Re: Abnormal behavior after bumping net.inet6.udp6.sendspace

[mlelstv%serpens.de@localhost (Michael van Elst)]

timo%math.uni-bonn.de@localhost (Timo Buhrmester) writes:

>> It may be that sbmax should default to a larger value.
>Wouldn't it be relatively straightforward to guard against changing the
>send/recvspace sysctls past sbmax?

Yes. The socket buffer cannot be larger than sb_max_adj
where sb_max_adj = sb_max * MCLBYTES / (MSIZE + MCLBYTES).

MCLBYTES is 2048 for all archs
MSIZE is 256 or 256 depending on arch.

For amd64:   262144 * 2048 / (512 + 2048) = 209715


Of course you could then start sanitizing lots of other
sysctl values. I'm not sure if this is really worth it.