On Tue, Nov 05, 2024 at 05:23:46PM +0100, Florian Westphal wrote: > Pablo Neira Ayuso <pablo@xxxxxxxxxxxxx> wrote: > > On Sun, Nov 03, 2024 at 11:26:36AM +0100, Nadia Pinaeva wrote: > > > I would like to provide some more context from the user point of view. > > > I am working on a tool that allows collecting network performance > > > metrics by using conntrack events. > > > Start time of a conntrack entry is used to evaluate seen_reply > > > latency, therefore the sooner it is timestamped, the better the > > > precision is. > > > In particular, when using this tool to compare the performance of the > > > same feature implemented using iptables/nftables/OVS it is crucial > > > to have the entry timestamped earlier to see any difference. > > > > > > I am not sure if current timestamping logic is used for anything, but > > > changing it would definitely help with my use case. > > > I am happy to provide more details, if you have any questions. > > > > The start time will be accurate. However, stop time will not be very > > accurate: the netlink message containing the SEEN_REPLY status flag > > can sit in the socket queue for some quite time until the userspace > > software has a chance to receive and parse it. > > > > @Florian: Would you explore instead to extend the nf_conntrack_ecache > > infrastructure to allow to provide timestamps for netlink events? This > > can be enabled via toggle. That would allow to have a more accurate > > delta between two events messages. > > Simply using current time in ctnetlink won't help, the NEW event comes > after confirm. > > It will help for SEEN_REPLY. But I don't see how it will avoid this > patch. Not current time from ctnetlink, but use the ecache extension to store the timestamp when the conntrack is allocated, ecache is already initialized from init_conntrack() path.
