On Fri, Jun 12, 2015 at 12:06:35PM +0100, Marc Zyngier wrote: > The current way we deal with the GIC world switch on arm64 is a bit > convoluted. As we have two possible backends, we rely on a couple of > indirections set at boot time. These indirection never change, so it > would make more sense to directly patch in a branch to the right > function. > > To do this, we introduce a new CPU feature that detect the presence of > a GICv3 CPU interface. When detected, we patch branches to the GICv3 > save/restore functions that otherwise point to the GICv2 backend. > > Tested on both GICv2 (Juno, Seattle) and GICv3 (FVP Base model). > > These patches have been on the various lists for quite a while as part > of a series dealing with instruction patching. Now that the > prerequisite are queued for 4.2, it would make sense for these to be > merged alongside. > > Catalin, would you mind taking them through the arm64 tree? > > Marc Zyngier (2): > arm64: alternative: Introduce feature for GICv3 CPU interface > arm64: KVM: Switch vgic save/restore to alternative_insn Applied. Thanks. -- Catalin _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
