Re: [BUG?] SIS IDE DMA errors

[Lionel Bouton]

Vojtech Pavlik said the following on 09/26/2003 07:53 PM:

> On Fri, Sep 26, 2003 at 07:27:35PM +0200, M?ns Rullg?rd wrote:
>  
>
> > Vojtech Pavlik <vojtech@xxxxxxx> writes:
> >
> >    
> >
> > > Actually, it's me who wrote the 961 and 963 support. It works fine for
> > > most people. Did you check you cabling?
> > >      
> > I'm dealing with a laptop, but I suppose I could wiggle the cables a
> > bit.  I still doubt it's a cable problem, since reading works
> > flawlessly.
> >    
>
> Hmm, that's indeed interesting and it'd point to a driver problem -
> when reading, the drive is dictating the timing, but when writing, it's
> the controllers turn.
>
> So if the controller timing is not correctly programmed, reads function,
> but writes don't.
>
> Can you send me the output of 'lspci -vvxxx' of the IDE device?
> I'll take a look to see if it looks correct.
>
>  
>
> > It appears to me that during heavy IO load, some DMA interrupts get
> > lost, for some reason.
> >    
>
> Well, I've got this feeling that not just IDE interrupts get lost under
> heavy IO load with recent kernels ...
>
>  

This could explain some odd reports. Amongst the usual causes like flaky 
hardware, kernel misconfiguration and the likes, I encountered some 
people for which IO-APIC support would throw their data away...

Now I always ask the users to recompile without IO-APIC, this usually 
brings other problems (awful ethernet perfs for one user comes to my 
mind) but tends to solve IDE instability.

Until today, I've not a single report where lspci -vxxx highlighted any 
IDE register misconfiguration, AFAICS your code *is* correct Vojtech.

LB.

--
Lionel Bouton - inet6
---------------------------------------------------------------------
  o              Siege social: 51, rue de Verdun - 92158 Suresnes
 /      _ __ _   Acces Bureaux: 33 rue Benoit Malon - 92150 Suresnes
/ /\  /_  / /_   France
\/  \/_  / /_/   Tel. +33 (0) 1 41 44 85 36
 Inetsys S.A.    Fax  +33 (0) 1 46 97 20 10



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/