* Alexandre Montplaisir ([email protected]) wrote: > On 11-03-16 10:34 AM, Mathieu Desnoyers wrote: > > * Harald Gustafsson ([email protected]) wrote: > >>> Sounds good. Which version of lttng are you using ? > >> I have tried this both on my x86 machine using the ubuntu ppa package > >> which uses kernel 2.6.35-24 and to me unknown lttng version. But I > >> also tried this on a ARM development board with kernel 2.6.34 and > >> lttng version 0.221 without success. > >> > >>> In your kernel config, under the "Kernel Hacking" menu, did you enable > >>> the options related to lock dependency debug, and lock proving and so > >>> on ? You should ensure that CONFIG_LOCKDEP is selected in your kernel > >>> config file. > >> In the ubuntu-ppa lttng package only LOCKDEP_SUPPORT is enabled, not > >> LOCKDEP and not LOCKDEP_DEBUG, neither DEBUG_LOCK_ALLOC, > >> PROVE_LOCKING, LOCK_STAT or any of the > >> INLINE_{READ|WRITE|SPIN}_LOCK_*. The same configuration have been used > >> on the ARM board. > >> > >> Hence, I should enable LOCKDEP, LOCKDEP_DEBUG, DEBUG_LOCK_ALLOC and > >> PROVE_LOCKING. > > PROVE_LOCKING might not be needed (and is quite heavyweight), but yes, > > enabling at the very least LOCKDEP is needed. > > > > Thanks, > > > > Mathieu > > > > Indeed, the packaged kernel is built using all the default LTTng options. > > Mathieu, is there a big overhead in enabling CONFIG_LOCKDEP? If not, > should it (or any other non-default options) be enabled in the package?
Yes, there is a big overhead to it. What I should do is to do not build the lockdep tracing module if CONFIG_LOCKDEP is not enabled. Mathieu > > -- > > Alexandre Montplaisir > DORSAL lab, > École Polytechnique de Montréal > -- Mathieu Desnoyers Operating System Efficiency R&D Consultant EfficiOS Inc. http://www.efficios.com _______________________________________________ ltt-dev mailing list [email protected] http://lists.casi.polymtl.ca/cgi-bin/mailman/listinfo/ltt-dev
