serialice PCI capture

[Joseph Smith]

Just a thought. YABEL can capture PCI read/writes from an option ROM, 
how hard would it be to get serialice to do the same with the bios?

This would solve many of life's mysteries.....

[Joseph Smith]

On Thu, 01 Oct 2009 10:06:31 -0400, Joseph Smith 
wrote:
> Just a thought. YABEL can capture PCI read/writes from an option ROM,
> how hard would it be to get serialice to do the same with the bios?
>
> This would solve many of life's mysteries.....

Hmm, it appears serialice has PCI access functions.... :-)

[Peter Stuge]

Joseph Smith wrote:
> how hard would it be to get serialice to do the same with the bios?

I think it already does. PCI is just IO, so it gets handled.

//Peter

[Joseph Smith]

On Fri, 2 Oct 2009 01:08:45 +0200, Peter Stuge  wrote:
> Joseph Smith wrote:
> > how hard would it be to get serialice to do the same with the bios?
>
> I think it already does. PCI is just IO, so it gets handled.

Huh? PCI has IO space allocated, but I am talking about actual PCI config
registers...

[Stefan Reinauer]

Joseph Smith wrote:
> On Fri, 2 Oct 2009 01:08:45 +0200, Peter Stuge  wrote:
> > Joseph Smith wrote:
> > > how hard would it be to get serialice to do the same with the bios?
> >
> > I think it already does. PCI is just IO, so it gets handled.
>
> Huh? PCI has IO space allocated, but I am talking about actual PCI config
> registers...

Just try it :-)

[Joseph Smith]

On Fri, 02 Oct 2009 01:20:09 +0200, Stefan Reinauer 
wrote:
> Joseph Smith wrote:
> > On Fri, 2 Oct 2009 01:08:45 +0200, Peter Stuge  wrote:
> > > Joseph Smith wrote:
> > > > how hard would it be to get serialice to do the same with the bios?
> > >
> > > I think it already does. PCI is just IO, so it gets handled.
> >
> > Huh? PCI has IO space allocated, but I am talking about actual PCI

config
> > registers...
>
> Just try it :-)

Ok, will do, how long does it take to get through a 512k bios image?

[Stefan Reinauer]

Joseph Smith wrote:
> Ok, will do, how long does it take to get through a 512k bios image?

Running a complete image will require several days...

[ron minnich]

not a couple days! Why, I'm up to SPD dump in  a few hours!

ron

[Stefan Reinauer]

ron minnich wrote:
> not a couple days! Why, I'm up to SPD dump in  a few hours!

Some BIOSes do an extensive RAM check, and the protocol limits checking
RAM (or copying) to something around 16MB/h

The checks can easily be circumvented with gdb, but stuff like
optimizing memory copy is tricky..

Stefan

[Joseph Smith]

On 10/01/2009 07:50 PM, ron minnich wrote:
> not a couple days! Why, I'm up to SPD dump in  a few hours!
>
> ron

SerialICE does SPD Dumps too?
Capturing SPD reads as well would make writing raminit a snap :-)

[Stefan Reinauer]

On Oct 2, 2009, at 3:24, Joseph Smith  wrote:
> On 10/01/2009 07:50 PM, ron minnich wrote:
> > not a couple days! Why, I'm up to SPD dump in  a few hours!
> >
> > ron
>
> SerialICE does SPD Dumps too?
> Capturing SPD reads as well would make writing raminit a snap :-)

It captures _everything_ ... :-)

In theory we could do ram init using a "replay attack" mechanism...

In practice it's more complicated than that as the log shows you init  
for your very configuration only... Modulo you don't easily recognize  
loops and their exit conditions...

Stefan
>

[Peter Stuge]

Joseph Smith wrote:
> > I think it already does. PCI is just IO, so it gets handled.
>
> Huh? PCI has IO space allocated, but I am talking about actual PCI
> config registers...

You get reads and writes, because they go over IO cf8/cfc.

//Peter

[Joseph Smith]

On Fri, 2 Oct 2009 01:20:12 +0200, Peter Stuge  wrote:
> Joseph Smith wrote:
> > > I think it already does. PCI is just IO, so it gets handled.
> >
> > Huh? PCI has IO space allocated, but I am talking about actual PCI
> > config registers...
>
> You get reads and writes, because they go over IO cf8/cfc.
>
> //Peter

Oh, ok, gotcha :-)