skbuf cache

From: Steve Smoot (smoot@bmrc.berkeley.edu)
Date: Wed Jan 29 2003 - 20:47:41 EST


What are the size limits on the skbuf cache? I'm trying to debug:
Jan 27 20:26:45 mybox kernel: eth1: Memory squeeze,deferring packet.
Jan 27 20:26:45 mybox kernel: eth1: NULL pointer encountered in Rx ring, skipping
Jan 27 20:26:45 mybox last message repeated 553 times
(later the kernel panics on a NULL pointer)

The message comes from the sis 900 device driver, chasing up, when the
alloc of a new skbuf fails. But I've failed to find where the size of this
cache comes from, it seems to fall into the generic kernel slab allocator,
(kmem cache) which if I understand it correctly is used for generic kernel
objects of a certain type. So, what controls the size of this pool, and is
there any reason to think that increasing it would either 1) have bad side
effects, or 2) not help.

(And if I'm really forced to guess, given the lack of checking in the code
I've scaned, I'm thinking this isn't really bounded except by physmem, and
thus more memory in the machine is needed - and I should add a vmstat call
to my logging info to check if that's what is up. That's cheap, I'll do
that.)

If it's any help, this all came about when trying to process > 90 Mbps of
multicast traffic through the NIC (100 Mbps switched net).

-s

-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html



This archive was generated by hypermail 2b29 : Fri Jan 31 2003 - 22:00:02 EST