Hi, Eiichi-san re-discovered the bug earlier found by He Zhe which we've failed to fix due to getting distracted by discussing how to untangle entry_64.S.
These 3 patches are basically a completion of the initial approach I suggested in that earlier thread: https://lkml.kernel.org/r/20190320221534.165ab...@oasis.local.home Since v1: - idtentry_part 'cleanup' - extra sanity check and comment - read_cr2=1 for do_double_fault - #BP vs IST cleanup - IDTENTRYx() C wrapper The thing boots on x86_64 with lockdep on, survives function-trace, selftests/x86 and perf-test.