On 08/09/19 at 06:33am, Jun Wang wrote:
The root cause is likely an infinite loop or corrupted data somewhere, but not too many vmcore dmessages. The fix of limiting vmcore-dmesg.txt size is mainly to defend kdump servers against the existing problem, as well as other future problems. It doesn’t prevent us from further investigating the root cause of the problem that is hard to reproduce either.
It is easy to add some assert in vmcore-dmesg.c to limit the write fd size and assert, so that do not add extra things in distribution.
Also for the size 100M is not necessarily a good value, it would be good to keep same max as kernel is using. Even if it is corrupted thing, it is still not good to have a 100M limitation.
Thanks Dave