From owner-freebsd-hackers  Fri Jan  6 09:25:09 1995
Return-Path: hackers-owner
Received: (from root@localhost) by freefall.cdrom.com (8.6.9/8.6.6) id JAA19153 for hackers-outgoing; Fri, 6 Jan 1995 09:25:09 -0800
Received: from iguana.reptiles.org (iguana.reptiles.org [142.57.253.130]) by freefall.cdrom.com (8.6.9/8.6.6) with SMTP id JAA19119; Fri, 6 Jan 1995 09:25:00 -0800
Received: by iguana.reptiles.org (/\==/\ Smail3.1.28.1 #28.4)
	id <m0rQIOZ-00033zC@iguana.reptiles.org>; Fri, 6 Jan 95 12:24 EST
Message-Id: <m0rQIOZ-00033zC@iguana.reptiles.org>
From: jim@reptiles.org (Jim Mercer)
Subject: 2.0-R bug in setgroups() or sys/nfs/nfs_socket.c?
To: bugs@freebsd.org, hackers@freebsd.org
Date: Fri, 6 Jan 1995 12:24:10 -0500 (EST)
X-Mailer: ELM [version 2.4 PL2]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Length: 1828      
Sender: hackers-owner@freebsd.org
Precedence: bulk


[ i'm re-doing this, as i think my original got lost in the
  majordomo failure last night ]

after installing 2.0, i enabled YP and mounted /home and /var/mail via
NFS (off of my SunOS 4.1.1U1 machine).

the system started panic'ing with a "nfsreq nogrps" message.
sys/nfs/nfs_socket.c:922

after adding a printf to the kernel, i discovered that smail3 was
attempting to trundle through one of the NFS mounted directories.

it was at this point that it panic'd.

also, as far as the kernel was concerned, the process doing the
trundling, had no gid.

i looked into the smail3 source and found that it makes the following
initialization:

smail-3.1.29.1:src/main.c:186

#ifdef  HAVE_SETGROUPS
    /* clear out all extra groups.  We don't want to have to deal with them */
    {
        gid_t dummy;
        (void) setgroups(0, &dummy);
    }
#endif  /* HAVE_SETGROUPS */

according to the smail3 conf files, HAVE_SETGROUPS is defined for these
systems: aix3.1 aux3.0 bsd4.2 bsd4.3 freebsd2.0 hp-ux7.0 hp-ux8.0 linux
mips-bsd4.3 next2.0 sco3.2 sys5.4

i cannot beleive that these other systems would behave the same (ie. panicing
on NFS if no gid).

questions:

should setgroups() set at least one group (ie. from the gid field in 
/etc/passwd)?

should NFS not be so sensitive to the lack of a gid?

is the smail code "bad"?

i've looked and this same code causes NFS to hang on 1.1.5.1 (but interestingly
enough, it does not panic.) the panic message in nfs_socket.c was added
sometime between 1.1.5.1 and 2.0-R.

-- 
[ Jim Mercer                 jim@reptiles.org              +1 416 506-0654 ]
[          Reptilian Research -- Longer Life through Colder Blood          ]
[   Never, ever forget to replace the toiletseat after use!!! A wet        ]
[   chinchilla is a very funny and pathetic sight.   -- alt.chinchilla     ]