From owner-freebsd-hackers Thu Feb 4 09:09:39 1999 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id JAA17812 for freebsd-hackers-outgoing; Thu, 4 Feb 1999 09:09:39 -0800 (PST) (envelope-from owner-freebsd-hackers@FreeBSD.ORG) Received: from detlev.UUCP (70-sweet.camalott.com [208.239.153.70]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id JAA17787 for ; Thu, 4 Feb 1999 09:09:23 -0800 (PST) (envelope-from joelh@gnu.org) Received: (from joelh@localhost) by detlev.UUCP (8.9.2/8.9.1) id LAA22755; Thu, 4 Feb 1999 11:08:41 -0600 (CST) (envelope-from joelh) To: Graham Wheeler Cc: hackers@FreeBSD.ORG Subject: Re: TCP keepinit time and connection timeouts References: <199902041251.OAA27910@cdsec.com> From: Joel Ray Holveck Date: 04 Feb 1999 11:08:40 -0600 In-Reply-To: Graham Wheeler's message of "Thu, 4 Feb 1999 14:51:10 +0200 (SAT)" Message-ID: <86aeyuqep3.fsf@detlev.UUCP> Lines: 21 X-Mailer: Gnus v5.5/Emacs 20.3 Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG > We have noticed that the connect() timeout when the target host does not > respond is much shorter than the value specified by the TCP keepinit time. > With TCP keepinit set to 25, connect() timesout after about 10 seconds, > while with TCP keepinit set to 75, connect() times out after about 37 seconds. > So it appears the keepinit time is being divided by some value in the range > 2.0 - 2.5 to obtain the actual timeout. All the TCP timers are in terms of PR_SLOWHZ, a kernel-wide constant which is generally 2. (This is noted near the top of /sys/netinet/tcp_timer.h.) That is because all TCP timeouts (except delayed acks) are handled by tcp_timer.c:tcp_slowtimo(), which is called PR_SLOWHZ times per second. Cheers, joelh -- Joel Ray Holveck - joelh@gnu.org Fourth law of programming: Anything that can go wrong wi sendmail: segmentation violation - core dumped To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message