+    int ret, i, n;
+
+    n = 0;
+    msrs[n++].entry = MSR_IA32_SYSENTER_CS;
+    msrs[n++].entry = MSR_IA32_SYSENTER_ESP;
+    msrs[n++].entry = MSR_IA32_SYSENTER_EIP;
+    msrs[n++].entry = MSR_IA32_TSC;
+#ifdef TARGET_X86_64
+    msrs[n++].entry = MSR_EFER;
+    msrs[n++].entry = MSR_STAR;
+    msrs[n++].entry = MSR_LSTAR;
+    msrs[n++].entry = MSR_CSTAR;
+    msrs[n++].entry = MSR_FMASK;
+    msrs[n++].entry = MSR_KERNELGSBASE;
+#endif
+    md.nr_msr = n;
+    ret = hax_sync_msr(env, &md, 0);
+    if (ret < 0) {
+        return ret;
+    }
+
+    for (i = 0; i < md.done; i++) {
+        switch (msrs[i].entry) {
+        case MSR_IA32_SYSENTER_CS:
+            env->sysenter_cs = msrs[i].value;
+            break;
+        case MSR_IA32_SYSENTER_ESP:
+            env->sysenter_esp = msrs[i].value;
+            break;
+        case MSR_IA32_SYSENTER_EIP:
+            env->sysenter_eip = msrs[i].value;
+            break;
+        case MSR_IA32_TSC:
+            env->tsc = msrs[i].value;
+            break;
+#ifdef TARGET_X86_64
+        case MSR_EFER:
+            env->efer = msrs[i].value;
+            break;
+        case MSR_STAR:
+            env->star = msrs[i].value;
+            break;
+        case MSR_LSTAR:
+            env->lstar = msrs[i].value;
+            break;
+        case MSR_CSTAR:
+            env->cstar = msrs[i].value;
+            break;
+        case MSR_FMASK:
+            env->fmask = msrs[i].value;
+            break;
+        case MSR_KERNELGSBASE:
+            env->kernelgsbase = msrs[i].value;
+            break;
+#endif
+        }
+    }
+
+    return 0;
+}
+
+static int hax_set_msrs(CPUArchState *env)
+{
+    struct hax_msr_data md;
+    struct vmx_msr *msrs;
+    msrs = md.entries;
+    int n = 0;
+
+    memset(&md, 0, sizeof(struct hax_msr_data));
+    hax_msr_entry_set(&msrs[n++], MSR_IA32_SYSENTER_CS, env->sysenter_cs);
+    hax_msr_entry_set(&msrs[n++], MSR_IA32_SYSENTER_ESP, env->sysenter_esp);
+    hax_msr_entry_set(&msrs[n++], MSR_IA32_SYSENTER_EIP, env->sysenter_eip);
+    hax_msr_entry_set(&msrs[n++], MSR_IA32_TSC, env->tsc);
+#ifdef TARGET_X86_64
+    hax_msr_entry_set(&msrs[n++], MSR_EFER, env->efer);
+    hax_msr_entry_set(&msrs[n++], MSR_STAR, env->star);
+    hax_msr_entry_set(&msrs[n++], MSR_LSTAR, env->lstar);
+    hax_msr_entry_set(&msrs[n++], MSR_CSTAR, env->cstar);
+    hax_msr_entry_set(&msrs[n++], MSR_FMASK, env->fmask);
+    hax_msr_entry_set(&msrs[n++], MSR_KERNELGSBASE, env->kernelgsbase);
+#endif
+    md.nr_msr = n;
+    md.done = 0;
+
+    return hax_sync_msr(env, &md, 1);
+}
+
+static int hax_get_fpu(CPUArchState *env)
+{
+    struct fx_layout fpu;
+    int i, ret;
+
+    ret = hax_sync_fpu(env, &fpu, 0);
+    if (ret < 0) {
+        return ret;
+    }
+
+    env->fpstt = (fpu.fsw >> 11) & 7;
+    env->fpus = fpu.fsw;
+    env->fpuc = fpu.fcw;
+    for (i = 0; i < 8; ++i) {
+        env->fptags[i] = !((fpu.ftw >> i) & 1);
+    }
+    memcpy(env->fpregs, fpu.st_mm, sizeof(env->fpregs));
+
+    for (i = 0; i < 8; i++) {
+        env->xmm_regs[i].ZMM_Q(0) = ldq_p(&fpu.mmx_1[i][0]);
+        env->xmm_regs[i].ZMM_Q(1) = ldq_p(&fpu.mmx_1[i][8]);
+        if (CPU_NB_REGS > 8) {
+            env->xmm_regs[i + 8].ZMM_Q(0) = ldq_p(&fpu.mmx_2[i][0]);
+            env->xmm_regs[i + 8].ZMM_Q(1) = ldq_p(&fpu.mmx_2[i][8]);
+        }
+    }
+    env->mxcsr = fpu.mxcsr;
+
+    return 0;
+}
+
+static int hax_set_fpu(CPUArchState *env)
+{
+    struct fx_layout fpu;
+    int i;
+
+    memset(&fpu, 0, sizeof(fpu));
+    fpu.fsw = env->fpus & ~(7 << 11);
+    fpu.fsw |= (env->fpstt & 7) << 11;
+    fpu.fcw = env->fpuc;
+
+    for (i = 0; i < 8; ++i) {
+        fpu.ftw |= (!env->fptags[i]) << i;
+    }
+
+    memcpy(fpu.st_mm, env->fpregs, sizeof(env->fpregs));
+    for (i = 0; i < 8; i++) {
+        stq_p(&fpu.mmx_1[i][0], env->xmm_regs[i].ZMM_Q(0));
+        stq_p(&fpu.mmx_1[i][8], env->xmm_regs[i].ZMM_Q(1));
+        if (CPU_NB_REGS > 8) {
+            stq_p(&fpu.mmx_2[i][0], env->xmm_regs[i + 8].ZMM_Q(0));
+            stq_p(&fpu.mmx_2[i][8], env->xmm_regs[i + 8].ZMM_Q(1));
+        }
+    }
+
+    fpu.mxcsr = env->mxcsr;
+
+    return hax_sync_fpu(env, &fpu, 1);
+}
+
+static int hax_arch_get_registers(CPUArchState *env)
+{
+    int ret;
+
+    ret = hax_sync_vcpu_register(env, 0);
+    if (ret < 0) {
+        return ret;
+    }
+
+    ret = hax_get_fpu(env);
+    if (ret < 0) {
+        return ret;
+    }
+
+    ret = hax_get_msrs(env);
+    if (ret < 0) {
+        return ret;
+    }
+
+    x86_update_hflags(env);
+    return 0;
+}
+
+static int hax_arch_set_registers(CPUArchState *env)
+{
+    int ret;
+    ret = hax_sync_vcpu_register(env, 1);
+
+    if (ret < 0) {
+        fprintf(stderr, "Failed to sync vcpu reg\n");
+        return ret;
+    }
+    ret = hax_set_fpu(env);
+    if (ret < 0) {
+        fprintf(stderr, "FPU failed\n");
+        return ret;
+    }
+    ret = hax_set_msrs(env);
+    if (ret < 0) {
+        fprintf(stderr, "MSR failed\n");
+        return ret;
+    }
+
+    return 0;
+}
+
+static void hax_vcpu_sync_state(CPUArchState *env, int modified)
+{
+    if (hax_enabled()) {
+        if (modified) {
+            hax_arch_set_registers(env);
+        } else {
+            hax_arch_get_registers(env);
+        }
+    }
+}
+
+/*
+ * much simpler than kvm, at least in first stage because:
+ * We don't need consider the device pass-through, we don't need
+ * consider the framebuffer, and we may even remove the bios at all
+ */
+int hax_sync_vcpus(void)
+{
+    if (hax_enabled()) {
+        CPUState *cpu;
+
+        cpu = first_cpu;
+        if (!cpu) {
+            return 0;
+        }
+
+        for (; cpu != NULL; cpu = CPU_NEXT(cpu)) {
+            int ret;
+
+            ret = hax_arch_set_registers(cpu->env_ptr);
+            if (ret < 0) {
+                return ret;
+            }
+        }
+    }
+
+    return 0;
+}
+
+void hax_reset_vcpu_state(void *opaque)
+{
+    CPUState *cpu;
+    for (cpu = first_cpu; cpu != NULL; cpu = CPU_NEXT(cpu)) {
+        cpu->hax_vcpu->tunnel->user_event_pending = 0;
+        cpu->hax_vcpu->tunnel->ready_for_interrupt_injection = 0;
+    }
+}
+
+static void hax_accel_class_init(ObjectClass *oc, void *data)
+{
+    AccelClass *ac = ACCEL_CLASS(oc);
+    ac->name = "HAX";
+    ac->init_machine = hax_accel_init;
+    ac->allowed = &hax_allowed;
+}
+
+static const TypeInfo hax_accel_type = {
+    .name = ACCEL_CLASS_NAME("hax"),
+    .parent = TYPE_ACCEL,
+    .class_init = hax_accel_class_init,
+};
+
+static void hax_type_init(void)
+{
+    type_register_static(&hax_accel_type);
+}
+
+type_init(hax_type_init);
diff --git a/target/i386/hax/hax-cpus.c b/target/i386/hax/hax-cpus.c
new file mode 100644 (file)
index 0000000..f72c85b
--- /dev/null
@@ -0,0 +1,85 @@
+/*
+ * QEMU HAX support
+ *
+ * Copyright IBM, Corp. 2008
+ *           Red Hat, Inc. 2008
+ *
+ * Authors:
+ *  Anthony Liguori   <aliguori@us.ibm.com>
+ *  Glauber Costa     <gcosta@redhat.com>
+ *
+ * Copyright (c) 2011 Intel Corporation
+ *  Written by:
+ *  Jiang Yunhong<yunhong.jiang@intel.com>
+ *  Xin Xiaohui<xiaohui.xin@intel.com>
+ *  Zhang Xiantao<xiantao.zhang@intel.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "qemu/error-report.h"
+#include "qemu/main-loop.h"
+#include "sysemu/runstate.h"
+#include "sysemu/cpus.h"
+#include "qemu/guest-random.h"
+
+#include "hax-cpus.h"
+
+static void *hax_cpu_thread_fn(void *arg)
+{
+    CPUState *cpu = arg;
+    int r;
+
+    rcu_register_thread();
+    qemu_mutex_lock_iothread();
+    qemu_thread_get_self(cpu->thread);
+
+    cpu->thread_id = qemu_get_thread_id();
+    current_cpu = cpu;
+    hax_init_vcpu(cpu);
+    cpu_thread_signal_created(cpu);
+    qemu_guest_random_seed_thread_part2(cpu->random_seed);
+
+    do {
+        if (cpu_can_run(cpu)) {
+            r = hax_smp_cpu_exec(cpu);
+            if (r == EXCP_DEBUG) {
+                cpu_handle_guest_debug(cpu);
+            }
+        }
+
+        qemu_wait_io_event(cpu);
+    } while (!cpu->unplug || cpu_can_run(cpu));
+    rcu_unregister_thread();
+    return NULL;
+}
+
+static void hax_start_vcpu_thread(CPUState *cpu)
+{
+    char thread_name[VCPU_THREAD_NAME_SIZE];
+
+    cpu->thread = g_malloc0(sizeof(QemuThread));
+    cpu->halt_cond = g_malloc0(sizeof(QemuCond));
+    qemu_cond_init(cpu->halt_cond);
+
+    snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/HAX",
+             cpu->cpu_index);
+    qemu_thread_create(cpu->thread, thread_name, hax_cpu_thread_fn,
+                       cpu, QEMU_THREAD_JOINABLE);
+#ifdef _WIN32
+    cpu->hThread = qemu_thread_get_handle(cpu->thread);
+#endif
+}
+
+const CpusAccel hax_cpus = {
+    .create_vcpu_thread = hax_start_vcpu_thread,
+    .kick_vcpu_thread = hax_kick_vcpu_thread,
+
+    .synchronize_post_reset = hax_cpu_synchronize_post_reset,
+    .synchronize_post_init = hax_cpu_synchronize_post_init,
+    .synchronize_state = hax_cpu_synchronize_state,
+    .synchronize_pre_loadvm = hax_cpu_synchronize_pre_loadvm,
+};
diff --git a/target/i386/hax/hax-cpus.h b/target/i386/hax/hax-cpus.h
new file mode 100644 (file)
index 0000000..ee8ab7a
--- /dev/null
@@ -0,0 +1,33 @@
+/*
+ * Accelerator CPUS Interface
+ *
+ * Copyright 2020 SUSE LLC
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ */
+
+#ifndef HAX_CPUS_H
+#define HAX_CPUS_H
+
+#include "sysemu/cpus.h"
+
+extern const CpusAccel hax_cpus;
+
+#include "hax-interface.h"
+#include "hax-i386.h"
+
+int hax_init_vcpu(CPUState *cpu);
+int hax_smp_cpu_exec(CPUState *cpu);
+int hax_populate_ram(uint64_t va, uint64_t size);
+
+void hax_cpu_synchronize_state(CPUState *cpu);
+void hax_cpu_synchronize_post_reset(CPUState *cpu);
+void hax_cpu_synchronize_post_init(CPUState *cpu);
+void hax_cpu_synchronize_pre_loadvm(CPUState *cpu);
+
+int hax_vcpu_destroy(CPUState *cpu);
+void hax_raise_event(CPUState *cpu);
+void hax_reset_vcpu_state(void *opaque);
+
+#endif /* HAX_CPUS_H */
diff --git a/target/i386/hax/hax-i386.h b/target/i386/hax/hax-i386.h
new file mode 100644 (file)
index 0000000..efbb346
--- /dev/null
@@ -0,0 +1,96 @@
+/*
+ * QEMU HAXM support
+ *
+ * Copyright (c) 2011 Intel Corporation
+ *  Written by:
+ *  Jiang Yunhong<yunhong.jiang@intel.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#ifndef HAX_I386_H
+#define HAX_I386_H
+
+#include "cpu.h"
+#include "sysemu/hax.h"
+
+#ifdef CONFIG_POSIX
+typedef int hax_fd;
+#endif
+
+#ifdef CONFIG_WIN32
+typedef HANDLE hax_fd;
+#endif
+
+extern struct hax_state hax_global;
+struct hax_vcpu_state {
+    hax_fd fd;
+    int vcpu_id;
+    struct hax_tunnel *tunnel;
+    unsigned char *iobuf;
+};
+
+struct hax_state {
+    hax_fd fd; /* the global hax device interface */
+    uint32_t version;
+    struct hax_vm *vm;
+    uint64_t mem_quota;
+    bool supports_64bit_ramblock;
+};
+
+#define HAX_MAX_VCPU 0x10
+
+struct hax_vm {
+    hax_fd fd;
+    int id;
+    int numvcpus;
+    struct hax_vcpu_state **vcpus;
+};
+
+#ifdef NEED_CPU_H
+/* Functions exported to host specific mode */
+hax_fd hax_vcpu_get_fd(CPUArchState *env);
+int valid_hax_tunnel_size(uint16_t size);
+
+/* Host specific functions */
+int hax_mod_version(struct hax_state *hax, struct hax_module_version *version);
+int hax_inject_interrupt(CPUArchState *env, int vector);
+struct hax_vm *hax_vm_create(struct hax_state *hax, int max_cpus);
+int hax_vcpu_run(struct hax_vcpu_state *vcpu);
+int hax_vcpu_create(int id);
+void hax_kick_vcpu_thread(CPUState *cpu);
+
+int hax_sync_vcpu_state(CPUArchState *env, struct vcpu_state_t *state,
+                        int set);
+int hax_sync_msr(CPUArchState *env, struct hax_msr_data *msrs, int set);
+int hax_sync_fpu(CPUArchState *env, struct fx_layout *fl, int set);
+#endif
+
+int hax_vm_destroy(struct hax_vm *vm);
+int hax_capability(struct hax_state *hax, struct hax_capabilityinfo *cap);
+int hax_notify_qemu_version(hax_fd vm_fd, struct hax_qemu_version *qversion);
+int hax_set_ram(uint64_t start_pa, uint32_t size, uint64_t host_va, int flags);
+
+/* Common host function */
+int hax_host_create_vm(struct hax_state *hax, int *vm_id);
+hax_fd hax_host_open_vm(struct hax_state *hax, int vm_id);
+int hax_host_create_vcpu(hax_fd vm_fd, int vcpuid);
+hax_fd hax_host_open_vcpu(int vmid, int vcpuid);
+int hax_host_setup_vcpu_channel(struct hax_vcpu_state *vcpu);
+hax_fd hax_mod_open(void);
+void hax_memory_init(void);
+
+
+#ifdef CONFIG_POSIX
+#include "hax-posix.h"
+#endif
+
+#ifdef CONFIG_WIN32
+#include "hax-windows.h"
+#endif
+
+#include "hax-interface.h"
+
+#endif
diff --git a/target/i386/hax/hax-interface.h b/target/i386/hax/hax-interface.h
new file mode 100644 (file)
index 0000000..537ae08
--- /dev/null
@@ -0,0 +1,369 @@
+/*
+ * QEMU HAXM support
+ *
+ * Copyright (c) 2011 Intel Corporation
+ *  Written by:
+ *  Jiang Yunhong<yunhong.jiang@intel.com>
+ *  Xin Xiaohui<xiaohui.xin@intel.com>
+ *  Zhang Xiantao<xiantao.zhang@intel.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+/* Interface with HAX kernel module */
+
+#ifndef HAX_INTERFACE_H
+#define HAX_INTERFACE_H
+
+/* fx_layout has 3 formats table 3-56, 512bytes */
+struct fx_layout {
+    uint16_t fcw;
+    uint16_t fsw;
+    uint8_t ftw;
+    uint8_t res1;
+    uint16_t fop;
+    union {
+        struct {
+            uint32_t fip;
+            uint16_t fcs;
+            uint16_t res2;
+        };
+        uint64_t fpu_ip;
+    };
+    union {
+        struct {
+            uint32_t fdp;
+            uint16_t fds;
+            uint16_t res3;
+        };
+        uint64_t fpu_dp;
+    };
+    uint32_t mxcsr;
+    uint32_t mxcsr_mask;
+    uint8_t st_mm[8][16];
+    uint8_t mmx_1[8][16];
+    uint8_t mmx_2[8][16];
+    uint8_t pad[96];
+} __attribute__ ((aligned(8)));
+
+struct vmx_msr {
+    uint64_t entry;
+    uint64_t value;
+} __attribute__ ((__packed__));
+
+/*
+ * Fixed array is not good, but it makes Mac support a bit easier by avoiding
+ * memory map or copyin staff.
+ */
+#define HAX_MAX_MSR_ARRAY 0x20
+struct hax_msr_data {
+    uint16_t nr_msr;
+    uint16_t done;
+    uint16_t pad[2];
+    struct vmx_msr entries[HAX_MAX_MSR_ARRAY];
+} __attribute__ ((__packed__));
+
+union interruptibility_state_t {
+    uint32_t raw;
+    struct {
+        uint32_t sti_blocking:1;
+        uint32_t movss_blocking:1;
+        uint32_t smi_blocking:1;
+        uint32_t nmi_blocking:1;
+        uint32_t reserved:28;
+    };
+    uint64_t pad;
+};
+
+typedef union interruptibility_state_t interruptibility_state_t;
+
+/* Segment descriptor */
+struct segment_desc_t {
+    uint16_t selector;
+    uint16_t _dummy;
+    uint32_t limit;
+    uint64_t base;
+    union {
+        struct {
+            uint32_t type:4;
+            uint32_t desc:1;
+            uint32_t dpl:2;
+            uint32_t present:1;
+            uint32_t:4;
+            uint32_t available:1;
+            uint32_t long_mode:1;
+            uint32_t operand_size:1;
+            uint32_t granularity:1;
+            uint32_t null:1;
+            uint32_t:15;
+        };
+        uint32_t ar;
+    };
+    uint32_t ipad;
+};
+
+typedef struct segment_desc_t segment_desc_t;
+
+struct vcpu_state_t {
+    union {
+        uint64_t _regs[16];
+        struct {
+            union {
+                struct {
+                    uint8_t _al, _ah;
+                };
+                uint16_t _ax;
+                uint32_t _eax;
+                uint64_t _rax;
+            };
+            union {
+                struct {
+                    uint8_t _cl, _ch;
+                };
+                uint16_t _cx;
+                uint32_t _ecx;
+                uint64_t _rcx;
+            };
+            union {
+                struct {
+                    uint8_t _dl, _dh;
+                };
+                uint16_t _dx;
+                uint32_t _edx;
+                uint64_t _rdx;
+            };
+            union {
+                struct {
+                    uint8_t _bl, _bh;
+                };
+                uint16_t _bx;
+                uint32_t _ebx;
+                uint64_t _rbx;
+            };
+            union {
+                uint16_t _sp;
+                uint32_t _esp;
+                uint64_t _rsp;
+            };
+            union {
+                uint16_t _bp;
+                uint32_t _ebp;
+                uint64_t _rbp;
+            };
+            union {
+                uint16_t _si;
+                uint32_t _esi;
+                uint64_t _rsi;
+            };
+            union {
+                uint16_t _di;
+                uint32_t _edi;
+                uint64_t _rdi;
+            };
+
+            uint64_t _r8;
+            uint64_t _r9;
+            uint64_t _r10;
+            uint64_t _r11;
+            uint64_t _r12;
+            uint64_t _r13;
+            uint64_t _r14;
+            uint64_t _r15;
+        };
+    };
+
+    union {
+        uint32_t _eip;
+        uint64_t _rip;
+    };
+
+    union {
+        uint32_t _eflags;
+        uint64_t _rflags;
+    };
+
+    segment_desc_t _cs;
+    segment_desc_t _ss;
+    segment_desc_t _ds;
+    segment_desc_t _es;
+    segment_desc_t _fs;
+    segment_desc_t _gs;
+    segment_desc_t _ldt;
+    segment_desc_t _tr;
+
+    segment_desc_t _gdt;
+    segment_desc_t _idt;
+
+    uint64_t _cr0;
+    uint64_t _cr2;
+    uint64_t _cr3;
+    uint64_t _cr4;
+
+    uint64_t _dr0;
+    uint64_t _dr1;
+    uint64_t _dr2;
+    uint64_t _dr3;
+    uint64_t _dr6;
+    uint64_t _dr7;
+    uint64_t _pde;
+
+    uint32_t _efer;
+
+    uint32_t _sysenter_cs;
+    uint64_t _sysenter_eip;
+    uint64_t _sysenter_esp;
+
+    uint32_t _activity_state;
+    uint32_t pad;
+    interruptibility_state_t _interruptibility_state;
+};
+
+/* HAX exit status */
+enum exit_status {
+    /* IO port request */
+    HAX_EXIT_IO = 1,
+    /* MMIO instruction emulation */
+    HAX_EXIT_MMIO,
+    /* QEMU emulation mode request, currently means guest enter non-PG mode */
+    HAX_EXIT_REAL,
+    /*
+     * Interrupt window open, qemu can inject interrupt now
+     * Also used when signal pending since at that time qemu usually need
+     * check interrupt
+     */
+    HAX_EXIT_INTERRUPT,
+    /* Unknown vmexit, mostly trigger reboot */
+    HAX_EXIT_UNKNOWN_VMEXIT,
+    /* HALT from guest */
+    HAX_EXIT_HLT,
+    /* Reboot request, like because of tripple fault in guest */
+    HAX_EXIT_STATECHANGE,
+    /* the vcpu is now only paused when destroy, so simply return to hax */
+    HAX_EXIT_PAUSED,
+    HAX_EXIT_FAST_MMIO,
+};
+
+/*
+ * The interface definition:
+ * 1. vcpu_run execute will return 0 on success, otherwise mean failed
+ * 2. exit_status return the exit reason, as stated in enum exit_status
+ * 3. exit_reason is the vmx exit reason
+ */
+struct hax_tunnel {
+    uint32_t _exit_reason;
+    uint32_t _exit_flag;
+    uint32_t _exit_status;
+    uint32_t user_event_pending;
+    int ready_for_interrupt_injection;
+    int request_interrupt_window;
+    union {
+        struct {
+            /* 0: read, 1: write */
+#define HAX_EXIT_IO_IN  1
+#define HAX_EXIT_IO_OUT 0
+            uint8_t _direction;
+            uint8_t _df;
+            uint16_t _size;
+            uint16_t _port;
+            uint16_t _count;
+            uint8_t _flags;
+            uint8_t _pad0;
+            uint16_t _pad1;
+            uint32_t _pad2;
+            uint64_t _vaddr;
+        } pio;
+        struct {
+            uint64_t gla;
+        } mmio;
+        struct {
+        } state;
+    };
+} __attribute__ ((__packed__));
+
+struct hax_module_version {
+    uint32_t compat_version;
+    uint32_t cur_version;
+} __attribute__ ((__packed__));
+
+/* This interface is support only after API version 2 */
+struct hax_qemu_version {
+    /* Current API version in QEMU */
+    uint32_t cur_version;
+    /* The minimum API version supported by QEMU */
+    uint32_t min_version;
+} __attribute__ ((__packed__));
+
+/* The mac specfic interface to qemu, mostly is ioctl related */
+struct hax_tunnel_info {
+    uint64_t va;
+    uint64_t io_va;
+    uint16_t size;
+    uint16_t pad[3];
+} __attribute__ ((__packed__));
+
+struct hax_alloc_ram_info {
+    uint32_t size;
+    uint32_t pad;
+    uint64_t va;
+} __attribute__ ((__packed__));
+
+struct hax_ramblock_info {
+    uint64_t start_va;
+    uint64_t size;
+    uint64_t reserved;
+} __attribute__ ((__packed__));
+
+#define HAX_RAM_INFO_ROM     0x01 /* Read-Only */
+#define HAX_RAM_INFO_INVALID 0x80 /* Unmapped, usually used for MMIO */
+struct hax_set_ram_info {
+    uint64_t pa_start;
+    uint32_t size;
+    uint8_t flags;
+    uint8_t pad[3];
+    uint64_t va;
+} __attribute__ ((__packed__));
+
+#define HAX_CAP_STATUS_WORKING     0x1
+#define HAX_CAP_STATUS_NOTWORKING  0x0
+#define HAX_CAP_WORKSTATUS_MASK    0x1
+
+#define HAX_CAP_FAILREASON_VT      0x1
+#define HAX_CAP_FAILREASON_NX      0x2
+
+#define HAX_CAP_MEMQUOTA           0x2
+#define HAX_CAP_UG                 0x4
+#define HAX_CAP_64BIT_RAMBLOCK     0x8
+
+struct hax_capabilityinfo {
+    /* bit 0: 1 - working
+     *        0 - not working, possibly because NT/NX disabled
+     * bit 1: 1 - memory limitation working
+     *        0 - no memory limitation
+     */
+    uint16_t wstatus;
+    /* valid when not working
+     * bit 0: VT not enabeld
+     * bit 1: NX not enabled*/
+    uint16_t winfo;
+    uint32_t pad;
+    uint64_t mem_quota;
+} __attribute__ ((__packed__));
+
+struct hax_fastmmio {
+    uint64_t gpa;
+    union {
+        uint64_t value;
+        uint64_t gpa2;  /* since HAX API v4 */
+    };
+    uint8_t size;
+    uint8_t direction;
+    uint16_t reg_index;
+    uint32_t pad0;
+    uint64_t _cr0;
+    uint64_t _cr2;
+    uint64_t _cr3;
+    uint64_t _cr4;
+} __attribute__ ((__packed__));
+#endif
diff --git a/target/i386/hax/hax-mem.c b/target/i386/hax/hax-mem.c
new file mode 100644 (file)
index 0000000..71e637c
--- /dev/null
@@ -0,0 +1,321 @@
+/*
+ * HAX memory mapping operations
+ *
+ * Copyright (c) 2015-16 Intel Corporation
+ * Copyright 2016 Google, Inc.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2.  See
+ * the COPYING file in the top-level directory.
+ */
+
+#include "qemu/osdep.h"
+#include "cpu.h"
+#include "exec/address-spaces.h"
+#include "qemu/error-report.h"
+
+#include "hax-cpus.h"
+#include "qemu/queue.h"
+
+#define DEBUG_HAX_MEM 0
+
+#define DPRINTF(fmt, ...) \
+    do { \
+        if (DEBUG_HAX_MEM) { \
+            fprintf(stdout, fmt, ## __VA_ARGS__); \
+        } \
+    } while (0)
+
+/**
+ * HAXMapping: describes a pending guest physical memory mapping
+ *
+ * @start_pa: a guest physical address marking the start of the region; must be
+ *            page-aligned
+ * @size: a guest physical address marking the end of the region; must be
+ *          page-aligned
+ * @host_va: the host virtual address of the start of the mapping
+ * @flags: mapping parameters e.g. HAX_RAM_INFO_ROM or HAX_RAM_INFO_INVALID
+ * @entry: additional fields for linking #HAXMapping instances together
+ */
+typedef struct HAXMapping {
+    uint64_t start_pa;
+    uint32_t size;
+    uint64_t host_va;
+    int flags;
+    QTAILQ_ENTRY(HAXMapping) entry;
+} HAXMapping;
+
+/*
+ * A doubly-linked list (actually a tail queue) of the pending page mappings
+ * for the ongoing memory transaction.
+ *
+ * It is used to optimize the number of page mapping updates done through the
+ * kernel module. For example, it's effective when a driver is digging an MMIO
+ * hole inside an existing memory mapping. It will get a deletion of the whole
+ * region, then the addition of the 2 remaining RAM areas around the hole and
+ * finally the memory transaction commit. During the commit, it will effectively
+ * send to the kernel only the removal of the pages from the MMIO hole after
+ * having computed locally the result of the deletion and additions.
+ */
+static QTAILQ_HEAD(, HAXMapping) mappings =
+    QTAILQ_HEAD_INITIALIZER(mappings);
+
+/**
+ * hax_mapping_dump_list: dumps @mappings to stdout (for debugging)
+ */
+static void hax_mapping_dump_list(void)
+{
+    HAXMapping *entry;
+
+    DPRINTF("%s updates:\n", __func__);
+    QTAILQ_FOREACH(entry, &mappings, entry) {
+        DPRINTF("\t%c 0x%016" PRIx64 "->0x%016" PRIx64 " VA 0x%016" PRIx64
+                "%s\n", entry->flags & HAX_RAM_INFO_INVALID ? '-' : '+',
+                entry->start_pa, entry->start_pa + entry->size, entry->host_va,
+                entry->flags & HAX_RAM_INFO_ROM ? " ROM" : "");
+    }
+}
+
+static void hax_insert_mapping_before(HAXMapping *next, uint64_t start_pa,
+                                      uint32_t size, uint64_t host_va,
+                                      uint8_t flags)
+{
+    HAXMapping *entry;
+
+    entry = g_malloc0(sizeof(*entry));
+    entry->start_pa = start_pa;
+    entry->size = size;
+    entry->host_va = host_va;
+    entry->flags = flags;
+    if (!next) {
+        QTAILQ_INSERT_TAIL(&mappings, entry, entry);
+    } else {
+        QTAILQ_INSERT_BEFORE(next, entry, entry);
+    }
+}
+
+static bool hax_mapping_is_opposite(HAXMapping *entry, uint64_t host_va,
+                                    uint8_t flags)
+{
+    /* removed then added without change for the read-only flag */
+    bool nop_flags = (entry->flags ^ flags) == HAX_RAM_INFO_INVALID;
+
+    return (entry->host_va == host_va) && nop_flags;
+}
+
+static void hax_update_mapping(uint64_t start_pa, uint32_t size,
+                               uint64_t host_va, uint8_t flags)
+{
+    uint64_t end_pa = start_pa + size;
+    HAXMapping *entry, *next;
+
+    QTAILQ_FOREACH_SAFE(entry, &mappings, entry, next) {
+        uint32_t chunk_sz;
+        if (start_pa >= entry->start_pa + entry->size) {
+            continue;
+        }
+        if (start_pa < entry->start_pa) {
+            chunk_sz = end_pa <= entry->start_pa ? size
+                                                 : entry->start_pa - start_pa;
+            hax_insert_mapping_before(entry, start_pa, chunk_sz,
+                                      host_va, flags);
+            start_pa += chunk_sz;
+            host_va += chunk_sz;
+            size -= chunk_sz;
+        } else if (start_pa > entry->start_pa) {
+            /* split the existing chunk at start_pa */
+            chunk_sz = start_pa - entry->start_pa;
+            hax_insert_mapping_before(entry, entry->start_pa, chunk_sz,
+                                      entry->host_va, entry->flags);
+            entry->start_pa += chunk_sz;
+            entry->host_va += chunk_sz;
+            entry->size -= chunk_sz;
+        }
+        /* now start_pa == entry->start_pa */
+        chunk_sz = MIN(size, entry->size);
+        if (chunk_sz) {
+            bool nop = hax_mapping_is_opposite(entry, host_va, flags);
+            bool partial = chunk_sz < entry->size;
+            if (partial) {
+                /* remove the beginning of the existing chunk */
+                entry->start_pa += chunk_sz;
+                entry->host_va += chunk_sz;
+                entry->size -= chunk_sz;
+                if (!nop) {
+                    hax_insert_mapping_before(entry, start_pa, chunk_sz,
+                                              host_va, flags);
+                }
+            } else { /* affects the full mapping entry */
+                if (nop) { /* no change to this mapping, remove it */
+                    QTAILQ_REMOVE(&mappings, entry, entry);
+                    g_free(entry);
+                } else { /* update mapping properties */
+                    entry->host_va = host_va;
+                    entry->flags = flags;
+                }
+            }
+            start_pa += chunk_sz;
+            host_va += chunk_sz;
+            size -= chunk_sz;
+        }
+        if (!size) { /* we are done */
+            break;
+        }
+    }
+    if (size) { /* add the leftover */
+        hax_insert_mapping_before(NULL, start_pa, size, host_va, flags);
+    }
+}
+
+static void hax_process_section(MemoryRegionSection *section, uint8_t flags)
+{
+    MemoryRegion *mr = section->mr;
+    hwaddr start_pa = section->offset_within_address_space;
+    ram_addr_t size = int128_get64(section->size);
+    unsigned int delta;
+    uint64_t host_va;
+    uint32_t max_mapping_size;
+
+    /* We only care about RAM and ROM regions */
+    if (!memory_region_is_ram(mr)) {
+        if (memory_region_is_romd(mr)) {
+            /* HAXM kernel module does not support ROMD yet  */
+            warn_report("Ignoring ROMD region 0x%016" PRIx64 "->0x%016" PRIx64,
+                        start_pa, start_pa + size);
+        }
+        return;
+    }
+
+    /* Adjust start_pa and size so that they are page-aligned. (Cf
+     * kvm_set_phys_mem() in kvm-all.c).
+     */
+    delta = qemu_real_host_page_size - (start_pa & ~qemu_real_host_page_mask);
+    delta &= ~qemu_real_host_page_mask;
+    if (delta > size) {
+        return;
+    }
+    start_pa += delta;
+    size -= delta;
+    size &= qemu_real_host_page_mask;
+    if (!size || (start_pa & ~qemu_real_host_page_mask)) {
+        return;
+    }
+
+    host_va = (uintptr_t)memory_region_get_ram_ptr(mr)
+            + section->offset_within_region + delta;
+    if (memory_region_is_rom(section->mr)) {
+        flags |= HAX_RAM_INFO_ROM;
+    }
+
+    /*
+     * The kernel module interface uses 32-bit sizes:
+     * https://github.com/intel/haxm/blob/master/API.md#hax_vm_ioctl_set_ram
+     *
+     * If the mapping size is longer than 32 bits, we can't process it in one
+     * call into the kernel. Instead, we split the mapping into smaller ones,
+     * and call hax_update_mapping() on each.
+     */
+    max_mapping_size = UINT32_MAX & qemu_real_host_page_mask;
+    while (size > max_mapping_size) {
+        hax_update_mapping(start_pa, max_mapping_size, host_va, flags);
+        start_pa += max_mapping_size;
+        size -= max_mapping_size;
+        host_va += max_mapping_size;
+    }
+    /* Now size <= max_mapping_size */
+    hax_update_mapping(start_pa, (uint32_t)size, host_va, flags);
+}
+
+static void hax_region_add(MemoryListener *listener,
+                           MemoryRegionSection *section)
+{
+    memory_region_ref(section->mr);
+    hax_process_section(section, 0);
+}
+
+static void hax_region_del(MemoryListener *listener,
+                           MemoryRegionSection *section)
+{
+    hax_process_section(section, HAX_RAM_INFO_INVALID);
+    memory_region_unref(section->mr);
+}
+
+static void hax_transaction_begin(MemoryListener *listener)
+{
+    g_assert(QTAILQ_EMPTY(&mappings));
+}
+
+static void hax_transaction_commit(MemoryListener *listener)
+{
+    if (!QTAILQ_EMPTY(&mappings)) {
+        HAXMapping *entry, *next;
+
+        if (DEBUG_HAX_MEM) {
+            hax_mapping_dump_list();
+        }
+        QTAILQ_FOREACH_SAFE(entry, &mappings, entry, next) {
+            if (entry->flags & HAX_RAM_INFO_INVALID) {
+                /* for unmapping, put the values expected by the kernel */
+                entry->flags = HAX_RAM_INFO_INVALID;
+                entry->host_va = 0;
+            }
+            if (hax_set_ram(entry->start_pa, entry->size,
+                            entry->host_va, entry->flags)) {
+                fprintf(stderr, "%s: Failed mapping @0x%016" PRIx64 "+0x%"
+                        PRIx32 " flags %02x\n", __func__, entry->start_pa,
+                        entry->size, entry->flags);
+            }
+            QTAILQ_REMOVE(&mappings, entry, entry);
+            g_free(entry);
+        }
+    }
+}
+
+/* currently we fake the dirty bitmap sync, always dirty */
+static void hax_log_sync(MemoryListener *listener,
+                         MemoryRegionSection *section)
+{
+    MemoryRegion *mr = section->mr;
+
+    if (!memory_region_is_ram(mr)) {
+        /* Skip MMIO regions */
+        return;
+    }
+
+    memory_region_set_dirty(mr, 0, int128_get64(section->size));
+}
+
+static MemoryListener hax_memory_listener = {
+    .begin = hax_transaction_begin,
+    .commit = hax_transaction_commit,
+    .region_add = hax_region_add,
+    .region_del = hax_region_del,
+    .log_sync = hax_log_sync,
+    .priority = 10,
+};
+
+static void hax_ram_block_added(RAMBlockNotifier *n, void *host, size_t size)
+{
+    /*
+     * We must register each RAM block with the HAXM kernel module, or
+     * hax_set_ram() will fail for any mapping into the RAM block:
+     * https://github.com/intel/haxm/blob/master/API.md#hax_vm_ioctl_alloc_ram
+     *
+     * Old versions of the HAXM kernel module (< 6.2.0) used to preallocate all
+     * host physical pages for the RAM block as part of this registration
+     * process, hence the name hax_populate_ram().
+     */
+    if (hax_populate_ram((uint64_t)(uintptr_t)host, size) < 0) {
+        fprintf(stderr, "HAX failed to populate RAM\n");
+        abort();
+    }
+}
+
+static struct RAMBlockNotifier hax_ram_notifier = {
+    .ram_block_added = hax_ram_block_added,
+};
+
+void hax_memory_init(void)
+{
+    ram_block_notifier_add(&hax_ram_notifier);
+    memory_listener_register(&hax_memory_listener, &address_space_memory);
+}
diff --git a/target/i386/hax/hax-posix.c b/target/i386/hax/hax-posix.c
new file mode 100644 (file)
index 0000000..735a749
--- /dev/null
@@ -0,0 +1,305 @@
+/*
+ * QEMU HAXM support
+ *
+ * Copyright (c) 2011 Intel Corporation
+ *  Written by:
+ *  Jiang Yunhong<yunhong.jiang@intel.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+/* HAX module interface - darwin version */
+#include "qemu/osdep.h"
+#include <sys/ioctl.h>
+
+#include "sysemu/cpus.h"
+#include "hax-cpus.h"
+
+hax_fd hax_mod_open(void)
+{
+    int fd = open("/dev/HAX", O_RDWR);
+    if (fd == -1) {
+        fprintf(stderr, "Failed to open the hax module\n");
+    }
+
+    qemu_set_cloexec(fd);
+
+    return fd;
+}
+
+int hax_populate_ram(uint64_t va, uint64_t size)
+{
+    int ret;
+
+    if (!hax_global.vm || !hax_global.vm->fd) {
+        fprintf(stderr, "Allocate memory before vm create?\n");
+        return -EINVAL;
+    }
+
+    if (hax_global.supports_64bit_ramblock) {
+        struct hax_ramblock_info ramblock = {
+            .start_va = va,
+            .size = size,
+            .reserved = 0
+        };
+
+        ret = ioctl(hax_global.vm->fd, HAX_VM_IOCTL_ADD_RAMBLOCK, &ramblock);
+    } else {
+        struct hax_alloc_ram_info info = {
+            .size = (uint32_t)size,
+            .pad = 0,
+            .va = va
+        };
+
+        ret = ioctl(hax_global.vm->fd, HAX_VM_IOCTL_ALLOC_RAM, &info);
+    }
+    if (ret < 0) {
+        fprintf(stderr, "Failed to register RAM block: ret=%d, va=0x%" PRIx64
+                ", size=0x%" PRIx64 ", method=%s\n", ret, va, size,
+                hax_global.supports_64bit_ramblock ? "new" : "legacy");
+        return ret;
+    }
+    return 0;
+}
+
+int hax_set_ram(uint64_t start_pa, uint32_t size, uint64_t host_va, int flags)
+{
+    struct hax_set_ram_info info;
+    int ret;
+
+    info.pa_start = start_pa;
+    info.size = size;
+    info.va = host_va;
+    info.flags = (uint8_t) flags;
+
+    ret = ioctl(hax_global.vm->fd, HAX_VM_IOCTL_SET_RAM, &info);
+    if (ret < 0) {
+        return -errno;
+    }
+    return 0;
+}
+
+int hax_capability(struct hax_state *hax, struct hax_capabilityinfo *cap)
+{
+    int ret;
+
+    ret = ioctl(hax->fd, HAX_IOCTL_CAPABILITY, cap);
+    if (ret == -1) {
+        fprintf(stderr, "Failed to get HAX capability\n");
+        return -errno;
+    }
+
+    return 0;
+}
+
+int hax_mod_version(struct hax_state *hax, struct hax_module_version *version)
+{
+    int ret;
+
+    ret = ioctl(hax->fd, HAX_IOCTL_VERSION, version);
+    if (ret == -1) {
+        fprintf(stderr, "Failed to get HAX version\n");
+        return -errno;
+    }
+
+    return 0;
+}
+
+static char *hax_vm_devfs_string(int vm_id)
+{
+    return g_strdup_printf("/dev/hax_vm/vm%02d", vm_id);
+}
+
+static char *hax_vcpu_devfs_string(int vm_id, int vcpu_id)
+{
+    return g_strdup_printf("/dev/hax_vm%02d/vcpu%02d", vm_id, vcpu_id);
+}
+
+int hax_host_create_vm(struct hax_state *hax, int *vmid)
+{
+    int ret;
+    int vm_id = 0;
+
+    if (hax_invalid_fd(hax->fd)) {
+        return -EINVAL;
+    }
+
+    if (hax->vm) {
+        return 0;
+    }
+
+    ret = ioctl(hax->fd, HAX_IOCTL_CREATE_VM, &vm_id);
+    *vmid = vm_id;
+    return ret;
+}
+
+hax_fd hax_host_open_vm(struct hax_state *hax, int vm_id)
+{
+    hax_fd fd;
+    char *vm_name = NULL;
+
+    vm_name = hax_vm_devfs_string(vm_id);
+    if (!vm_name) {
+        return -1;
+    }
+
+    fd = open(vm_name, O_RDWR);
+    g_free(vm_name);
+
+    qemu_set_cloexec(fd);
+
+    return fd;
+}
+
+int hax_notify_qemu_version(hax_fd vm_fd, struct hax_qemu_version *qversion)
+{
+    int ret;
+
+    if (hax_invalid_fd(vm_fd)) {
+        return -EINVAL;
+    }
+
+    ret = ioctl(vm_fd, HAX_VM_IOCTL_NOTIFY_QEMU_VERSION, qversion);
+
+    if (ret < 0) {
+        fprintf(stderr, "Failed to notify qemu API version\n");
+        return ret;
+    }
+    return 0;
+}
+
+/* Simply assume the size should be bigger than the hax_tunnel,
+ * since the hax_tunnel can be extended later with compatibility considered
+ */
+int hax_host_create_vcpu(hax_fd vm_fd, int vcpuid)
+{
+    int ret;
+
+    ret = ioctl(vm_fd, HAX_VM_IOCTL_VCPU_CREATE, &vcpuid);
+    if (ret < 0) {
+        fprintf(stderr, "Failed to create vcpu %x\n", vcpuid);
+    }
+
+    return ret;
+}
+
+hax_fd hax_host_open_vcpu(int vmid, int vcpuid)
+{
+    char *devfs_path = NULL;
+    hax_fd fd;
+
+    devfs_path = hax_vcpu_devfs_string(vmid, vcpuid);
+    if (!devfs_path) {
+        fprintf(stderr, "Failed to get the devfs\n");
+        return -EINVAL;
+    }
+
+    fd = open(devfs_path, O_RDWR);
+    g_free(devfs_path);
+    if (fd < 0) {
+        fprintf(stderr, "Failed to open the vcpu devfs\n");
+    }
+    qemu_set_cloexec(fd);
+    return fd;
+}
+
+int hax_host_setup_vcpu_channel(struct hax_vcpu_state *vcpu)
+{
+    int ret;
+    struct hax_tunnel_info info;
+
+    ret = ioctl(vcpu->fd, HAX_VCPU_IOCTL_SETUP_TUNNEL, &info);
+    if (ret) {
+        fprintf(stderr, "Failed to setup the hax tunnel\n");
+        return ret;
+    }
+
+    if (!valid_hax_tunnel_size(info.size)) {
+        fprintf(stderr, "Invalid hax tunnel size %x\n", info.size);
+        ret = -EINVAL;
+        return ret;
+    }
+
+    vcpu->tunnel = (struct hax_tunnel *) (intptr_t) (info.va);
+    vcpu->iobuf = (unsigned char *) (intptr_t) (info.io_va);
+    return 0;
+}
+
+int hax_vcpu_run(struct hax_vcpu_state *vcpu)
+{
+    return ioctl(vcpu->fd, HAX_VCPU_IOCTL_RUN, NULL);
+}
+
+int hax_sync_fpu(CPUArchState *env, struct fx_layout *fl, int set)
+{
+    int ret, fd;
+
+    fd = hax_vcpu_get_fd(env);
+    if (fd <= 0) {
+        return -1;
+    }
+
+    if (set) {
+        ret = ioctl(fd, HAX_VCPU_IOCTL_SET_FPU, fl);
+    } else {
+        ret = ioctl(fd, HAX_VCPU_IOCTL_GET_FPU, fl);
+    }
+    return ret;
+}
+
+int hax_sync_msr(CPUArchState *env, struct hax_msr_data *msrs, int set)
+{
+    int ret, fd;
+
+    fd = hax_vcpu_get_fd(env);
+    if (fd <= 0) {
+        return -1;
+    }
+    if (set) {
+        ret = ioctl(fd, HAX_VCPU_IOCTL_SET_MSRS, msrs);
+    } else {
+        ret = ioctl(fd, HAX_VCPU_IOCTL_GET_MSRS, msrs);
+    }
+    return ret;
+}
+
+int hax_sync_vcpu_state(CPUArchState *env, struct vcpu_state_t *state, int set)
+{
+    int ret, fd;
+
+    fd = hax_vcpu_get_fd(env);
+    if (fd <= 0) {
+        return -1;
+    }
+
+    if (set) {
+        ret = ioctl(fd, HAX_VCPU_SET_REGS, state);
+    } else {
+        ret = ioctl(fd, HAX_VCPU_GET_REGS, state);
+    }
+    return ret;
+}
+
+int hax_inject_interrupt(CPUArchState *env, int vector)
+{
+    int fd;
+
+    fd = hax_vcpu_get_fd(env);
+    if (fd <= 0) {
+        return -1;
+    }
+
+    return ioctl(fd, HAX_VCPU_IOCTL_INTERRUPT, &vector);
+}
+
+void hax_kick_vcpu_thread(CPUState *cpu)
+{
+    /*
+     * FIXME: race condition with the exit_request check in
+     * hax_vcpu_hax_exec
+     */
+    cpu->exit_request = 1;
+    cpus_kick_thread(cpu);
+}
diff --git a/target/i386/hax/hax-posix.h b/target/i386/hax/hax-posix.h
new file mode 100644 (file)
index 0000000..fb7c644
--- /dev/null
@@ -0,0 +1,61 @@
+/*
+ * QEMU HAXM support
+ *
+ * Copyright (c) 2011 Intel Corporation
+ *  Written by:
+ *  Jiang Yunhong<yunhong.jiang@intel.com>
+ *  Xin Xiaohui<xiaohui.xin@intel.com>
+ *  Zhang Xiantao<xiantao.zhang@intel.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#ifndef TARGET_I386_HAX_POSIX_H
+#define TARGET_I386_HAX_POSIX_H
+
+#include <sys/ioctl.h>
+
+#define HAX_INVALID_FD  (-1)
+static inline int hax_invalid_fd(hax_fd fd)
+{
+    return fd <= 0;
+}
+
+static inline void hax_mod_close(struct hax_state *hax)
+{
+    close(hax->fd);
+}
+
+static inline void hax_close_fd(hax_fd fd)
+{
+    close(fd);
+}
+
+/* HAX model level ioctl */
+#define HAX_IOCTL_VERSION _IOWR(0, 0x20, struct hax_module_version)
+#define HAX_IOCTL_CREATE_VM _IOWR(0, 0x21, uint32_t)
+#define HAX_IOCTL_DESTROY_VM _IOW(0, 0x22, uint32_t)
+#define HAX_IOCTL_CAPABILITY _IOR(0, 0x23, struct hax_capabilityinfo)
+
+#define HAX_VM_IOCTL_VCPU_CREATE _IOWR(0, 0x80, uint32_t)
+#define HAX_VM_IOCTL_ALLOC_RAM _IOWR(0, 0x81, struct hax_alloc_ram_info)
+#define HAX_VM_IOCTL_SET_RAM _IOWR(0, 0x82, struct hax_set_ram_info)
+#define HAX_VM_IOCTL_VCPU_DESTROY _IOW(0, 0x83, uint32_t)
+#define HAX_VM_IOCTL_NOTIFY_QEMU_VERSION _IOW(0, 0x84, struct hax_qemu_version)
+#define HAX_VM_IOCTL_ADD_RAMBLOCK _IOW(0, 0x85, struct hax_ramblock_info)
+
+#define HAX_VCPU_IOCTL_RUN  _IO(0, 0xc0)
+#define HAX_VCPU_IOCTL_SET_MSRS _IOWR(0, 0xc1, struct hax_msr_data)
+#define HAX_VCPU_IOCTL_GET_MSRS _IOWR(0, 0xc2, struct hax_msr_data)
+
+#define HAX_VCPU_IOCTL_SET_FPU  _IOW(0, 0xc3, struct fx_layout)
+#define HAX_VCPU_IOCTL_GET_FPU  _IOR(0, 0xc4, struct fx_layout)
+
+#define HAX_VCPU_IOCTL_SETUP_TUNNEL _IOWR(0, 0xc5, struct hax_tunnel_info)
+#define HAX_VCPU_IOCTL_INTERRUPT _IOWR(0, 0xc6, uint32_t)
+#define HAX_VCPU_SET_REGS       _IOWR(0, 0xc7, struct vcpu_state_t)
+#define HAX_VCPU_GET_REGS       _IOWR(0, 0xc8, struct vcpu_state_t)
+
+#endif /* TARGET_I386_HAX_POSIX_H */
diff --git a/target/i386/hax/hax-windows.c b/target/i386/hax/hax-windows.c
new file mode 100644 (file)
index 0000000..6c82dfb
--- /dev/null
@@ -0,0 +1,485 @@
+/*
+ * QEMU HAXM support
+ *
+ * Copyright (c) 2011 Intel Corporation
+ *  Written by:
+ *  Jiang Yunhong<yunhong.jiang@intel.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "cpu.h"
+#include "hax-cpus.h"
+
+/*
+ * return 0 when success, -1 when driver not loaded,
+ * other negative value for other failure
+ */
+static int hax_open_device(hax_fd *fd)
+{
+    uint32_t errNum = 0;
+    HANDLE hDevice;
+
+    if (!fd) {
+        return -2;
+    }
+
+    hDevice = CreateFile("\\\\.\\HAX",
+                         GENERIC_READ | GENERIC_WRITE,
+                         0, NULL, CREATE_ALWAYS, FILE_ATTRIBUTE_NORMAL, NULL);
+
+    if (hDevice == INVALID_HANDLE_VALUE) {
+        fprintf(stderr, "Failed to open the HAX device!\n");
+        errNum = GetLastError();
+        if (errNum == ERROR_FILE_NOT_FOUND) {
+            return -1;
+        }
+        return -2;
+    }
+    *fd = hDevice;
+    return 0;
+}
+
+/* hax_fd hax_mod_open */
+ hax_fd hax_mod_open(void)
+{
+    int ret;
+    hax_fd fd = NULL;
+
+    ret = hax_open_device(&fd);
+    if (ret != 0) {
+        fprintf(stderr, "Open HAX device failed\n");
+    }
+
+    return fd;
+}
+
+int hax_populate_ram(uint64_t va, uint64_t size)
+{
+    int ret;
+    HANDLE hDeviceVM;
+    DWORD dSize = 0;
+
+    if (!hax_global.vm || !hax_global.vm->fd) {
+        fprintf(stderr, "Allocate memory before vm create?\n");
+        return -EINVAL;
+    }
+
+    hDeviceVM = hax_global.vm->fd;
+    if (hax_global.supports_64bit_ramblock) {
+        struct hax_ramblock_info ramblock = {
+            .start_va = va,
+            .size = size,
+            .reserved = 0
+        };
+
+        ret = DeviceIoControl(hDeviceVM,
+                              HAX_VM_IOCTL_ADD_RAMBLOCK,
+                              &ramblock, sizeof(ramblock), NULL, 0, &dSize,
+                              (LPOVERLAPPED) NULL);
+    } else {
+        struct hax_alloc_ram_info info = {
+            .size = (uint32_t) size,
+            .pad = 0,
+            .va = va
+        };
+
+        ret = DeviceIoControl(hDeviceVM,
+                              HAX_VM_IOCTL_ALLOC_RAM,
+                              &info, sizeof(info), NULL, 0, &dSize,
+                              (LPOVERLAPPED) NULL);
+    }
+
+    if (!ret) {
+        fprintf(stderr, "Failed to register RAM block: va=0x%" PRIx64
+                ", size=0x%" PRIx64 ", method=%s\n", va, size,
+                hax_global.supports_64bit_ramblock ? "new" : "legacy");
+        return ret;
+    }
+
+    return 0;
+}
+
+int hax_set_ram(uint64_t start_pa, uint32_t size, uint64_t host_va, int flags)
+{
+    struct hax_set_ram_info info;
+    HANDLE hDeviceVM = hax_global.vm->fd;
+    DWORD dSize = 0;
+    int ret;
+
+    info.pa_start = start_pa;
+    info.size = size;
+    info.va = host_va;
+    info.flags = (uint8_t) flags;
+
+    ret = DeviceIoControl(hDeviceVM, HAX_VM_IOCTL_SET_RAM,
+                          &info, sizeof(info), NULL, 0, &dSize,
+                          (LPOVERLAPPED) NULL);
+
+    if (!ret) {
+        return -EFAULT;
+    } else {
+        return 0;
+    }
+}
+
+int hax_capability(struct hax_state *hax, struct hax_capabilityinfo *cap)
+{
+    int ret;
+    HANDLE hDevice = hax->fd;        /* handle to hax module */
+    DWORD dSize = 0;
+    DWORD err = 0;
+
+    if (hax_invalid_fd(hDevice)) {
+        fprintf(stderr, "Invalid fd for hax device!\n");
+        return -ENODEV;
+    }
+
+    ret = DeviceIoControl(hDevice, HAX_IOCTL_CAPABILITY, NULL, 0, cap,
+                          sizeof(*cap), &dSize, (LPOVERLAPPED) NULL);
+
+    if (!ret) {
+        err = GetLastError();
+        if (err == ERROR_INSUFFICIENT_BUFFER || err == ERROR_MORE_DATA) {
+            fprintf(stderr, "hax capability is too long to hold.\n");
+        }
+        fprintf(stderr, "Failed to get Hax capability:%luu\n", err);
+        return -EFAULT;
+    } else {
+        return 0;
+    }
+}
+
+int hax_mod_version(struct hax_state *hax, struct hax_module_version *version)
+{
+    int ret;
+    HANDLE hDevice = hax->fd; /* handle to hax module */
+    DWORD dSize = 0;
+    DWORD err = 0;
+
+    if (hax_invalid_fd(hDevice)) {
+        fprintf(stderr, "Invalid fd for hax device!\n");
+        return -ENODEV;
+    }
+
+    ret = DeviceIoControl(hDevice,
+                          HAX_IOCTL_VERSION,
+                          NULL, 0,
+                          version, sizeof(*version), &dSize,
+                          (LPOVERLAPPED) NULL);
+
+    if (!ret) {
+        err = GetLastError();
+        if (err == ERROR_INSUFFICIENT_BUFFER || err == ERROR_MORE_DATA) {
+            fprintf(stderr, "hax module verion is too long to hold.\n");
+        }
+        fprintf(stderr, "Failed to get Hax module version:%lu\n", err);
+        return -EFAULT;
+    } else {
+        return 0;
+    }
+}
+
+static char *hax_vm_devfs_string(int vm_id)
+{
+    return g_strdup_printf("\\\\.\\hax_vm%02d", vm_id);
+}
+
+static char *hax_vcpu_devfs_string(int vm_id, int vcpu_id)
+{
+    return g_strdup_printf("\\\\.\\hax_vm%02d_vcpu%02d", vm_id, vcpu_id);
+}
+
+int hax_host_create_vm(struct hax_state *hax, int *vmid)
+{
+    int ret;
+    int vm_id = 0;
+    DWORD dSize = 0;
+
+    if (hax_invalid_fd(hax->fd)) {
+        return -EINVAL;
+    }
+
+    if (hax->vm) {
+        return 0;
+    }
+
+    ret = DeviceIoControl(hax->fd,
+                          HAX_IOCTL_CREATE_VM,
+                          NULL, 0, &vm_id, sizeof(vm_id), &dSize,
+                          (LPOVERLAPPED) NULL);
+    if (!ret) {
+        fprintf(stderr, "Failed to create VM. Error code: %lu\n",
+                GetLastError());
+        return -1;
+    }
+    *vmid = vm_id;
+    return 0;
+}
+
+hax_fd hax_host_open_vm(struct hax_state *hax, int vm_id)
+{
+    char *vm_name = NULL;
+    hax_fd hDeviceVM;
+
+    vm_name = hax_vm_devfs_string(vm_id);
+    if (!vm_name) {
+        fprintf(stderr, "Failed to open VM. VM name is null\n");
+        return INVALID_HANDLE_VALUE;
+    }
+
+    hDeviceVM = CreateFile(vm_name,
+                           GENERIC_READ | GENERIC_WRITE,
+                           0, NULL, CREATE_ALWAYS, FILE_ATTRIBUTE_NORMAL, NULL);
+    if (hDeviceVM == INVALID_HANDLE_VALUE) {
+        fprintf(stderr, "Open the vm device error:%s, ec:%lu\n",
+                vm_name, GetLastError());
+    }
+
+    g_free(vm_name);
+    return hDeviceVM;
+}
+
+int hax_notify_qemu_version(hax_fd vm_fd, struct hax_qemu_version *qversion)
+{
+    int ret;
+    DWORD dSize = 0;
+    if (hax_invalid_fd(vm_fd)) {
+        return -EINVAL;
+    }
+    ret = DeviceIoControl(vm_fd,
+                          HAX_VM_IOCTL_NOTIFY_QEMU_VERSION,
+                          qversion, sizeof(struct hax_qemu_version),
+                          NULL, 0, &dSize, (LPOVERLAPPED) NULL);
+    if (!ret) {
+        fprintf(stderr, "Failed to notify qemu API version\n");
+        return -1;
+    }
+    return 0;
+}
+
+int hax_host_create_vcpu(hax_fd vm_fd, int vcpuid)
+{
+    int ret;
+    DWORD dSize = 0;
+
+    ret = DeviceIoControl(vm_fd,
+                          HAX_VM_IOCTL_VCPU_CREATE,
+                          &vcpuid, sizeof(vcpuid), NULL, 0, &dSize,
+                          (LPOVERLAPPED) NULL);
+    if (!ret) {
+        fprintf(stderr, "Failed to create vcpu %x\n", vcpuid);
+        return -1;
+    }
+
+    return 0;
+}
+
+hax_fd hax_host_open_vcpu(int vmid, int vcpuid)
+{
+    char *devfs_path = NULL;
+    hax_fd hDeviceVCPU;
+
+    devfs_path = hax_vcpu_devfs_string(vmid, vcpuid);
+    if (!devfs_path) {
+        fprintf(stderr, "Failed to get the devfs\n");
+        return INVALID_HANDLE_VALUE;
+    }
+
+    hDeviceVCPU = CreateFile(devfs_path,
+                             GENERIC_READ | GENERIC_WRITE,
+                             0, NULL, CREATE_ALWAYS, FILE_ATTRIBUTE_NORMAL,
+                             NULL);
+
+    if (hDeviceVCPU == INVALID_HANDLE_VALUE) {
+        fprintf(stderr, "Failed to open the vcpu devfs\n");
+    }
+    g_free(devfs_path);
+    return hDeviceVCPU;
+}
+
+int hax_host_setup_vcpu_channel(struct hax_vcpu_state *vcpu)
+{
+    hax_fd hDeviceVCPU = vcpu->fd;
+    int ret;
+    struct hax_tunnel_info info;
+    DWORD dSize = 0;
+
+    ret = DeviceIoControl(hDeviceVCPU,
+                          HAX_VCPU_IOCTL_SETUP_TUNNEL,
+                          NULL, 0, &info, sizeof(info), &dSize,
+                          (LPOVERLAPPED) NULL);
+    if (!ret) {
+        fprintf(stderr, "Failed to setup the hax tunnel\n");
+        return -1;
+    }
+
+    if (!valid_hax_tunnel_size(info.size)) {
+        fprintf(stderr, "Invalid hax tunnel size %x\n", info.size);
+        ret = -EINVAL;
+        return ret;
+    }
+    vcpu->tunnel = (struct hax_tunnel *) (intptr_t) (info.va);
+    vcpu->iobuf = (unsigned char *) (intptr_t) (info.io_va);
+    return 0;
+}
+
+int hax_vcpu_run(struct hax_vcpu_state *vcpu)
+{
+    int ret;
+    HANDLE hDeviceVCPU = vcpu->fd;
+    DWORD dSize = 0;
+
+    ret = DeviceIoControl(hDeviceVCPU,
+                          HAX_VCPU_IOCTL_RUN,
+                          NULL, 0, NULL, 0, &dSize, (LPOVERLAPPED) NULL);
+    if (!ret) {
+        return -EFAULT;
+    } else {
+        return 0;
+    }
+}
+
+int hax_sync_fpu(CPUArchState *env, struct fx_layout *fl, int set)
+{
+    int ret;
+    hax_fd fd;
+    HANDLE hDeviceVCPU;
+    DWORD dSize = 0;
+
+    fd = hax_vcpu_get_fd(env);
+    if (hax_invalid_fd(fd)) {
+        return -1;
+    }
+
+    hDeviceVCPU = fd;
+
+    if (set) {
+        ret = DeviceIoControl(hDeviceVCPU,
+                              HAX_VCPU_IOCTL_SET_FPU,
+                              fl, sizeof(*fl), NULL, 0, &dSize,
+                              (LPOVERLAPPED) NULL);
+    } else {
+        ret = DeviceIoControl(hDeviceVCPU,
+                              HAX_VCPU_IOCTL_GET_FPU,
+                              NULL, 0, fl, sizeof(*fl), &dSize,
+                              (LPOVERLAPPED) NULL);
+    }
+    if (!ret) {
+        return -EFAULT;
+    } else {
+        return 0;
+    }
+}
+
+int hax_sync_msr(CPUArchState *env, struct hax_msr_data *msrs, int set)
+{
+    int ret;
+    hax_fd fd;
+    HANDLE hDeviceVCPU;
+    DWORD dSize = 0;
+
+    fd = hax_vcpu_get_fd(env);
+    if (hax_invalid_fd(fd)) {
+        return -1;
+    }
+    hDeviceVCPU = fd;
+
+    if (set) {
+        ret = DeviceIoControl(hDeviceVCPU,
+                              HAX_VCPU_IOCTL_SET_MSRS,
+                              msrs, sizeof(*msrs),
+                              msrs, sizeof(*msrs), &dSize, (LPOVERLAPPED) NULL);
+    } else {
+        ret = DeviceIoControl(hDeviceVCPU,
+                              HAX_VCPU_IOCTL_GET_MSRS,
+                              msrs, sizeof(*msrs),
+                              msrs, sizeof(*msrs), &dSize, (LPOVERLAPPED) NULL);
+    }
+    if (!ret) {
+        return -EFAULT;
+    } else {
+        return 0;
+    }
+}
+
+int hax_sync_vcpu_state(CPUArchState *env, struct vcpu_state_t *state, int set)
+{
+    int ret;
+    hax_fd fd;
+    HANDLE hDeviceVCPU;
+    DWORD dSize;
+
+    fd = hax_vcpu_get_fd(env);
+    if (hax_invalid_fd(fd)) {
+        return -1;
+    }
+
+    hDeviceVCPU = fd;
+
+    if (set) {
+        ret = DeviceIoControl(hDeviceVCPU,
+                              HAX_VCPU_SET_REGS,
+                              state, sizeof(*state),
+                              NULL, 0, &dSize, (LPOVERLAPPED) NULL);
+    } else {
+        ret = DeviceIoControl(hDeviceVCPU,
+                              HAX_VCPU_GET_REGS,
+                              NULL, 0,
+                              state, sizeof(*state), &dSize,
+                              (LPOVERLAPPED) NULL);
+    }
+    if (!ret) {
+        return -EFAULT;
+    } else {
+        return 0;
+    }
+}
+
+int hax_inject_interrupt(CPUArchState *env, int vector)
+{
+    int ret;
+    hax_fd fd;
+    HANDLE hDeviceVCPU;
+    DWORD dSize;
+
+    fd = hax_vcpu_get_fd(env);
+    if (hax_invalid_fd(fd)) {
+        return -1;
+    }
+
+    hDeviceVCPU = fd;
+
+    ret = DeviceIoControl(hDeviceVCPU,
+                          HAX_VCPU_IOCTL_INTERRUPT,
+                          &vector, sizeof(vector), NULL, 0, &dSize,
+                          (LPOVERLAPPED) NULL);
+    if (!ret) {
+        return -EFAULT;
+    } else {
+        return 0;
+    }
+}
+
+static void CALLBACK dummy_apc_func(ULONG_PTR unused)
+{
+}
+
+void hax_kick_vcpu_thread(CPUState *cpu)
+{
+    /*
+     * FIXME: race condition with the exit_request check in
+     * hax_vcpu_hax_exec
+     */
+    cpu->exit_request = 1;
+    if (!qemu_cpu_is_self(cpu)) {
+        if (!QueueUserAPC(dummy_apc_func, cpu->hThread, 0)) {
+            fprintf(stderr, "%s: QueueUserAPC failed with error %lu\n",
+                    __func__, GetLastError());
+            exit(1);
+        }
+    }
+}
diff --git a/target/i386/hax/hax-windows.h b/target/i386/hax/hax-windows.h
new file mode 100644 (file)
index 0000000..a5ce12d
--- /dev/null
@@ -0,0 +1,88 @@
+/*
+ * QEMU HAXM support
+ *
+ * Copyright IBM, Corp. 2008
+ *
+ * Authors:
+ *  Anthony Liguori   <aliguori@us.ibm.com>
+ *
+ * Copyright (c) 2011 Intel Corporation
+ *  Written by:
+ *  Jiang Yunhong<yunhong.jiang@intel.com>
+ *  Xin Xiaohui<xiaohui.xin@intel.com>
+ *  Zhang Xiantao<xiantao.zhang@intel.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#ifndef TARGET_I386_HAX_WINDOWS_H
+#define TARGET_I386_HAX_WINDOWS_H
+
+#include <winioctl.h>
+#include <windef.h>
+
+#include "hax-cpus.h"
+
+#define HAX_INVALID_FD INVALID_HANDLE_VALUE
+
+static inline void hax_mod_close(struct hax_state *hax)
+{
+    CloseHandle(hax->fd);
+}
+
+static inline void hax_close_fd(hax_fd fd)
+{
+    CloseHandle(fd);
+}
+
+static inline int hax_invalid_fd(hax_fd fd)
+{
+    return (fd == INVALID_HANDLE_VALUE);
+}
+
+#define HAX_DEVICE_TYPE 0x4000
+
+#define HAX_IOCTL_VERSION       CTL_CODE(HAX_DEVICE_TYPE, 0x900, \
+                                         METHOD_BUFFERED, FILE_ANY_ACCESS)
+#define HAX_IOCTL_CREATE_VM     CTL_CODE(HAX_DEVICE_TYPE, 0x901, \
+                                         METHOD_BUFFERED, FILE_ANY_ACCESS)
+#define HAX_IOCTL_CAPABILITY    CTL_CODE(HAX_DEVICE_TYPE, 0x910, \
+                                         METHOD_BUFFERED, FILE_ANY_ACCESS)
+
+#define HAX_VM_IOCTL_VCPU_CREATE   CTL_CODE(HAX_DEVICE_TYPE, 0x902, \
+                                            METHOD_BUFFERED, FILE_ANY_ACCESS)
+#define HAX_VM_IOCTL_ALLOC_RAM     CTL_CODE(HAX_DEVICE_TYPE, 0x903, \
+                                            METHOD_BUFFERED, FILE_ANY_ACCESS)
+#define HAX_VM_IOCTL_SET_RAM       CTL_CODE(HAX_DEVICE_TYPE, 0x904, \
+                                            METHOD_BUFFERED, FILE_ANY_ACCESS)
+#define HAX_VM_IOCTL_VCPU_DESTROY  CTL_CODE(HAX_DEVICE_TYPE, 0x905, \
+                                            METHOD_BUFFERED, FILE_ANY_ACCESS)
+#define HAX_VM_IOCTL_ADD_RAMBLOCK  CTL_CODE(HAX_DEVICE_TYPE, 0x913, \
+                                            METHOD_BUFFERED, FILE_ANY_ACCESS)
+
+#define HAX_VCPU_IOCTL_RUN      CTL_CODE(HAX_DEVICE_TYPE, 0x906, \
+                                         METHOD_BUFFERED, FILE_ANY_ACCESS)
+#define HAX_VCPU_IOCTL_SET_MSRS CTL_CODE(HAX_DEVICE_TYPE, 0x907, \
+                                         METHOD_BUFFERED, FILE_ANY_ACCESS)
+#define HAX_VCPU_IOCTL_GET_MSRS CTL_CODE(HAX_DEVICE_TYPE, 0x908, \
+                                         METHOD_BUFFERED, FILE_ANY_ACCESS)
+#define HAX_VCPU_IOCTL_SET_FPU  CTL_CODE(HAX_DEVICE_TYPE, 0x909, \
+                                         METHOD_BUFFERED, FILE_ANY_ACCESS)
+#define HAX_VCPU_IOCTL_GET_FPU  CTL_CODE(HAX_DEVICE_TYPE, 0x90a, \
+                                         METHOD_BUFFERED, FILE_ANY_ACCESS)
+
+#define HAX_VCPU_IOCTL_SETUP_TUNNEL  CTL_CODE(HAX_DEVICE_TYPE, 0x90b, \
+                                              METHOD_BUFFERED, FILE_ANY_ACCESS)
+#define HAX_VCPU_IOCTL_INTERRUPT     CTL_CODE(HAX_DEVICE_TYPE, 0x90c, \
+                                              METHOD_BUFFERED, FILE_ANY_ACCESS)
+#define HAX_VCPU_SET_REGS            CTL_CODE(HAX_DEVICE_TYPE, 0x90d, \
+                                              METHOD_BUFFERED, FILE_ANY_ACCESS)
+#define HAX_VCPU_GET_REGS            CTL_CODE(HAX_DEVICE_TYPE, 0x90e, \
+                                              METHOD_BUFFERED, FILE_ANY_ACCESS)
+
+#define HAX_VM_IOCTL_NOTIFY_QEMU_VERSION CTL_CODE(HAX_DEVICE_TYPE, 0x910, \
+                                                  METHOD_BUFFERED,        \
+                                                  FILE_ANY_ACCESS)
+#endif /* TARGET_I386_HAX_WINDOWS_H */
diff --git a/target/i386/hax/meson.build b/target/i386/hax/meson.build
new file mode 100644 (file)
index 0000000..77ea431
--- /dev/null
@@ -0,0 +1,7 @@
+i386_softmmu_ss.add(when: 'CONFIG_HAX', if_true: files(
+  'hax-all.c',
+  'hax-mem.c',
+  'hax-cpus.c',
+))
+i386_softmmu_ss.add(when: ['CONFIG_HAX', 'CONFIG_POSIX'], if_true: files('hax-posix.c'))
+i386_softmmu_ss.add(when: ['CONFIG_HAX', 'CONFIG_WIN32'], if_true: files('hax-windows.c'))
index 62cd042915179e8e2220b82714571f5522fd0e05..284d52ab8135e8685b7fe5a36b68e246f5df08c4 100644 (file)
@@ -27,15 +27,9 @@ i386_softmmu_ss.add(files(
   'machine.c',
   'monitor.c',
 ))
-i386_softmmu_ss.add(when: 'CONFIG_HAX', if_true: files(
-  'hax-all.c',
-  'hax-mem.c',
-  'hax-cpus.c',
-))
-i386_softmmu_ss.add(when: ['CONFIG_HAX', 'CONFIG_POSIX'], if_true: files('hax-posix.c'))
-i386_softmmu_ss.add(when: ['CONFIG_HAX', 'CONFIG_WIN32'], if_true: files('hax-windows.c'))
 
 subdir('kvm')
+subdir('hax')
 subdir('whpx')
 subdir('hvf')