On Wed, 13 Dec 2017 12:16:03 +0000 Mark Brown <broonie@xxxxxxxxxx> wrote: > On Wed, Dec 13, 2017 at 12:04:26PM +0100, Alexandre Belloni wrote: > > > Also, I think we could try having only the time64_t in the ring buffer. > > Maybe I'm wrong but I think tools reading that buffer can do the > > conversion themselves. Maybe I don't understand correctly how > > tracepoints work and this doesn't make sense, tell me. > > Tools reading the buffer can do the conversion themselves but it's also > useful for users to just view the log directly via tracing/trace > sometimes. OTOH the whole point is to be low overhead so... Ideally, we want TP_fast_assign() do as little as possible, and put all the work in TP_printk() as the printk happens on read of the trace buffer and TP_fast_assign() happens at the tracepoint invocation. -- Steve
