Pavel Machek <pavel@xxxxxxx> wrote: > This goes from "kill kernel with backtrace" to "silently return > failure". Should WARN_ON() be preserved here? No need. The error would propagate back to userspace via nfnetlink. So it would be 'running this command fails with error x' vs. 'running this command makes kernel crash'. That being said, I did not observe this BUG from triggering, ever. I only removed it because I did not see any reason for it.
