On Mon, 07 Mar 2022 19:07:37 +0100 Toke Høiland-Jørgensen wrote: > Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx> writes: > > > On 2022-03-07 17:50:04 [+0100], Toke Høiland-Jørgensen wrote: > >> > >> Right, looking at the code again, the id is only assigned in the path > >> that doesn't return NULL from __xdp_reg_mem_model(). > >> > >> Given that the trace points were put in specifically to be able to pair > >> connect/disconnect using the IDs, I don't think there's any use to > >> creating the events if there's no ID, so I think we should fix it by > >> skipping the trace event entirely if xdp_alloc is NULL. > > > > This sounds like a reasonable explanation. If nobody disagrees then I > > post a new patch tomorrow and try to recycle some of what you wrote :) > > SGTM :) Was the patch posted? This seems to be a 5.17 thing, so it'd be really really good if the fix was in net by tomorrow morning! :S
