| From: |
| Christoph Hellwig <hch-AT-lst.de> |
| To: |
| mst-AT-redhat.com, jasowang-AT-redhat.com |
| Subject: |
| automatic IRQ affinity for virtio V3 |
| Date: |
| Sun, 5 Feb 2017 18:15:17 +0100 |
| Message-ID: |
| <20170205171526.6224-1-hch@lst.de> |
| Cc: |
| axboe-AT-kernel.dk, pbonzini-AT-redhat.com, virtualization-AT-lists.linux-foundation.org, linux-block-AT-vger.kernel.org, linux-kernel-AT-vger.kernel.org |
Hi Michael, hi Jason,
This patches applies a few cleanups to the virtio PCI interrupt handling
code, and then converts the virtio PCI code to use the automatic MSI-X
vectors spreading, as well as using the information in virtio-blk
and virtio-scsi to automatically align the blk-mq queues to the MSI-X
vectors.
Changes since V2:
- remove a redundant callback check
- calculate ->msix_vectors correctly
- add a few Reviewed-by: tags for Jason
Changes since V1:
- dropped the patches already merged for 4.10-rc
- new patch to remove struct virtio_pci_vq_info
- improve probe error handling by unwinding step by step
- new patch to convert virtio-scsi
