On 29/03/2017 23:28, Michael S. Tsirkin wrote:
>> Something like this?
>>
>> diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
>> index 03592c5..73e69ac 100644
>> --- a/hw/virtio/virtio.c
>> +++ b/hw/virtio/virtio.c
>> @@ -176,6 +176,10 @@ err_used:
>> address_space_cache_destroy(&new->desc);
>> err_desc:
>> g_free(new);
>> + atomic_rcu_set(&vq->vring.caches, NULL);
>> + if (old) {
>> + call_rcu(old, virtio_free_region_cache, rcu);
>> + }
>> }
>>
>> /* virt queue functions */
> I would be worried about call_rcu here - this means
> something can hang on to and use the old cache,
> and reset really must act as a sync/flush point.
The flush is done later in virtio_reset. Here it's just reacting
asynchronously to the IOMMU reset.
I'm thinking of adding a global generation count for IOMMU mappings, and
forcing an update when the IOMMU mappings have changed.
Paolo
