I'm having a strange problem with one of my machines... basically,
AppleTalk traffic causes the skbuff_head_cache size reported in
/proc/slabinfo to expand, to the point where it will consume all
available memory and crash the machine after a few days. (it's a
16MB machine)
This started happening when I upgraded from 2.1.119 (after ~60 days of
uptime!) to 2.1.130, though that's probably just coincidence, since I
also started using afpd much more heavily after that point.
I'm currently using 2.1.131ac8, which doesn't appear to fix the problem.
The machine looks like this:
eth0 - outside world (no AppleTalk) (DE450, de4x5.o)
eth1 - internal network (AppleTalk, TCP/IP) (NE2000, ne.o)
lt0 - internal network (AppleTalk, AppleTalkIP) (LTPC, ltpc.o)
This is with Netatalk 1.4b5+asun2.1 as well as 1.4b5+asun2.0b8.
The machine is also doing IP Masquerading and some extra firewalling
stuff, plus a caching-only DNS.
This only happens when Netatalk is up, and there is AppleTalk traffic
on the network - when someone accesses the AFP server, it grows steadily
for a while (by about 10-15 buffers each time), then levels off at a
larger size.
Curiously enough, this doesn't seem to be happening on my machine at home,
which is similarly configured (but LocalTalk-only). It doesn't matter how
much I access the AFP server, skbuff_head_cache never grows.
Can anyone suggest where I should start looking for the memory leak, and
how (aside from simply reading through all the relevant code)?
Thanks.
--
5.15: Is C an acronym?
A: Yes, it stands for "C". It's another of those funky recursive
acronyms.
(From the "comp.lang.c Infrequently Asked Questions list")
-
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/