On Fri, 2006-06-16 at 16:28 -0500, Gian Paolo Mureddu wrote: > Lee Revell escribió: > > I would recommend you attack one issue at a time. But yes, that will > > work. > > > > It might not even be the case anymore that reiser3 causes latency > > problems - the -rt kernel and the introduction of the preemptible BKL > > probably improved the situation. It has been a long time since I tested > > it. > > > > Lee > > > > > > > Hmm... it might be interesting to have a test case of this... > It should be trivial to make one. The -rt kernel is of course unaffected, as all filesystem operations are offloaded to IRQ threads and softirqd. For testing the mainline kernel, use 2.6.16 with this patch: http://people.redhat.com/mingo/latency-tracing-patches/latency-tracing-v2.6.16.patch Generate a lot of disk activity (dbench, tiobench, etc) and check /proc/latency_trace for reiserfs related entries. Lee
