On Mon, 4 Nov 2024 at 10:48, Dave Hansen <dave.hansen@xxxxxxxxx> wrote: > > Next time, could we please not run the patch subjects through the > marketing department before sending them out? ;) > > This might be "zero overhead" for the BPF program, but it adds new code > (and overhead) to the x86 page fault code. The new check is *not* zero > overhead. In patch 1, the kernel is about to panic after entering the branch where the check is inserted. Branch is marked "unlikely". Surely this is not the fast path of the fault handler? The patch subject is about zero overhead of PROBE_MEM.
