traps.c

xen虚拟机源代码安装包

}

void __devinit subarch_percpu_traps_init(void)
{
    struct tss_struct *tss = &doublefault_tss;
    asmlinkage int hypercall(void);

    if ( smp_processor_id() != 0 )
        return;

    /* The hypercall entry vector is only accessible from ring 1. */
    _set_gate(idt_table+HYPERCALL_VECTOR, 14, 1, &hypercall);

    /*
     * Make a separate task for double faults. This will get us debug output if
     * we blow the kernel stack.
     */
    memset(tss, 0, sizeof(*tss));
    tss->ds     = __HYPERVISOR_DS;
    tss->es     = __HYPERVISOR_DS;
    tss->ss     = __HYPERVISOR_DS;
    tss->esp    = (unsigned long)&doublefault_stack[DOUBLEFAULT_STACK_SIZE];
    tss->__cr3  = __pa(idle_pg_table);
    tss->cs     = __HYPERVISOR_CS;
    tss->eip    = (unsigned long)do_double_fault;
    tss->eflags = 2;
    tss->bitmap = IOBMP_INVALID_OFFSET;
    _set_tssldt_desc(
        gdt_table + __DOUBLEFAULT_TSS_ENTRY - FIRST_RESERVED_GDT_ENTRY,
        (unsigned long)tss, 235, 9);

    set_task_gate(TRAP_double_fault, __DOUBLEFAULT_TSS_ENTRY<<3);
}

void init_int80_direct_trap(struct vcpu *v)
{
    struct trap_info *ti = &v->arch.guest_context.trap_ctxt[0x80];

    /*
     * We can't virtualise interrupt gates, as there's no way to get
     * the CPU to automatically clear the events_mask variable. Also we
     * must ensure that the CS is safe to poke into an interrupt gate.
     *
     * When running with supervisor_mode_kernel enabled a direct trap
     * to the guest OS cannot be used because the INT instruction will
     * switch to the Xen stack and we need to swap back to the guest
     * kernel stack before passing control to the system call entry point.
     */
    if ( TI_GET_IF(ti) || !guest_gate_selector_okay(v->domain, ti->cs) ||
         supervisor_mode_kernel )
    {
        v->arch.int80_desc.a = v->arch.int80_desc.b = 0;
        return;
    }

    v->arch.int80_desc.a = (ti->cs << 16) | (ti->address & 0xffff);
    v->arch.int80_desc.b =
        (ti->address & 0xffff0000) | 0x8f00 | ((TI_GET_DPL(ti) & 3) << 13);

    if ( v == current )
        set_int80_direct_trap(v);
}

#ifdef CONFIG_X86_SUPERVISOR_MODE_KERNEL
static void do_update_sysenter(void *info)
{
    xen_callback_t *address = info;

    wrmsr(MSR_IA32_SYSENTER_CS, address->cs, 0);
    wrmsr(MSR_IA32_SYSENTER_EIP, address->eip, 0);
}
#endif

static long register_guest_callback(struct callback_register *reg)
{
    long ret = 0;
    struct vcpu *v = current;

    fixup_guest_code_selector(v->domain, reg->address.cs);

    switch ( reg->type )
    {
    case CALLBACKTYPE_event:
        v->arch.guest_context.event_callback_cs     = reg->address.cs;
        v->arch.guest_context.event_callback_eip    = reg->address.eip;
        break;

    case CALLBACKTYPE_failsafe:
        v->arch.guest_context.failsafe_callback_cs  = reg->address.cs;
        v->arch.guest_context.failsafe_callback_eip = reg->address.eip;
        if ( reg->flags & CALLBACKF_mask_events )
            set_bit(_VGCF_failsafe_disables_events,
                    &v->arch.guest_context.flags);
        else
            clear_bit(_VGCF_failsafe_disables_events,
                      &v->arch.guest_context.flags);
        break;

#ifdef CONFIG_X86_SUPERVISOR_MODE_KERNEL
    case CALLBACKTYPE_sysenter_deprecated:
        if ( !cpu_has_sep )
            ret = -EINVAL;
        else if ( on_each_cpu(do_update_sysenter, &reg->address, 1, 1) != 0 )
            ret = -EIO;
        break;

    case CALLBACKTYPE_sysenter:
        if ( !cpu_has_sep )
            ret = -EINVAL;
        else
            do_update_sysenter(&reg->address);
        break;
#endif

    case CALLBACKTYPE_nmi:
        ret = register_guest_nmi_callback(reg->address.eip);
        break;

    default:
        ret = -ENOSYS;
        break;
    }

    return ret;
}

static long unregister_guest_callback(struct callback_unregister *unreg)
{
    long ret;

    switch ( unreg->type )
    {
    case CALLBACKTYPE_event:
    case CALLBACKTYPE_failsafe:
#ifdef CONFIG_X86_SUPERVISOR_MODE_KERNEL
    case CALLBACKTYPE_sysenter_deprecated:
    case CALLBACKTYPE_sysenter:
#endif
        ret = -EINVAL;
        break;

    case CALLBACKTYPE_nmi:
        ret = unregister_guest_nmi_callback();
        break;

    default:
        ret = -ENOSYS;
        break;
    }

    return ret;
}


long do_callback_op(int cmd, XEN_GUEST_HANDLE(const_void) arg)
{
    long ret;

    switch ( cmd )
    {
    case CALLBACKOP_register:
    {
        struct callback_register reg;

        ret = -EFAULT;
        if ( copy_from_guest(&reg, arg, 1) )
            break;

        ret = register_guest_callback(&reg);
    }
    break;

    case CALLBACKOP_unregister:
    {
        struct callback_unregister unreg;

        ret = -EFAULT;
        if ( copy_from_guest(&unreg, arg, 1) )
            break;

        ret = unregister_guest_callback(&unreg);
    }
    break;

    default:
        ret = -ENOSYS;
        break;
    }

    return ret;
}

long do_set_callbacks(unsigned long event_selector,
                      unsigned long event_address,
                      unsigned long failsafe_selector,
                      unsigned long failsafe_address)
{
    struct callback_register event = {
        .type = CALLBACKTYPE_event,
        .address = { event_selector, event_address },
    };
    struct callback_register failsafe = {
        .type = CALLBACKTYPE_failsafe,
        .address = { failsafe_selector, failsafe_address },
    };

    register_guest_callback(&event);
    register_guest_callback(&failsafe);

    return 0;
}

static void hypercall_page_initialise_ring0_kernel(void *hypercall_page)
{
    extern asmlinkage int hypercall(void);
    char *p;
    int i;

    /* Fill in all the transfer points with template machine code. */

    for ( i = 0; i < (PAGE_SIZE / 32); i++ )
    {
        p = (char *)(hypercall_page + (i * 32));

        *(u8  *)(p+ 0) = 0x9c;      /* pushf */
        *(u8  *)(p+ 1) = 0xfa;      /* cli */
        *(u8  *)(p+ 2) = 0xb8;      /* mov $<i>,%eax */
        *(u32 *)(p+ 3) = i;
        *(u8  *)(p+ 7) = 0x9a;      /* lcall $__HYPERVISOR_CS,&hypercall */
        *(u32 *)(p+ 8) = (u32)&hypercall;
        *(u16 *)(p+12) = (u16)__HYPERVISOR_CS;
        *(u8  *)(p+14) = 0xc3;      /* ret */
    }

    /*
     * HYPERVISOR_iret is special because it doesn't return and expects a
     * special stack frame. Guests jump at this transfer point instead of
     * calling it.
     */
    p = (char *)(hypercall_page + (__HYPERVISOR_iret * 32));
    *(u8  *)(p+ 0) = 0x50;      /* push %eax */
    *(u8  *)(p+ 1) = 0x9c;      /* pushf */
    *(u8  *)(p+ 2) = 0xfa;      /* cli */
    *(u8  *)(p+ 3) = 0xb8;      /* mov $<i>,%eax */
    *(u32 *)(p+ 4) = __HYPERVISOR_iret;
    *(u8  *)(p+ 8) = 0x9a;      /* lcall $__HYPERVISOR_CS,&hypercall */
    *(u32 *)(p+ 9) = (u32)&hypercall;
    *(u16 *)(p+13) = (u16)__HYPERVISOR_CS;
}

static void hypercall_page_initialise_ring1_kernel(void *hypercall_page)
{
    char *p;
    int i;

    /* Fill in all the transfer points with template machine code. */

    for ( i = 0; i < (PAGE_SIZE / 32); i++ )
    {
        p = (char *)(hypercall_page + (i * 32));
        *(u8  *)(p+ 0) = 0xb8;    /* mov  $<i>,%eax */
        *(u32 *)(p+ 1) = i;
        *(u16 *)(p+ 5) = 0x82cd;  /* int  $0x82 */
        *(u8  *)(p+ 7) = 0xc3;    /* ret */
    }

    /*
     * HYPERVISOR_iret is special because it doesn't return and expects a 
     * special stack frame. Guests jump at this transfer point instead of 
     * calling it.
     */
    p = (char *)(hypercall_page + (__HYPERVISOR_iret * 32));
    *(u8  *)(p+ 0) = 0x50;    /* push %eax */
    *(u8  *)(p+ 1) = 0xb8;    /* mov  $__HYPERVISOR_iret,%eax */
    *(u32 *)(p+ 2) = __HYPERVISOR_iret;
    *(u16 *)(p+ 6) = 0x82cd;  /* int  $0x82 */
}

void hypercall_page_initialise(struct domain *d, void *hypercall_page)
{
    memset(hypercall_page, 0xCC, PAGE_SIZE);
    if ( is_hvm_domain(d) )
        hvm_hypercall_page_initialise(d, hypercall_page);
    else if ( supervisor_mode_kernel )
        hypercall_page_initialise_ring0_kernel(hypercall_page);
    else
        hypercall_page_initialise_ring1_kernel(hypercall_page);
}

/*
 * Local variables:
 * mode: C
 * c-set-style: "BSD"
 * c-basic-offset: 4
 * tab-width: 4
 * indent-tabs-mode: nil
 * End:
 */