[ipxe-devel] [gPXE-devel] [PATCH 1/4] [io] Move get_memmap() to the I/O API group

Michael Brown mcb30 at ipxe.org
Tue Aug 17 00:32:57 UTC 2010

(CC to gpxe-devel stripped due to list censorship by Marty Connor)

On Tuesday 17 Aug 2010 01:24:22 Piotr Jaroszyński wrote:
> I appreciate your comments, in fact moving the b44 to driver arch/i386
> sounds like a good option to me, but I would like to get Michael's
> opinion on that as he suggested moving get_memmap to the I/O API group
> in the first place and already merged the patchset in ipxe :)
> Wrt the second option it just seems to be a bigger undertaking which I
> think should be covered in a separate patchset.

I don't think b44 should be moved to arch/i386, since there's nothing i386-
specific about it.  The problem of which DMA masks are supported by the 
hardware (and by the drivers, in the case of the majority of drivers that are 
not 64-bit aware) is a generic problem that is orthogonal to the question of 
CPU architecture.

I would support the idea of extending malloc_dma() to include the ability to 
specify an upper bound on the allocated address.  This would seem to fit 
naturally on top of the multiple memory pool patch due to be merged as part of 
the userspace patch series.


More information about the ipxe-devel mailing list