linux-origin
[Top] [All Lists]

Re: 2.3.41

To: ralf@xxxxxxxxxxxxxx (Ralf Baechle), kanoj@xxxxxxxxxxxxxxxxxxx (Kanoj Sarcar)
Subject: Re: 2.3.41
From: dagum@xxxxxxxxxxxxxxxxxxx (Leo Dagum)
Date: Sat, 29 Jan 2000 17:47:14 -0800
Cc: linux-origin@xxxxxxxxxxx
Sender: owner-linux-origin@xxxxxxxxxxx
> > 
> > Take a look at linux/Documentation/DMA-mapping.txt:
> > 
> > >                       Dynamic DMA mapping
> > >                       ===================
> > > 
> > >                David S. Miller <davem@xxxxxxxxxx>
> > >                Richard Henderson <rth@xxxxxxxxxx>
> > >                 Jakub Jelinek <jakub@xxxxxxxxxx>
> > 
> > This text documents an API which we now need to examine if suitable for
> > SNx.
> > 
> > Kanoj, re our phone discussion - it says:
> > 
> > > Drivers converted fully to this interface should not use virt_to_bus any
> > > longer, nor should they use bus_to_virt.
> > 
> >   Ralf
> > 
> 
> Leo, looks like other people are starting to get worried about the dma
> and pio issues too. After talking to Ralf, it is clear that some of us
> have to understand how irix handled this, issues involved in the sn0/sn1
> platforms, and then get in touch with David and Jakub ... fast.
> We probably want to reach consensus internally about the general
> structure of the apis first though. Its time to start looking at the 
> old papers people put up on bablyon regarding io apis in irix ....

This is a good thing really, even if it doesn't go fully our way
I'm guessing (w/o having seen the document yet) that it will be a
lot easier as a starting point than virt_to_bus/bus_to_virt and
should give us much better driver code capture than we could
otherwise hope for.

> 
> The sad truth is, we have to keep supporting older drivers which do
> not follow the apis ...
> 

Sure, but we were expecting that anyway.

- leo

> Kanoj
> 
Leo Dagum    SGI  Mountain View, CA 94043 (650-933-2179)

<Prev in Thread] Current Thread [Next in Thread>