From owner-freebsd-hackers  Tue Jul  4 21:07:12 1995
Return-Path: hackers-owner
Received: (from majordom@localhost)
          by freefall.cdrom.com (8.6.10/8.6.6) id VAA01799
          for hackers-outgoing; Tue, 4 Jul 1995 21:07:12 -0700
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.34])
          by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id VAA01785
          for <hackers@FreeBSD.org>; Tue, 4 Jul 1995 21:06:55 -0700
Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id NAA14946; Wed, 5 Jul 1995 13:48:50 +1000
Date: Wed, 5 Jul 1995 13:48:50 +1000
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199507050348.NAA14946@godzilla.zeta.org.au>
To: davidg@Root.COM, jmb@kryten.atinc.com
Subject: Re: splnet and if_ep.c
Cc: hackers@FreeBSD.org
Sender: hackers-owner@FreeBSD.org
Precedence: bulk

>>the 3c509 is reported to be buggy.  though i have not had any problems 
>>with it.  however, if writing smcioctl() for the smc91c92 chip, i see 
>>that if_ze.c:ze_ioctl() calls splnet() at entry.  if_ep.c:ep_ioctl() does 
>>not.   anybody seen any problems with the ix or zp drivers?  from grep's 
>>it appears that splnet is a higher priority than splimp.

>   In general, the ioctl routines of the drivers should be protected with
>splimp(). This is because they often write to device registers or modify the
>board state in some other way that must not be interrupted.

Yes, splnet is a _lower_ priority than splimp and probably should never be
used in the hardware parts of network drivers.  It is only used in if_zp.c
(among network card drivers) and that use seems to be a bug.

The networking part of lpt.c seems to have a bug in the opposite direction.
It uses splclock() instead of splimp() so it may break clock interrupt
handling.

Bruce