Hard Reset of IXP1200 over PCI?

[Heaton, Eric D]

Scott,

The text you refered to in the Programmer's Reference Manual Section 4.7.13
is actually incorrect.  You may indeed write to the IXP1200_RESET register
to reset the chip over PCI, as you saw in your experiments.  The
documentation in the IXP1200 Datasheet Section 5.6.3 has the correct
information concerning this issue.

     Eric Heaton
     Technical Marketing Engineer
     Intel Corporation - PNPO

-----Original Message-----
From: Scott C. Karlin [mailto:scott at CS.Princeton.EDU]
Sent: Tuesday, February 20, 2001 10:27 AM
To: ixp1200 at CS.Princeton.EDU
Subject: Re: Hard Reset of IXP1200 over PCI?


I performed the experiment and answered my own question.

It seems that writing 0xFFFFFFFF to the IXP1200_RESET register
from the PCI bus does indeed reset the IXP1200.  As expected,
the reset knocks the IXP1200 off of the PCI bus.  However, once
I restored the PCI configuration registers to their original
state, the board worked fine.

Can anyone from Intel confirm (to the list) that the only reason
a PCI device should not write to the IXP1200_RESET register is
that the board will be deconfigured from the PCI bus?

Scott



> From: "Scott C. Karlin" <scott at CS.Princeton.EDU>
> Subject: Hard Reset of IXP1200 over PCI?
> To: ixp1200 at CS.Princeton.EDU
> Date: Wed, 14 Feb 2001 14:50:07 -0500 (EST)
> 
> I'd like to be able to perform a hard reset of the
> IXP1200 over the PCI bus.  Most of the documentation
> seems to indicate that I cannot do this.  However,
> the IXP1200 Programmer's Reference Manual seems to
> indicate that it might be possible (but not allowed).
> 
> Section 4.7.13 IXP1200_RESET indicates that the
> IXP1200_RESET register can be read and written
> from offset 0x7C.  However, the paragraph directly
> above says that "PCI devices must not write to this
> register."
> 
> What happens if I do?  (I assume that the IXP1200
> will reset and clear the PCI registers including the
> BAR set by the host machine.  If I promise to put the
> board back in the same place in the PCI memory map,
> will everything be OK?)
> 
> Scott