Re: disk corruptions on "tuned" disks Was: APM killing low-latency performance on BX mainboard

Benno Senoner (sbenno@gardena.net)
Thu, 11 Nov 1999 02:04:16 +0100


On Thu, 11 Nov 1999, Alan Cox wrote:
> > for example "tune for realtime performance"
> > turns APM off and turns on DMA / unmaskirq /32bit access
> > on all EIDE device.
>
> At which point you just corrupted the hard disk on my old 486. Some things
> unfortunately really cannot be let loose like that.

Yes I know this, but this EIDE tuning is REQUIRED in order to
get rock solid scheduling performance out of your box.

Of course the "tuning tool" should check for problematic hardware,
and warn the user about potential problems, especially if he is
running Linux on old HW)

It would be nice if we could collect some data about problematic
drives / mainboards.
Alan any know drives ?
( If I get some data I will write a little tool which automates this)

Is this more a harddisk issue or an mainboard issue ?

On Pentium+ boxes, with not too old disks, the problem should be
pratically inexistent.

I understad this why most distros do not tune the disks in the initscripts.

But such an automatic tuning/check has soon to be implemented into distros,
in order to get the maximum performance out of the HW.
( Or are you not willing to try out the 5th gear of your Ferrari ? :-) )

Benno.

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