Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Posting-Version: version B 2.10 5/3/83; site mcvax.UUCP Path: utzoo!linus!philabs!mcvax!jim From: jim@mcvax.UUCP Newsgroups: net.unix-wizards,net.news.group,net.bugs.4bsd Subject: Re: 4.1c subgroup? Message-ID: <5260@mcvax.UUCP> Date: Thu, 30-Jun-83 00:04:23 EDT Article-I.D.: mcvax.5260 Posted: Thu Jun 30 00:04:23 1983 Date-Received: Thu, 30-Jun-83 12:39:35 EDT References: <1519@allegra.UUCP> Organization: Math.Centre, Amsterdam Lines: 36 Yes, I think we could get by with using net.bugs.4bsd, there isn't much traffic on it (we don't get much anyway). As a starter, has anyone tried to bind() to a socket in the UNIX domain? Fun, isn't it? The exact fix depends on what you think the format of an address in the UNIX domain should be, whether it should have "family" specifier or just the pathname (I plump for the latter). And isn't 14 characters a bit short for the FULL pathname? But anyway, you need to do a check on whether namei() sets u.u_error in unp_bind() in uipc_usrreq.c, otherwise maknode() tries to do something with the illegal value set into u.u_pdir by namei(), and SPLAT! So I fixed that, and the system decided to go into a tight loop if you try to interrupt a process while it is listen()ing but has had no connect()ions yet. I know where it is looping, but haven't got round to fixing it yet. To save time we changed the application program to use the TCP/IP domain, and, what do you know, in the same state as above, we got "panic: tcp_output"! A fix for this one is to add a case TCPS_CLOSED: tcp_usrclosed() in tcp_usrreq.c before the line tp = tcp_close(); (This bug occured, though the code is not the same, in 4.1A too). These things are not so easy to track down when /etc/savecore continues to say "dumping 0 bytes of core in vmcore.7" on reboot. Anyone have a fix for that? Do we all have these bugs? Are we all running the same version of 4.1C (ours dates from beginning of April)? Not afraid to let the hedgehogs call him Jungle Jim, Jim McKie ....decvax!mcvax!jim ...philabs!mcvax!jim