On 2019-Sep-11, at 08:15, Mark Johnston <markj at freebsd.org> wrote: > On Wed, Sep 11, 2019 at 07:57:26AM -0700, Mark Millard wrote: >> >> >> On 2019-Sep-11, at 07:31, Mark Johnston <markj at freebsd.org> wrote: >> >>> On Tue, Sep 10, 2019 at 10:58:05PM -0700, Mark Millard wrote: >>>> In a context with: >>>> >>>> # cpuset -g >>>> pid -1 mask: 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27 >>>> pid -1 domain policy: first-touch mask: 0, 1 >>>> >>>> I get: >>>> >>>> # cpuset -l0 -n prefer:0 COMMAND >>>> cpuset: setdomain: Invalid argument >>>> >>>> # cpuset -l0 -n prefer:2 COMMAND >>>> cpuset: setdomain: Invalid argument >>>> >>>> But one prefer:? value does allow the COMMAND >>>> to run: >>>> >>>> # cpuset -l0 -n prefer:1 COMMAND >>>> >>>> This seem odd to me. Am I missing something? >>>> >>>> For reference: I'm using a ThreadRipper 1950X >>>> with a head -r351227 based context for this >>>> activity. The above happens to have been run >>>> in a Windows 10 Pro HyperV session, instead >>>> of in a native-boot of the same media. (A >>>> native-boot would have had 32 CPUs.) >>> >>> Can you please show the output of "sysctl vm.phys_segs" from this >>> setup? >> >> Sure: > > I was wondering if you had only one domain populated, but it seems not > to be the case. Could you try updating to r351672 or later and see if > the behaviour persists? It may be a bit before I do that. FYI: I had set MAXMEMDOM to match the number of actual domains for the context: /usr/src/sys/amd64/conf/GENERIC-DBG:options MAXMEMDOM=2 /usr/src/sys/amd64/conf/GENERIC-NODBG:options MAXMEMDOM=2 (These kernel configuration files include GENERIC.) === Mark Millard marklmi at yahoo.com ( dsl-only.net went away in early 2018-Mar)Received on Wed Sep 11 2019 - 15:11:52 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:41:21 UTC