On Thu, Jul 18, 2013 at 08:32:18PM +0200, Karol Lewandowski wrote: > Agreed, this is why I have been naively thinking that caching wouldn't > be that bad... However, sending all the required information in message > itself is clearly the best solution to this problem.
It is nice to avoid the race for sure but unless I am not using perf record/top correctly or perf itself is broken on ARM, the bottleneck is really not the opening of files. For me glibc's _int_malloc remains at the top. So even if we skip a log of reading of files the journald will still create strings for KEY=VALUE. So maybe we could put key/val separtaely into the iovec array and avoid the extra strdup and then in the journal-file we can avoid the memchr for '='? Anyway I have not looked at journal-file.c much yet... holger _______________________________________________ systemd-devel mailing list systemd-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/systemd-devel