On 02/24/10 00:39, Joerg Sonnenberger wrote: > On Wed, Feb 24, 2010 at 12:18:13AM +0100, Jean-Yves Migeon wrote: >> But this doesn't tell me: should I go the "typdef uint64_t paddr_t" >> route? This may have unexpected overhead (which I cannot assess), >> and might consume extra stack/register just to store bytes which >> will never be used in non-PAE situation. > > I'd go for it. I suggest to carefully check a few places like twa(4) to > ensure they don't force 64bit PCI on !PAE kernels.
Then 64 bits it is. -- Jean-Yves Migeon [email protected]
