Powerbase IDE: DMA & ohare-ata

Powerbase IDE: DMA & ohare-ata

[Ty Hedrick]

I just noticed the thread from a few days ago on greatly speeding up IDE
transfer rates by poking at ohare and thought I'd give it a shot
on my Powerbase. The first thing I noticed is that my IDE drive was still
operating in PIO mode, even with BenH.'s latest test kernel.  

It seems that even though the Powerbase is perfectly capable of DMA, the
test in ide-pmac.c to setup DMA 

#ifdef CONFIG_BLK_DEV_IDEDMA_PMAC
                if (np->n_addrs >= 2 && np->n_intrs >= 2) {

fails because n_intrs is 0.  Risking my mostly worthless IDE drive I
removed the test and recompiled.  DMA appears to work without a hitch -
transfer rates are almost identical to those under PIO, but cpu usage was 
much lower.

After looking at DMA I decided to enable the ohare register mentioned
earlier, stuffing the new value in also resulted in a working system, and
approximately doubled the IDE transfer rate as measured by hdparm - to a
whopping 3.7 MB/s.

Is there any way to be kind to all the Powerbase owners out there and
enable DMA by default without breaking any other systems?  Also, the mod
to ohare and heathrow does seem to help lots of platforms.

Ty    


[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to  Cc linuxppc-dev  if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]

Re: Powerbase IDE: DMA & ohare-ata

[Benjamin Herrenschmidt]

On Fri, Jan 15, 1999, Ty Hedrick <thedrick@netspace.org> wrote:

>After looking at DMA I decided to enable the ohare register mentioned
>earlier, stuffing the new value in also resulted in a working system, and
>approximately doubled the IDE transfer rate as measured by hdparm - to a
>whopping 3.7 MB/s.
>
>Is there any way to be kind to all the Powerbase owners out there and
>enable DMA by default without breaking any other systems?  Also, the mod
>to ohare and heathrow does seem to help lots of platforms.

The test kernel I have on my web site will soon have code for both ohare
and heathrow. I've added an option do disable the speed boost in case of
troubles (a command line option, currently implemented as a "hack"
because I didn't want to touch non-arch files). I'll add the fix for
not-checking the DMA interrupts, they are not used anyway.

I'll release this version with diffs as soon as I have finished some
other little things in progress, probably during the week-end.

If the speed-boost proves to be harmless to everyone, I'll leave it in
and probably submit it 'as-is'. If it doesn't, I'll either change the
command-line arg to optionally enable it 
instead of optionally disable it, or add some drive-specific parameters
or some kind of "exception list".


-- 
           E-Mail: <mailto:bh40@calva.net>
BenH.      Web   : <http://calvaweb.calvacom.fr/bh40/>





[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to  Cc linuxppc-dev  if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]

Re: Powerbase IDE: DMA & ohare-ata

[Paul Mackerras]

Ty Hedrick <thedrick@netspace.org> wrote:

> It seems that even though the Powerbase is perfectly capable of DMA, the
> test in ide-pmac.c to setup DMA 
> 
> #ifdef CONFIG_BLK_DEV_IDEDMA_PMAC
>                 if (np->n_addrs >= 2 && np->n_intrs >= 2) {
> 
> fails because n_intrs is 0.  Risking my mostly worthless IDE drive I

Could you get a copy of lsprop (ftp://samba.anu.edu.au/pub/linux-pmac/misc)
and do `lsprop /proc/device-tree/bandit/ohare/ide' (or whatever the
node for your IDE controller is called) and send it to me?

In fact we don't use the DMA interrupt, so the test for np->n_intrs
could be removed.

Paul.

[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to  Cc linuxppc-dev  if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]