[ovs-dev] [PATCH v6 00/13] Support multi-segment mbufs

Ben Pfaff blp at ovn.org
Tue Jul 24 17:43:56 UTC 2018


On Tue, Jul 24, 2018 at 07:55:39PM +0300, Ilya Maximets wrote:
> Hi.
> Just wanted to add some comments for the use-cases and testing methodology.
> See inline.
> 
> And I'm actually not sure if there any profit from this patch set?
> It looks like an internal mbuf handling rework that only degrades
> the performance and complicates the code.
> 
> Please, don't consider me as merge blocker. I just want to understand
> why you think we need this 1200 LOCs?
> 
> ---
> About 'resize()' related discussion:
> Maybe it's worth to allow dp_packet APIs to return different dp_packet.
> In this case we'll be able to just clone the packet to malloced memory
> and resize in cases of not enough headroom available.
> Like:
> 	packet = eth_push_vlan(packet, vlan->vlan_tpid, vlan->vlan_tci);
> or
> 	eth_push_vlan(&packet, vlan->vlan_tpid, vlan->vlan_tci);
> 
> This will have a little performance penalty in compare with data shifting
> inside the mbuf, but will be much more elegant, and will allow to eliminate
> all the OVS_NOT_REACHED cases.
> 
> 
> Best regards, Ilya Maximets.

Hmm, this is the second person from whom I've heard serious misgivings
about this patch series.  Tiago, Ian, would you like to respond?  I'm a
little nervous about merging this patch series, especially relatively
late before branching, given that some people have technical objections
to it.

Thanks,

Ben.


More information about the dev mailing list