Actually, I just bought one vm from public cloud provider and run into this problem. after reading code and compare pci device info, I reproduce this scenario. Since common users cannot change msi vector numbers, so I suggest blk_mq_map_queues to be more friendly. blk_mq_map_queues may be the last choice. At 2019-03-27 16:16:19, "Christoph Hellwig" <hch@xxxxxx> wrote: >On Tue, Mar 26, 2019 at 03:55:10PM +0800, luferry wrote: >> >> >> >> At 2019-03-26 15:39:54, "Christoph Hellwig" <hch@xxxxxx> wrote: >> >Why isn't this using the automatic PCI-level affinity assignment to >> >start with? >> >> When enable virtio-blk with multi queues but with only 2 msix-vector. >> vp_dev->per_vq_vectors will be false, vp_get_vq_affintity will return NULL directly >> so blk_mq_virtio_map_queues will fallback to blk_mq_map_queues. > >What is the point of the multiqueue mode if you don't have enough >(virtual) MSI-X vectors?
