Re: DMA APIs gumble grumble

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

> pci_alloc_consistent() is not allowed from atomic contexts.

Yes, but some drivers did it anyway, though I can't remember under which
circumstances (IDE probe possibly ? It's a usual culprit for that sort
of thing). This is why most implementations use GFP_ATOMIC (including
sparc64 :-)

> > I'm splitting it into a pci_do_alloc_consistent that takes a gfp arg,
> > and a pair of pci_alloc_consistent & dma_alloc_consistent wrappers.
> > 
> > Do you think I should have the former pass GFP_KERNEL like the current
> > implementation does or switch it to GFP_ATOMIC like everybody does ? In
> > this case, should I also change the kmalloc done in there to allocate a
> > struct resource to use the gfp argument ? (It's currently doing
> > GFP_KERNEL).
> 
> pci_alloc_consistent() really cannot be allowed to use GFP_ATOMIC.

Oh well, I have no problem with leaving sparc32 do GFP_KERNEL indeed, I
can't remember for sure the reason why we have most architectures do
GFP_ATOMIC, but it probably never hit sparc32.

Ben.