Re: [PATCH-2.5.46] IDE BIOS timings

[Vojtech Pavlik]

On Mon, Nov 11, 2002 at 12:10:13PM -0500, Bill Davidsen wrote:
> On 8 Nov 2002, Alan Cox wrote:
> 
> > On Fri, 2002-11-08 at 16:56, Jens Axboe wrote:
> > > > Linus please drop this patch for now. Its not been tested on enough
> > > > controllers, its making things unneccessarily ugly and its also just
> > > > going to make updates hard.
> > > 
> > > Alan, the patch is pretty much straight forward. Cleaning up the magic
> > > numbers and ->autotune consistencies is a good thing, imo.
> > 
> > You can clean up the naming but it still hasn't been tested, not all
> > bioses neccessarily give us timings we can trust either.  I'm not
> > opposed to the concept but after the previous IDE mess in 2.5 merging
> > something that isnt tested on lots of controllers and might have weird
> > effects does both me a bit
> 
> This is one of those things which we should allow at user risk. After all,
> you can shoot yourself in the foot with hdparm as well, there are many
> unwise things allowed.
> 
> Having seen all the warnings from bad setup of MPS and ACPI in dmesg, I
> would say it's more likely that the BIOS get these settings right, since
> they may be used by that other operating system.

I can tell you that many VIA (namely older) boards simply crash after
the first DMA access when you don't fix the timings/fifo settings of the
chip after what mess the BIOS left there.

-- 
Vojtech Pavlik
SuSE Labs