Messages in this thread |  | | Date | Sat, 22 Mar 2008 16:28:01 -0700 | From | "Yinghai Lu" <> | Subject | Re: ohci1394 problem (MMIO broken) (was 2.6.25-rc6-git6: Reported regressions from 2.6.24) |
| |
On Sat, Mar 22, 2008 at 2:59 PM, H. Peter Anvin <hpa@kernel.org> wrote: > Linus Torvalds wrote: > > > > Ingo/Thomas - should ioremap*() perhaps take "resource_size_t" or a "u64" > > for the address (and then "__ioremap()" should probably take a PFN, not a > > physical address, and that one can remain just a "unsigned long"?) > > > > resource_size_t seems to make sense here. In the case when we have > 64-bit resources but no PAE support we should error our when the > resource is out of range rather than silently fail. > > > > Has anybody ever had a working 64-bit BAR on x86? Ivan? Maybe I'm missing > > something.. > > 64-bit BAR, certainly. 64-bit BAR with a value beyond 4 GB I would find > questionable at best... I suspect the answer is "no". > Opteron system with coprocessor on socket or HTX slot. will use 64bit BAR above 4g and size is more than 4g too.
YH
|  |