Date: Wed, 30 Jun 2004 10:32:52 -0700 From: Sean Chittenden <sean@chittenden.org> To: fabulous <fabulous@t7ds.com.br> Cc: freebsd-performance@freebsd.org Subject: Re: net.inet.tcp.sendspace Message-ID: <83A6DC4F-CABB-11D8-B344-000A95C705DC@chittenden.org> In-Reply-To: <40E2EF4D.3050508@t7ds.com.br> References: <40E1A47D.6020709@t7ds.com.br> <DFEBD5EB-C9F2-11D8-B73A-000A95C705DC@chittenden.org> <40E2EF4D.3050508@t7ds.com.br>
next in thread | previous in thread | raw e-mail | index | archive | help
> Thank you very very much.. but sysctl sendspace/recvspace will be the > limits then? Yes. > can I set SO_SNDBUF to something higher than tcp.sendspace? Hrm... tcp(4) is a bit ambiguous about this: MIB Variables The TCP protocol implements a number of variables in the net.inet.tcp branch of the sysctl(3) MIB. TCPCTL_SENDSPACE (sendspace) Maximum TCP send window. TCPCTL_RECVSPACE (recvspace) Maximum TCP receive window. I don't know if setsockopt(2) will allow you to exceed these values. I don't see anything in setsockopt(2) that suggests that you can't exceed TCPCTL_SENDSPACE with setsockopt(2). > I'm thinking of setting tcp.sendspace to 64k and using setsockopt on > ircd to set it to 2k (why an irc connection would need more than > that?), so apache will use 64k and ircd 2k.. am I right? :P This is a wise course of action given the uncertainty of setsockopt(2) and the net.inet.tcp.* interaction. I'd grep through the source and give you a definitive answer but am busy/too lazy at the moment. :) -sc -- Sean Chittenden
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?83A6DC4F-CABB-11D8-B344-000A95C705DC>