On 05.01.2012 15:49, Gleb Smirnoff wrote:
mpd has many open sockets, and each socket allocates that much recvspace
and sendspace. Since there are a lot of them, it hits per-user resource
limits, I suppose.
We need to:
1) Change mpd to use one control socket with many hooks.
It is partially done. mpd uses common sockets to connect ppp, ccp and
ecp nodes of all links. But some link types still could be improved.
This also requires improving ng_socket, to have a fast hook lookup
method.
Agree.
2) Implement SO_SNDBUF/SO_RCVBUF for AF_NETGRAPH socket, and utilize
it in ngctl, so there would be no need to bump global tunables.
It would be nice. The biggest consumer of socket buffers I remember is
`ngctl list`. May be it could be reimplemented somehow to transfer
information in smaller chunks.
By the way about features: there were request recently to allow PPPoE
over ng_eiface interface. I think it should not be very difficult to add
orphan hook and several control requests to it to behave alike to ng_ether.
--
Alexander Motin
_______________________________________________
freebsd-net@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-net
To unsubscribe, send any mail to "freebsd-net-unsubscr...@freebsd.org"
