lkml.org 
[lkml]   [2000]   [Mar]   [29]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: The 2.3.x Job List (Updated)
On Tue, 28 Mar 2000, Alan Cox wrote:

> > > > Use PCI DMA 'lost interrupt' problem with some hw [which ?]
> > >
> > > wow !
> > > I do have 'Use PCI DMA' in my 2.2.15pre14
> > > and I do get 'lost interrupt' when trying to restore data from
> > > my ide-tape OnStream 30Go. of course it doesn't work.
> > >
> > > I have to try disabling 'Use PCI DMA' though.
> > >
> > >
> > > So your answer could include 'OnStream 30Go ide-tape'
> >
> > The OnStream DI-30 runs flawless in PIO or DMA mode.
>
> Then why are people seeing lost interrupts ?
>
> Right at the moment I suspect most of the lost interrupt stuff is PCI/APIC
> stuff, certainly on 2.3.9x. You can't simply say 'it runs flawlessly' when
> for someone it doesn't.

Whelp, this has been fun. I reported eariler that I was able to
get rid of the DMA timeouts by setting udma mode 0. I tried this
based on a completely flawed assumption.

The assumtion was that the values reported for tDMA in hdparm have
anything to do with udma modes. They do not - they are used to set
timings for multiword DMA modes (I don't have the Intel and
ATAPI-5 spec here today, so I'm not writing precisely.)

At least on the PIIX4, there are very few parameters that have
anything to do with UDMA. The CT values, which of course set the
UDMA mode and some other flags which I'm not sure if they even
effect UDMA. Specically PPE, DTE and TIME. IORDY also, but use of
IORDY is clearly linked to some features of the ATAPI drive.
Deciding whether or not it should be set is pretty easy.

And that's about it. There just isn't all sorts of magic needed.

I was reading the ATAPI-5 spec, and it talks about the use of an 80
wire cable when the PCB trace length is "long". This led me to the
idea that if my drive or motherboard was somewhat out of spec,
perhaps the better grounding supplied by the 80 wire cable would
help. Perhaps the IDE cables are running too close to the SCSI
cables. Maybe all the fans spinning in the case slam too many
electrons into photons affecting subspace.

Using the 80 wire cable allows UDMA/33 to work with my ASUS P2B-DS
and my Seagate ST39140A. Please note that the cable I took out was
most likely the cable that came with the Seagate. It is 18", with
two drive connectors about 6" apart. The ending connector was
attached to the drive.

This is on kernel 2.2.15pre14, with the ide patch from a few days
ago applied.

Perhaps we need a way to change the IDE controller's mode? Similar
to how hdparm can cause the features (transfer mode) of the IDE
device to be changed.

Mike


-
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/

\
 
 \ /
  Last update: 2005-03-22 13:57    [from the cache]
©2003-2014 Jasper Spaans. Advertise on this site