- write stress tests - take ideas from brad.chen's testing plan - integrate into "crashme" framework - arrange automated runs on available RH hardware
Is this bug still relevant given the grind *probes have gone through in the past 3 years?
> Is this bug still relevant given the grind *probes have gone through in the past > 3 years? While the code has certainly improved, we still don't have broad tests that try to confirm the sufficiency of the kernel-side kprobes blacklist. (Such broad systemtap-level tests, corresponding to broad wildcard probes have problems of their own.)
Masami's 'perf-probe' based tool is probably the best we'll have. Perhaps someday the kernel will pass that test. http://lkml.org/lkml/2010/2/15/197