On Tue, Nov 04, 2014 at 04:52:31PM -0200, Marcelo Ricardo Leitner wrote: > >That will catch possible miscalculations too, I don't like we're > >polluting this with many WARN_ONCE, but I don't see any better way to > >catch this size miscalculation bugs, so ahead add it. > > > >BTW, we should also signal the userspace when we fail to build the > >message via: > > > >nfnetlink_set_err(net, 0, group, -ENOBUFS); > > > >so it knows that we're losing log messages for whatever reason. > >Basically, userspace hits -ENOBUFS when calling recv(), which means > >netlink is losing messages. I don't think we really need the > >statistics. > > Cool. Then we probably don't need the WARN_ON unless we really want > those numbers, or even so. As we will be calling nfnetlink_set_err, > we have other ways of debugging. Wouldn't be as ready as a WARN_ONCE > in there, yes.. but a systemtap script can easily get a hold in > there. > > So no WARN_ONCE, just the nfnetlink_set_err() call? Like this? The ENOBUFS notice won't be enough. This can be triggered if the rate of log message is too high, or the userspace process is too slow to empty the socket queue. I think we need both WARN_ONCE (tells us that miscalculation is happening, ie. we have a bug) and the nfnetlink_set_err (tells userspace it's losing log messages, so logging becomes unreliable/approximative). -- To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
