[ELDK] Jumbo Frames, sil24 SATA driver, and kswapd0 page allocation failures

[Dave Cogley]

What is the size of your Jumbo Frame MTU? I have had problems on the
440EP with an MTU larger than 4080 related to the MAL allocating
multiple buffer descriptors. Try shrinking your MTU to below 4K and see
if the problem still occurs.

Dave 

On Wed, 2009-08-12 at 15:50 -0600, Jonathan Haws wrote:
> > > Everything seems to point to the fact that when write() is called, 
> > > memory i s allocated from RAM to cache data before it is 
> > physically written to disk.
> > >   However, it appears (again, I am speculating here - if 
> > someone can 
> > > enligh ten me that would be great) that memory is not 
> > getting released 
> > > - simply be cause when the emac_poll_rx tries to allocate 
> > memory for 
> > > its next skb it fa ils and for all intents and purposes 
> > crashes the kernel.
> > > 
> > > I am almost certain that the issue is not in my code, but in the 
> > > kernel som ewhere.  But digging through libAta, I do not 
> > find anything 
> > > that would caus e this.  It appears that when memory is 
> > allocated, it 
> > > is released as soon a t it is finished using it.
> > 
> > Well, it would be definitely helpful if you first checked 
> > your code and added proper error checking to all system calls 
> > and library function calls, and bail out after an error occurs.
> 
> Okay, I have gone through my code, and added checks to every system
> call.  If there is erroneous behavior, then I call it out and kill
> everything.  However, I still see the crash and I never see any 
> errors come back from the system calls.  That tells me that it is not
> the application code, but something else.
> 
> Has anyone seen something similar to this before?
> 
> Thanks,
> 
> Jonathan
> _______________________________________________
> eldk mailing list
> eldk at lists.denx.de
> http://lists.denx.de/mailman/listinfo/eldk