On Thu, Feb 17, 2022 at 08:27:14AM +0000, Gautam Dawar wrote: > [GD>>] Yes , I agree a complete solution that will support the > emulated virtio device with in_order rx/tx virtqueue functions will > definitely be better but at the same time it will take considerable > amount of time and effort. I also noticed that something similar > (https://patchew.org/QEMU/1533833677-27512-1-git-send-email-i.maxim...@samsung.com/) > was proposed years ago which got dropped for similar reasons and it > has been status quo since then.
Not applying a patch until it's complete is really the only leverage maintainers have to push for complete patches. Otherwise people keep adding half-baked code and hoping that someone else does the rest of the work. > So, unless we know that this work is > already in progress & will be up-streamed soon and if you don’t see > any side-effects of this patch, we can get it integrated first and > then this can be reverted when the complete solution is available. > This would help in benchmarking performance boosts achieved with > IN_ORDER feature. You can just carry the patch in private for benchmarking I guess? -- MST
