Re: [libvirt] [PATCH RESEND] qemu: Process RDMA GID state change event

On Mon, Nov 05, 2018 at 05:42:10PM +0100, Michal Privoznik wrote: > [There is no need to resend your patch just to put yourself onto CC > list. The review policy is always "reply to all"] Hi Michal, Actually i posted the "RESEND" as my original mail was not shown up in https://www.redhat.com/archives/libvir-list so I followed the recommendations (register to mailing list etc) and re-post.

> > On 11/05/2018 11:50 AM, Yuval Shaia wrote: >> This event is emitted on the monitor when a GID table in pvrdma device >> is modified and the change needs to be propagate to the backend RDMA >> device's GID table. > > Not yet, your qemu patches are not merged yet ;-) I will provide review > anyway, but even if this patch was good as is we couldn't merge it > unless it's merged into qemu repo first. We have our history with that. Yeah, not yet merged but assuming both should be merged at the same time due to dependency i had to start the reviews in parallel. The qemu patches will probably merged in the in the next few weeks (v3.2) so I'd like to do the best to be ready with this one too.

Anyways, appreciate much your review i took all you comments and will post v2 soon. Please check my answers/questions below. > >>

> > Frankly, I don't know about RDMA that much, but the qemu interface looks > weird to me. If it wants to send an IP address to us it's doing that in > a cryptic way. Can you elaborate more on that? Actually the RDMA device does not send IP addresses, from the RDMA device perspective those uint64_ts (subnet_prefix and interface_id) are called GID and the only way to add new GID to RDMA device is by adding the "equivalent" IP(v6) address to the Ethernet function of the device. Hope it make sense.