[plug] hdparm troubles

[Bernd Felsche]

On Saturday 17 July 2004 17:28, Steve Boak wrote:
> On Sat, 17 Jul 2004 04:20 pm, Bernd Felsche wrote:
> > Just for you, I've just assembled a system with EPIA V C3 at 1GHz;
> > and 512MB -- gross overkill but components were on the shelf except
> > for the mainboard which was cheaper than the Eden.
>
> Wish I had that sort of kit 'just laying around' :-) Your processor is

I don't; usually. Just happened to have picked up the mainboard from
the post office. It's unfortunate that I had to order it in the
first place.

> twice the speed of mine, but that shouldn't make significant difference
> to the dma speed. I also have 512MB of Ram.

> > Those are the default settings detected at boot-up. I've not forced
> > any parameters. Just let the drivers do their thing.

> > clutch:/proc/ide # cat via
> > ----------VIA BusMastering IDE Configuration----------------
> > Driver Version:                     3.29
>
> I have version 3.37
>
> > South Bridge:                       VIA vt8231
> > Revision:                           ISA 0x10 IDE 0x6
> > Highest DMA rate:                   UDMA100
> > BM-DMA base:                        0xe000
>
> and 0xd000 here

I can't see a reason for that to be different unless perhaps you
have something plugged into the PCI. Unplug everything from the PCI
and disable everything you're not using in the BIOS.

BIOS is version 1.06 on the 1GHz V-Series.

> > -----------------------Primary IDE-------Secondary IDE------
> > Cable Type:                   80w                 40w

Make sure it's recognized the 80-wire cable on your system.

If you have access to another Linux system, try the hard drive on
that... could be a flakey drive.

> min:~# hdparm -i /dev/hda
> /dev/hda:

[snip]

>  DMA modes:  mdma0 mdma1 mdma2
>  UDMA modes: udma0 udma1 udma2 udma3 udma4 *udma5
>
> Hmmm - it's gone to udma5 on it's own now.

I work on the base assumption that the accummulated knowledge
presented in the drivers is far greater than my own tenuous
understanding.

You might want to try a smaller MaxMultSect on WDC drives. See
hdparm -m option description in the man pages.

> >  Timing buffer-cache reads:   128 MB in  1.25 seconds =102.40 MB/sec
> >  Timing buffered disk reads:  64 MB in  3.16 seconds = 20.25 MB/sec
>
>  Timing buffer-cache reads:   180 MB in  2.04 seconds =  88.24 MB/sec
>  Timing buffered disk reads:   12 MB in  3.27 seconds =   3.67 MB/sec

This does look a bit sus'. Why had hdparm chosen a smaller size of
transfer? The driver seems to have detected and chosen udma5; yet
selected a smaller transfer volume appropriate for mdma or even pio.

> > i.e. It looks like DMA works. Indeed more automagically than with
> > the EPIA-M I tested earlier.

> Wish it did on mine :-)

-- 
/"\ Bernd Felsche - Innovative Reckoning, Perth, Western Australia
\ /  ASCII ribbon campaign | I'm a .signature virus!
 X   against HTML mail     | Copy me into your ~/.signature
/ \  and postings          | to help me spread!