Okay, thanks. I guess this brings up a more general question. For the purposes of the testsuite, I'm wondering if it should count this as a PASS or a FAIL. Currently, the test suite fails, because it was designed to work against all the CPUs. If the non-offlineability of cpu0 is intended as the proper behavior, I want to modify the testsuite to give passes on these architectures. I could modify the testsuite such that if the online file does not exist for cpu0, it ignores it. However I'm concerned this may hide other bugs in the future. Is there a better way to detect which cpu's support hotplugging operations? Thanks, Bryce On Tue, May 02, 2006 at 07:50:34AM -0700, Raj, Ashok wrote: > Yes, cpu0 was not offlineable since inception. (both in x86 and x86_64) > due to certain other dependencies that would need more cleanup in kernel > code. > > Even though you had an online file earlier, its was a no-op.. basically > when you echo 1 to offline to cpu0, we just returned error and did > nothing. > > Cheers, > ashok raj > - Open Source Technology Center > > > >-----Original Message----- > >From: Bryce Harrington [mailto:bryce@xxxxxxxx] > >Sent: Monday, May 01, 2006 5:38 PM > >To: Raj, Ashok > >Cc: Mary Edie Meredith; Hotplug SIG; lhcs-devel@xxxxxxxxxxxxxxxxxxxxx > >Subject: Hotplug CPU issue? > > > >On Tue, Apr 11, 2006 at 10:47:16AM -0700, Raj, Ashok wrote: > >> > >> > > >> > * Patch to fix the CPU issue we discovered with the > lhcs_regression > >> > testsuite has been validated to fix the issue. > >> > > >> > >> > >> Yes this patch is now included in mainline. > > > >Hi ashok, > > > >Thanks, I verified this is now working in mainline kernels. However > >I've noticed a new issue. > > > >I've been running my CPU hotplug regression test against recent -rc, > >-mm, and -git kernels, on a 2-cpu machine. But I notice that while > cpu1 > >can be onlined/offlined, cpu0 cannot. This is because there is no > >'online' file for cpu0: > > > > cl009 cpu # ls /sys/devices/system/cpu/cpu* > > /sys/devices/system/cpu/cpu0: > > topology > > > > /sys/devices/system/cpu/cpu1: > > online topology > > > >Why would I be seeing the online file only for cpu1? Is x86 incapable > >of offlining cpu0? > > > >Thanks, > >Bryce
