On Sunday 13 February 2005 11:32, Oded Arbel wrote:
> > > Trying not to sound like a noob, but ... how do I set it ? I hope
> > > not by recompiling my kernel which is something I loathe to do on a
> > > production machine, you understand - I haven't even upgraded to
> > > 2.6.

This is in the kernel code itself, so it's a *change* of the code
and than recompile...

> > Afraid so. It might be already set, or there might be a module option
> > to turn it on - which kernel are you using and which NIC driver?
> 
> eepro100 on 2.4.25

At least on 2.6 (that's what sitting on my disk) both drivers
available for eepro (Becker's eepro100.c and Intel's e100.c)
do not use these interrupts as entropy source. As Oleg explained
in a previous post -- this is a wise thing (Network interrupts
may be scheduled by a potential attacker).

Weirdly enough, some network drivers do use SA_SAMPLE_RANDOM
(e.g: e1000.c from Intel and 3c523.c).

> > > I looked at sysctl -a, but nothing looked related. Also I hear that
> > > unlike SCSI, IDE drives do supposed to contribute to entropy, but
> > > while I do have an IDE drive there doesn't seem to be any entropy
> > > generated from it - is it possible to turn this on as well ?
> >
> > Same answer I'm afraid. Which IDE module?
> 
> ahmm.. the standard one ?

Just checked (on my 2.6.10) -- the IDE irq grabbing is common
code (on drivers/ide/ide-probe.c) and *does not* contribute
to random number entropy. Also, from the SCSI variety, almost
none is used (drivers/scsi/nsp32.c does, but I don't know this
chip/device).

Can anybody explain why not? My only guess is that block devices
are too much regularized by the block layer (elevator etc.) and
so don't have enough randomness. Any better explanation?

Actually, since this is a policy issue, maybe this should be
tunable (per driver? per IRQ?). Looks like a future patch
is pending here :-)

-- 
Oron Peled                             Voice/Fax: +972-4-8228492
[EMAIL PROTECTED]                  http://www.actcom.co.il/~oron
ICQ UIN: 16527398

"It's almost like we're doing Windows users a favor by charging them money
for something they could get for free, because they get confused otherwise."
 - Larry Wall.

=================================================================
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word "unsubscribe" in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]

Reply via email to