Date: Mon, 10 Jan 2000 13:09:45 +1000 From: Stanley.Hopcroft@ipaustralia.gov.au To: questions@freebsd.org Cc: isp@freebsd.org, grog@lemis.com Subject: Getting TCP connections prematurely FINned with duplxity mismatch - was read() fails (called from apache 1.3.9) on FreeBSD 3.2 -RELEASE Message-ID: <4A256862.00115FA7.00@noteshub01.aipo.gov.au>
next in thread | raw e-mail | index | archive | help
To: questions@freebsd.org cc: isp@freebsd.org, grog@lemis.com Dear Ladies and Gentlemen, I am writing to say the question recently asked about read() failing intermittently when called by apache 1.3.9 (and only with an RFC 1867 file upload) has an answer (courtesy Mr Carl Makin). Browser clients would report "connection reset by browser" when the browser attempted to upload a file over about one megabyte. tcpdump shows the FreeBSD/apache host sending a FIN packet (then a couple of RESEST) to the browsers TCP connection before the file upload is done. The problem seems not to be caused by . RFC 1323 extensions . RFC 1646 extensions It apparently *is* caused by a duplexity mismatch between the FreeBSD box and the adjacent (FreeBSD) router via a cross over cable. ifconfig fxp1 media 100baseTX mediaopt full-duplex removes all signs of the problem So, if anyone asks what causes read() on a socket to fail, you won't get fooled (like me). I suppose netstat -i fxp1 -w 1 when the problem was happening might have helped. Thank you, Yours sincerely, To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4A256862.00115FA7.00>