Messages in this thread |  | | Subject | RE: [PATCH][2.6] fix i386 idle routine selection | Date | Wed, 11 Aug 2004 13:52:21 -0700 | From | "Pallipadi, Venkatesh" <> |
| |
>-----Original Message----- >From: linux-kernel-owner@vger.kernel.org >[mailto:linux-kernel-owner@vger.kernel.org] On Behalf Of Zwane >Mwaikambo > >Hi Venkatesh, > >On Tue, 10 Aug 2004, Pallipadi, Venkatesh wrote: > >> The original idea of setting it back to default_idle(), >> was to handle the cases: >> 1) CPU 0 says it can do mwait and CPU 1 later says it >> cannot do mwait >> 2) CPU 0 says it cannot do mwait and later CPU 1 says >> that it can do mwait >> >> Not that I know of any system like that. But, ideally >> mwait_idle should be set only when all CPUs say that >> they can do it. > >Hmm. Something like mwait/monitor really should be supported on all >processors in a system running the same executive, if you do happen to >come across a system with mixed capabilities then we should clear the >capability during the processor setup so that we never even select the >mwait idle. So we should defer handling such a scenario until someone >ships something, but generally the lowest common denominator with >respect to capabilities should be the BSP. >
Agreed that handling asymmetric CPUs here is a bit of overkill. We can stick to the original patch + few more cleanups (attached).
Thanks, Venki
[unhandled content-type:application/octet-stream]
|  |