Michal Hocko wrote: > > If you can rebuild your kernel, calling dump_tasks() in mm/oom_kill.c when > > you hit warn_alloc() warnings might help. > > I do not really see how this would help much. If anything watching for > /proc/vmstat counters would tell us much more. Under memory pressure, read()/write() syscalls might involve significant delay (including reading from /sys/kernel/debug/tracing/trace_pipe and writing to a log file). Unless all problems are contained in a cgroup (which means that administrators can diagnose from outside of that cgroup), it is silly to try to read memory related information of a stalling system using userspace interface. Therefore, automatic printk() is helpful than trying to start "cat /proc/vmstat" from a shell. I wish there is a kernel function which does "cat /proc/some_file" and/or "cat /sys/kernel/debug/tracing/trace_pipe" and sends the output to printk() so that such actions will not involve significant delay under memory pressure. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>
