Re: 0-order allocation failed in 2.4.10-pre8

[Daniel Phillips]

On September 16, 2001 10:03 am, Jens Axboe wrote:
> On Sun, Sep 16 2001, Daniel Phillips wrote:
> > > Use the
> > > 
> > > 
*.kernel.org/pub/linux/kernel/people/axboe/patches/2.4.9/block-highmem-all
> > > 
> > > patch and you can use highmem without having to worry about failed
> > > 0-order bounce pages allocations.
> > 
> > Right, by using 64 bit DMA instead of bounce buffers.  But aren't there 
cases
> > where the 64 bit capable hardware isn't there but somebody still wants to 
use
> > highmem?
> 
> Yes of course. The common case is not 64-bit dma here though, it's just
> being able to DMA to highmem pages (just full 32-bit dma instead of low
> memory dma). And that should cover most systems out there.

Right, but that does not mean we can forget about bounce buffers, does it.  
Most users will probably be able to use full 32-bit dma and users with more 
than 4 GB of memory really should go to the effort of making sure their 
hardware supports 64 bit dma.  But there will still be a few people who have 
to use bounce buffers.

I'm just confirming that we really do have to push on and get bounce buffers 
working reliably, even if most people will be able to use your far nicer 
alternative.

--
Daniel