提交 8db4936b 编写于 作者: P Paolo Bonzini

kvm-all: kvm_irqchip_create is not expected to fail

KVM_CREATE_IRQCHIP should never fail, and so should its userspace
wrapper kvm_irqchip_create.  The function does not do anything
if the irqchip capability is not available, as is the case for PPC.

With this patch, kvm_arch_init can allocate memory and it will not
be leaked.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
上级 38bfe691
......@@ -1345,27 +1345,31 @@ int kvm_irqchip_remove_irqfd_notifier(KVMState *s, EventNotifier *n, int virq)
false);
}
static int kvm_irqchip_create(MachineState *machine, KVMState *s)
static void kvm_irqchip_create(MachineState *machine, KVMState *s)
{
int ret;
if (!machine_kernel_irqchip_allowed(machine) ||
(!kvm_check_extension(s, KVM_CAP_IRQCHIP) &&
(kvm_vm_enable_cap(s, KVM_CAP_S390_IRQCHIP, 0) < 0))) {
return 0;
if (kvm_check_extension(s, KVM_CAP_IRQCHIP)) {
;
} else if (kvm_check_extension(s, KVM_CAP_S390_IRQCHIP)) {
ret = kvm_vm_enable_cap(s, KVM_CAP_S390_IRQCHIP, 0);
if (ret < 0) {
fprintf(stderr, "Enable kernel irqchip failed: %s\n", strerror(-ret));
exit(1);
}
} else {
return;
}
/* First probe and see if there's a arch-specific hook to create the
* in-kernel irqchip for us */
ret = kvm_arch_irqchip_create(s);
if (ret < 0) {
return ret;
} else if (ret == 0) {
if (ret == 0) {
ret = kvm_vm_ioctl(s, KVM_CREATE_IRQCHIP);
if (ret < 0) {
fprintf(stderr, "Create kernel irqchip failed\n");
return ret;
}
}
if (ret < 0) {
fprintf(stderr, "Create kernel irqchip failed: %s\n", strerror(-ret));
exit(1);
}
kvm_kernel_irqchip = true;
......@@ -1376,8 +1380,6 @@ static int kvm_irqchip_create(MachineState *machine, KVMState *s)
kvm_halt_in_kernel_allowed = true;
kvm_init_irq_routing(s);
return 0;
}
/* Find number of supported CPUs using the recommended
......@@ -1594,9 +1596,8 @@ static int kvm_init(MachineState *ms)
goto err;
}
ret = kvm_irqchip_create(ms, s);
if (ret < 0) {
goto err;
if (machine_kernel_irqchip_allowed(ms)) {
kvm_irqchip_create(ms, s);
}
kvm_state = s;
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册