> On Fri, 2002-11-29 at 13:17, Rasmus Bøg Hansen wrote:
> > I just tried turning off DMA on the server disk (this is just a low-end
> > IDE-system): No errors in files (compressing the file thrice).
> >
> > So it does not at all seem to be a NFS-issue!
> >
> > I have no idea what is wrong. If the disk, cable or IDE controller does
> > bit-flipping when DMA is turned on, why is the problem only seen with
> > NFS? I have never seem corrupted files or metadata with DMA turned
> > (except once long ago, when I experimented with high-transfer-modes - I
> > haven't done that since)...
>
> More likely it changes the timings. There is at least one other
> possibility though. With some via bridges using slightly too slow DDR
> RAM at a 133MHz clock works reliably _until_ you get a mix of CPU and
> DMA traffic. It'll even pass memtest86.
This is not DDR, it is SD-RAM (K6-2/300, 256MB SD-RAM in one stick).
But there is not much CPU traffic when making the files. The NFS-client
is doing the compression, but the corruption seems to happen on the
NFS-server (at least I can avoid it by turning off DMA on teh NFS
server), which has not got any special CPU usage at that time.
> So if its a VIA box, turn DMA back on, stick the bios into its load
> failsafe defaults mode and see if that has an affect.
I will try to look into BIOS settings (it does not have a 'load failsafe
option' IIRC, but I'm pretty sure I can change RAM timings) - but I
won't be near the machine before tomorrow.
Regards
/Rasmus
-- -- [ Rasmus "Møffe" Bøg Hansen ] --------------------------------------- There is no insanity, just different perceptions of reality. ----------------------------------[ moffe at amagerkollegiet dot dk ] --- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/