home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.unix.sysv386
- Path: sparky!uunet!elroy.jpl.nasa.gov!nntp-server.caltech.edu!egg!nat
- From: nat@egg.gg.caltech.edu (Nathan Inada)
- Subject: ISC 3.2/TCP-IP 1.3 connections hang
- Message-ID: <nat.715529534@egg.gg.caltech.edu>
- Summary: tcpip server connections hang. get protocol err on accept
- Keywords: tcp-ip, socket
- Sender: news@cco.caltech.edu
- Nntp-Posting-Host: egg.gg.caltech.edu
- Organization: California Institute of Technology, Pasadena
- Distribution: all
- Date: Thu, 3 Sep 1992 14:12:14 GMT
- Lines: 32
-
- i have set up a tcp connection server which "selects" on
- a set of "listen"ing sockets
-
- every now and again, under a heavy barrage of connection requests, i get
- errno = 71 "Protocol error"
- when "accept"ing on a selected socket. the message:
- "tcpcmds: non-fatal error in unknown event: 0000000B"
- gets sent to the console
-
- after this select no longer returns on connection requests
- though it does return for activity on already connected sockets.
- any requests that do come in are handled by what ever magic it is that actually
- deals with the tcp protocol. that is, subsequent connections ARE established,
- and data is passed (i know. i sniffed.), but select never hears about them.
- this lead to the following problem of closing connections with data in the
- queues... after a while the requesting client times out and closes the
- connection. this leaves a closed tcp connection with data in the queue.
- TCP/IP v 1.3 does not seem to deal with this situation well either.
- even when a tcp packet with the Reset bit set in its flags field is sent
- the protocol handler doesnt flush its buffers.
-
-
- has anybody run into this sort of thing ?
- the TCP/IP v 1.3 release notes say in its bug fixes "Multiple, simultanious
- incoming connections no longer hang a server"
- am i missing something ?
-
- any help would be greatly appreciated.
-
- nathan
- (nat@gg.caltech.edu)
-
-