Messages in this thread | | | Date | Tue, 12 Jan 2010 11:57:38 -0800 | Subject | Re: [RFC PATCH 1/5] use nr_cpus= to set nr_cpu_ids early | From | Yinghai Lu <> |
| |
On Tue, Jan 12, 2010 at 11:32 AM, Christoph Lameter <cl@linux-foundation.org> wrote: > > On Tue, 12 Jan 2010, Yinghai Lu wrote: > >> on x86, before prefill_possible_map(), nr_cpu_ids will be NR_CPUS aka CONFIG_NR_CPUS >> >> add nr_cpus= to set nr_cpu_ids. so we can simulate cpus <=8 on normal config. >> instead of change NR_CPUS directly. > > We already have > > maxcpus=x? > > Why do we need this twice?
maxcpus only change setup_max_cpus., and if you are using maxcpus=1, and you have 8 cpus installed, you can put other cpus back online via /sys/interface.
nr_cpus= is hard limit nr_cpu_ids, so if you have 16 cpus installed, nr_cpus=8 will make your nr_cpu_ids=8, and you can not put back other 8 back. and apic mode could stay with logical flat. this is used to simulate some debug case. for example you have kernel support physflat, and flat, with CONFIG_NR_CPUS=255. to run that on system that only have 8 cpus, you will have apic mode in logical flat. but if you have run the kernel on system with 32 cpus installed, it will switch to physflat even you have maxcpus=8 appedded.
YH
| |