Does the following code from networking/ping.c (appears in two places)
serve any useful purpose?
/* set recv buf for broadcast pings */
sockopt = 48 * 1024; /* explain why 48k? */
setsockopt(pingsock, SOL_SOCKET, SO_RCVBUF, &sockopt, sizeof(sockopt));
On some kernels (our 2.6.16.26 mips kernel for example), it breaks large
like this:
ping -s 64000 -c1 <dest>
This problem can be fixed either removing this setsockopt entirely or
by using a 96k receive bufsize instead of 48k. I picked 96k because
that's what the ping on my Redhat desktop (from iputils-20020927) uses.
I'm not sure exactly what this number is used for. You'd think it would
set the maximum size of a packet that could be received on the socket, but
on the above mentioned kernel, I can use a ping size of up to 57712 before
it breaks, and on a 2.6.9 x86 desktop kernel, it doesn't break at all.
It's not really a big deal that really large pings break, unless you
don't know that the bug is in ping and not in your networking stack,
and then proceed to waste a large chunk of time debugging the latter :-)
So, as the comment asks, can anybody "explain why 48k?".
Thanks,
Doug.
_______________________________________________
busybox mailing list
[email protected]
http://busybox.net/cgi-bin/mailman/listinfo/busybox
