Re: kern/34507: bge(4) doesn't contribute to the rnd pool

[Christoph Badura <bad%bsd.de@localhost>]

The following reply was made to PR kern/34507; it has been noted by GNATS.

From: Christoph Badura <bad%bsd.de@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: 
Subject: Re: kern/34507: bge(4) doesn't contribute to the rnd pool
Date: Sun, 24 Feb 2008 14:45:36 +0100

 On Sat, Feb 23, 2008 at 10:40:02AM +0000, Michael van Elst wrote:
 >  I also don't know what data in the driver could be considered
 >  random. The status word, which is used by other drivers, is almost
 >  always a constant when an interrupt occurs.
 
 The data doesn't need to be random. rnd_add_uint32() doesn't use the
 actual data.  It is interested only in the jitter between the events.
 
 I've looked at adding rnd(4) support for bge before.  I gave up because
 determining if the interrupt routine was called because of a interrupt
 from the device was expensive and, IIRC, not trivial without layering
 violations.  However, this is necessary, because the interrupt vector is
 often shared with other devices on the PCI bus (e.g. 8 other drivers on my
 IBM T43).
 
 --chris