echo 1 > /proc/sys/vm/oom_dump_tasks
que parece ser o máximo que você pode fazer com que o kernel seja exibido em erros de falta de memória.
Enables a system-wide task dump (excluding kernel threads) to be produced when the kernel performs an OOM-killing and includes such information as pid, uid, tgid, vm size, rss, nr_ptes, swapents, oom_score_adj score, and name. This is helpful to determine why the OOM killer was invoked, to identify the rogue task that caused it, and to determine why the OOM killer chose the task it did to kill.
If this is set to zero, this information is suppressed. On very large systems with thousands of tasks it may not be feasible to dump the memory state information for each one. Such systems should not be forced to incur a performance penalty in OOM conditions when the information may not be desired.
If this is set to non-zero, this information is shown whenever the OOM killer actually kills a memory-hogging task.