[PATCH] Fix LPC decode ranges on SB700

[Rudolf Marek]

Hello,

Attached patch fixes the LPC decode ranges of SB700. We enable early only 
Serial/SIO/RTC. Everything else needs to be done by lpc.c Problem was that early 
settings survived, because the lpc.c is doing ORs only...
Hence we decode quite a lot and even strange ranges like IO port 0x4600 etc...

Also, if some port which does not fit to predefined set is requested, like 0x290 
for Hardware monitor, the wide port is done, but in our case it has range 512 
bytes which means we decode in fact 0x290 - 0x490. And if we hit GPU in the 
0x3bx range I receive MCE exception if I do isadump -f 0x300 which is bad.
Therefore If I detect that the requested range is small (16 bytes) I 
additionally set the small wide io region so only 16 bytes is decoded.

While at it, I fix spelling typos and I init the regs so we don't write random 
garbage to regs even if we don't enable them later.

Signed-off-by: Rudolf Marek

Thanks,
Rudolf

[Stefan Reinauer]

* Rudolf Marek  [110126 00:01]:
> Hello,
>
> Attached patch fixes the LPC decode ranges of SB700. We enable early
> only Serial/SIO/RTC. Everything else needs to be done by lpc.c
> Problem was that early settings survived, because the lpc.c is doing
> ORs only...
> Hence we decode quite a lot and even strange ranges like IO port 0x4600 etc...
>
> Also, if some port which does not fit to predefined set is
> requested, like 0x290 for Hardware monitor, the wide port is done,
> but in our case it has range 512 bytes which means we decode in fact
> 0x290 - 0x490. And if we hit GPU in the 0x3bx range I receive MCE
> exception if I do isadump -f 0x300 which is bad.
> Therefore If I detect that the requested range is small (16 bytes) I
> additionally set the small wide io region so only 16 bytes is
> decoded.
>
> While at it, I fix spelling typos and I init the regs so we don't
> write random garbage to regs even if we don't enable them later.
>
> Signed-off-by: Rudolf Marek

Acked-by: Stefan Reinauer

Stefan

[Marc Jones]

On Tue, Jan 25, 2011 at 7:03 PM, Stefan Reinauer
 wrote:
> Rudolf Marek  [110126 00:01]:
> > Hello,
> >
> > Attached patch fixes the LPC decode ranges of SB700. We enable early
> > only Serial/SIO/RTC. Everything else needs to be done by lpc.c
> > Problem was that early settings survived, because the lpc.c is doing
> > ORs only...
> > Hence we decode quite a lot and even strange ranges like IO port 0x4600 etc...
> >
> > Also, if some port which does not fit to predefined set is
> > requested, like 0x290 for Hardware monitor, the wide port is done,
> > but in our case it has range 512 bytes which means we decode in fact
> > 0x290 - 0x490. And if we hit GPU in the 0x3bx range I receive MCE
> > exception if I do isadump -f 0x300 which is bad.
> > Therefore If I detect that the requested range is small (16 bytes) I
> > additionally set the small wide io region so only 16 bytes is
> > decoded.
> >
> > While at it, I fix spelling typos and I init the regs so we don't
> > write random garbage to regs even if we don't enable them later.
> >
> > Signed-off-by: Rudolf Marek
>
> Acked-by: Stefan Reinauer
>
> Stefan

I assume that sb600 and sb800 have the same bug.

[Rudolf Marek]

Oh yes seems it is quite similar. I could not test this, I'm attaching the changes.

Signed-off-by: Rudolf Marek

Thanks,
Rudolf

[Marc Jones]

On Fri, Jan 28, 2011 at 2:31 PM, Rudolf Marek  wrote:
> Oh yes seems it is quite similar. I could not test this, I'm attaching the
> changes.
>
> Signed-off-by: Rudolf Marek

Acked-by: Marc Jones