Re: Bug: Latest (rc2 and beyond) kernel fails on Dreamcast

[Adrian McMenamin <adrian@newgolddream.dyndns.info>]

On Tue, 2009-01-20 at 20:40 +0000, Adrian McMenamin wrote:
> 2009/1/20 Paul Mundt <lethal@linux-sh.org>:
> > On Mon, Jan 19, 2009 at 10:33:57PM +0000, Adrian McMenamin wrote:
> >> 2009/1/19 Adrian McMenamin <lkmladrian@gmail.com>:
> >> > I haven't had time to look at this, but the most up to date kernel
> >> > (and the one of around a week ago also) fails to boot, seemingly
> >> > seeking an IP address for the 8139too based NIC.
> >>
> >> > [    4.664236] Sending DHCP requests .<3>eth0: PCI Bus error 2200.
> >> > [    4.669582] eth0: PCI Bus error 2200.
> >>
> >> And the bisection reveals the following:
> >>
> >> adrian@bossclass:~/linux-2.6$ git bisect bad
> >> 58c6d3dfe436eb8cfb451981d8fdc9044eaf42da is first bad commit
> >> commit 58c6d3dfe436eb8cfb451981d8fdc9044eaf42da
> >> Author: Johannes Weiner <hannes@cmpxchg.org>
> >> Date:   Tue Jan 6 14:43:10 2009 -0800
> >>
> >>     dma-coherent: catch oversized requests to dma_alloc_from_coherent()
> >>
> 
> 
> 
> >
> > The dreamcast uses a 32kB coherent area for PCI DMA, which all of the
> > 8139too allocations should fit under just fine. What does
> > dma_alloc_from_coherent() say the requested size is, vs the per-device
> > memory size for each allocation?
> >
> 
> Hmmm...
> 
> >From printks I stuck in the code
> 
> [    3.641267] Size is 0x1800, mem->size is 0x8
> [    3.644996] Size is 0x2810, mem->size is 0x8
> 
> The sizes are correct - the first is the TX buffer, the second RX, but
> I have no idea why mem->size is set so low, I assume that is a bug
> somewhere else in the Dreamcast code. Any clues gratefully received.
> --


Actually, it looks to me that the patch was wrong.

This is from kernel/dma-coherent.c

dev->dma_mem->size = pages;

Yet this is the test applied in dma_alloc_from_coherent:

if (unlikely(size > mem->size))

Should surely be:

if (unlikely (size > (mem->size >> PAGE_SHIFT)))

Will patch that and try it




> To unsubscribe from this list: send the line "unsubscribe linux-sh" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html