On 16/03/2016 19:18, Stefan Hajnoczi wrote: > Looks good overall. I'm a little nervous about merging it for QEMU 2.6 > but the block job, NBD, and data plane tests should give it a good > workout.
Apart from QEMU nearing hard freeze, I totally understand not wanting to commit to merging part 1 of n where n will probably be a dozen or so. I'm open to experimenting with different models for handling long-term contributions. For example, each part will probably have an uncontroversial and generally useful prefix---for example patches 1-4 in this case, or the change to a single linux-aio context per iothread. You could merge those only, and for the rest, I will maintain myself a branch with R-b from maintainers. Master will be periodically merged into it, but not too frequently---it could be only after each part is accepted, or when there is some important bugfix to catch. Once the whole multiqueue thing gets somewhere I would send you a pull request with the entire feature, which would consist of say 200 patches all with a Reviewed-by already. This is just a possibility; if you have any other idea, I'd be happy to follow it. Paolo
