> I've been playing a bit with KVM on an A15 model, and did some changes > to support booting on an SMP host (SMP guest is currently unsupported, > but I'm looking at it). great, I am also looking at it a little. > > I've rebased Christoffer's patches on top of Catalin's v3.1 branch > (which includes support for LPAE as well as a bunch of other A15 related > patches like timers, platform support...). I did a rebase myself of Catalin's master branch. Was planning on releasing a v5 version of the patches this coming week based on c2ea487a643594c273e71881666b7ca629cf066b. Perhaps once I have this in a consistent state, I will push that as a branch that we can work on simultaenously by simply merging on top of that, and I can do the work of rebasing and integrating into a v6 patch series once that time comes and so on... What do you think? > > My changes mainly revolve around getting the secondary CPUs to be > correctly initialized (non-secure mode, GIC setup, HYP stack pages) and > a change to use the virtualized MPIDR so that a SMP-on-UP kernel boots > correctly. > > The kernel code is located on the following branch: > git://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git kvm looks good, I merged your things, and will give detailed feedback separately. One question though, the "fixup! ARM: local timers: Add A15 architected timer support" - is that required for KVM on smp, and if so, why? (I am still quite oblivious to the architected timer support). > > and the boot-wrapper is stashed there: > git://linux-arm.org/boot-wrapper.git kvm-smp > > Please do not consider any of these as stable branches, as I indent to > rebase them frequently. does this make sense? Consider the model I suggested above.
