On Fri, Oct 09, 2020 at 10:59:45AM +0100, John Keeping wrote: > No, it's not, although I would have saved several days debugging if it > was! I backported the lockdep warning to prove that it caught this > issue. > > The evidence it is possible to see on vanilla 5.4.x is: > > $ trace-cmd report -l > irq/43-e-280 0....2 74.017658: softirq_raise: vec=3 > [action=NET_RX] > > Note the missing "d" where this should be "0d...2" to indicate hardirqs > disabled.
Cool, makes sense.
