Hi Drew, On 3/11/20 4:38 PM, Andrew Jones wrote: > On Wed, Mar 11, 2020 at 02:51:11PM +0100, Eric Auger wrote: >> +/* must be called after gicv3_enable_defaults */ >> +void its_enable_defaults(void) >> +{ >> + int i; >> + >> + /* Allocate LPI config and pending tables */ >> + gicv3_lpi_alloc_tables(); >> + >> + for (i = 0; i < nr_cpus; i++) >> + gicv3_lpi_rdist_enable(i); > > You still haven't explained what's wrong with for_each_present_cpu. The previous comment you did was related to a spurious change I made in gicv3_lpi_alloc_tables. This change was removed in v5: [kvm-unit-tests PATCH v5 05/13] arm/arm64: gicv3: Set the LPI config and pending tables I did not understand from your comment you wanted all locations to use for_each_present_cpu(). I have nothing against it ;-) Also, > I see you've added 'i < nr_cpus' loops in arm/gic.c too. I'd prefer we not > assume that all cpu's are present (even though, currently, they must be), > because we may want to integrate cpu hotplug tests with these tests at > some point. OK Thanks Eric > >> + >> + writel(GITS_CTLR_ENABLE, its_data.base + GITS_CTLR); >> +} >> -- >> 2.20.1 >> >> > _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/mailman/listinfo/kvmarm