Re: problem with blk_queue_bounce_limit()

[David Mosberger <davidm@napali.hpl.hp.com>]

>>>>> On 05 Jun 2003 00:20:53 -0700, "David S. Miller" <davem@redhat.com> said:

  DaveM> On Wed, 2003-06-04 at 23:42, David Mosberger wrote:
  >> On platforms with I/O MMU hardware, memory above 4GB, and IDE hard disks,
  >> this check:

  >> BUG_ON(dma_addr < BLK_BOUNCE_ISA);

  DaveM> Doesn't panic on sparc64, let this be your guiding light :-)

I checked with Dave in private email and, like I suspected, this is a
(potential) problem on sparc64 as well.  Let me be try to be even more
clear: on many 64 bit platforms, BLK_BOUNCE_ISA will be bigger than
4GB, so the BUG_ON() will trigger for IDE controllers that can DMA
only to, say, 4GB (and it's probably not even an IDE-only problem, I
think we don't see it on SCSI because our machines have SCSI
controllers that can DMA >4GB).

Manfred, I'm readdressed this mail to you because according to google,
you're the original author of the patch
(http://www.cs.helsinki.fi/linux/linux-kernel/2002-02/0032.html).

Like I stated earlier, I think this code simply makes no sense on a
platform with I/O MMU.  Hence my suggestion to deal with this via
dma_supported().

	--david