RE: [PATCH] [scsi] Remove __GFP_DMA

[James Bottomley <James.Bottomley@SteelEye.com>]

On Thu, 2007-05-24 at 09:24 -0400, Salyzyn, Mark wrote:
> So, is the sequence:
> 
> 	p = kmalloc(upsg->sg[i].count,GFP_KERNEL);
> 	. . .
> 	addr = pci_map_single(dev->pdev, p, upsg->sg[i].count,
> data_dir);
> 
> Going to ensure that we have a 31 bit (not 32 bit) physical address?

No, unfortunately.  Implementing kmalloc_mask() and kmalloc_dev() was
something I said I'd do ... about two years ago.

> If not, then I reject this patch. We can not consider replacement with
> pci_alloc_consistent until it works on AMD respecting the DMA masks.

It should, I believe ... x86_64 has a complex allocation scheme where
for masks < 32 bit it first tries in GFP_DMA32 and sees if it gets lucky
before falling back to GFP_DMA. i386 just goes straight to GFP_DMA.

James


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/