VCPU with vcpu->vcpu_id has highest probability of being stored in
kvm->vcpus[vcpu->vcpu_id]. Other common case, sparse sequential
vcpu_id, is more likely to find a match downwards from vcpu->vcpu_id.
Random distribution does not matter so we first search slots
[vcpu->vcpu_id..0] and then slots (vcpu->vcpu_id..kvm->online_vcpus).
If we value cycles over memory, a direct map between vcpu_id and
vcpu->vcpu_id would be better.
(Like kvm_for_each_vcpu, the code avoid the kvm->lock by presuming that
kvm->online_vcpus doesn't shrink and that the vcpu pointer is set up
before incrementing. kvm_free_vcpus() breaks that presumption, but vm
is destroyed only after the fd has been released.)
Signed-off-by: Radim Krčmář <rkrcmar@xxxxxxxxxx>
---
virt/kvm/kvm_main.c | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index 024428b64812..7d532591d5af 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -2650,7 +2650,7 @@ int kvm_vm_ioctl_user_exit(struct kvm *kvm, struct kvm_user_exit *info)
* KVM_CREATE_VCPU, where we cast from unsigned long.
*/
int vcpu_id = info->vcpu_id;
- int idx;
+ int idx, first;
struct kvm_vcpu *vcpu;
const struct kvm_user_exit valid = {.vcpu_id = info->vcpu_id};
@@ -2659,7 +2659,11 @@ int kvm_vm_ioctl_user_exit(struct kvm *kvm, struct kvm_user_exit *info)
if (memcmp(info, &valid, sizeof(valid)))
return -EINVAL;
- kvm_for_each_vcpu(idx, vcpu, kvm)
+ for (idx = first = min(vcpu_id, atomic_read(&kvm->online_vcpus) - 1);
+ idx >= 0 ? (vcpu = kvm_get_vcpu(kvm, idx)) != NULL
+ : ++first < atomic_read(&kvm->online_vcpus) &&
+ (vcpu = kvm_get_vcpu(kvm, first)) != NULL;
+ idx--)
if (vcpu->vcpu_id == vcpu_id) {
kvm_make_request(KVM_REQ_EXIT, vcpu);
kvm_vcpu_kick(vcpu);
--
2.5.0
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
