>>>>> "Alan" == Alan Curry <pacman-kernel@cqc.com> writes:
> Why do you assume ENOMEM results in death? The whole point is that it doesn't
> need to. Critical processes like init can sleep and retry after ENOMEM.
And it's exactly in those cases that noovercommit bites you: So let's say
syslogd is written carefully not to die when receiving ENOMEM, but to wait a
while and try again (in the meant time, syslog messages coming from the network
need to be dropped, obviously since we have no memory to store them into).
Now syslogd hits a bug of some kind, goes crazy and gets stuck in loop
eating more memory.
What happens then ? Every app starts getting ENOMEM, so they all either die
or keep asking for more ==> deadlock.
Stefan
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Tue Mar 07 2000 - 21:00:22 EST