Date: Thu, 13 Dec 2001 10:05:01 -0800 (PST) From: John Polstra <jdp@polstra.com> To: stable@freebsd.org Cc: thz@Lennartz-electronic.de Subject: Re: TCP stack still hosed? Message-ID: <200112131805.fBDI51c14845@vashon.polstra.com> In-Reply-To: <20011213185153.A365@mezcal.tue.le> References: <20011213105451.A738@mezcal.tue.le> <200112131102.fBDB2WQ66827@apollo.backplane.com> <20011213185153.A365@mezcal.tue.le>
next in thread | previous in thread | raw e-mail | index | archive | help
In article <20011213185153.A365@mezcal.tue.le>, Thomas Zenker <thz@Lennartz-electronic.de> wrote: > the most impacting change is in tcp_usrreq.c > ---------------------------- > revision 1.51.2.10 > date: 2001/11/30 19:54:05; author: alfred; state: Exp; lines: +3 -3 > MFC: 1.64 (larger default tcp send/recieve buffers) > ---------------------------- > > It seems that the switch and/or the USB/ethernet adaptor is overrun > by the larger sendspace. So setting sysctl net.inet.tcp.xxxspace=16384 > on the USB/ethernet system solves the problem mostly. Try reducing the sysctl "net.inet.tcp.local_slowstart_flightsize" to 1 or 2. Does that also solve the problem? There seems to be mounting evidence that disabling slow start for local destinations is not a good strategy in general. It might have been OK in the days of hubs, but it seems to cause problems with many switches. John -- John Polstra John D. Polstra & Co., Inc. Seattle, Washington USA "Disappointment is a good sign of basic intelligence." -- Chögyam Trungpa To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200112131805.fBDI51c14845>