Hi Michael, Understood. Thanks.
Best Regards, Rick Zhong -----邮件原件----- 发件人: Michael S. Tsirkin <m...@redhat.com> 发送时间: 2024年2月13日 17:53 收件人: Rick Zhong <zhaoyong.zh...@nephogine.com> 抄送: Eugenio Perez Martin <epere...@redhat.com>; qemu-devel@nongnu.org; Jason Wang <jasow...@redhat.com>; Peter Xu <pet...@redhat.com>; Guo Zhi <qtxuning1...@sjtu.edu.cn>; Xinying Yu <xinying...@nephogine.com>; Wentao Jia <wentao....@nephogine.com>; Shujing Dong <shujing.d...@nephogine.com>; Kyle Xu <zhenbing...@nephogine.com> 主题: Re: 回复: FW: [PATCH] vhost-user: add VIRTIO_F_IN_ORDER and VIRTIO_F_NOTIFICATION_DATA feature On Fri, Feb 02, 2024 at 10:27:33AM +0000, Rick Zhong wrote: > Hi Eugenio and Michael, > > Let me make it more clear about the target for this patch. Currently Corigine > is developing the vDPA features on NIC which are based on the QEMU > vhost-vdpa/vhost-user backend. These two virtio features are helpful in data > plane performance. > > In my understanding, these two virtio features are defined as part of the > basic facilities of a common virtio device, which means they can be utilized > by virtio-net, virtio-blk, virtio-fs... whatever backend. To implement, it is > beyond the team's knowledge to handle these for all kinds of backends. So I'd > prefer to set them off by default and raise an warning for other type of > backends, except vhost-vdpa/vhost-user. > > Best Regards, > Rick Zhong Yes, you should set it off by default. No, just skipping implementation won't cut it. It is understandable that you just want your own use-case addressed and it is annoying to get roped in to do some work on qemu. However, such is the cost of doing this business. If instead we add hacks like the warning you mention then the codebase quickly becomes a mess of special cases. If you need this feature, you have to make it nicely orthogonal and palatable to everyone. It is really not a lot of coding work, mostly testing. -- MST
