Messages in this thread | | | Date | Mon, 09 Jun 2008 18:02:42 -0300 | From | Glauber Costa <> | Subject | Re: [PATCH 11/15] x86: move enabling of io_apic to prepare_cpus |
| |
Maciej W. Rozycki wrote: > On Mon, 9 Jun 2008, Glauber Costa wrote: > >> This one does enable_IO_APIC in the init_uniprocessor too, and should >> account for the !smp case. > > Hmm, it looks a little bit better, but why do you want to call > enable_IO_APIC() separately in the first place? There is a comment > stating: "Enable IO APIC before setting up error vector," but why is it > needed on 64-bit systems? Especially as the very same system may run a > 32-bit kernel and then it suddenly would not have to do this anymore? > Strange...
This was reported by Yinghai, but I think he already answered to that.
> Also since you are cleaning up this code -- why don't you actually take > the opportunity and get rid of the horrible #ifdefs interspersed > throughout? throughout where? They're all over the place ;)
My next target would be per-cpu data. But that's because there's _a lot_ of code in the tree that got ifdefs between 32 and 64-bit because of differences in that, specially irq statistics. A macro would do, but if we're gonna do it, let's do it right.
| |