On 10/01/13 06:58, xiangyou xie wrote: > Hi Marc: > > I have tried your kvm-arm64/soc-armv8-kvm branch. Host boot > successfully, but it’s failed to run guest, I don’t know what’s wrong > with it. The following is the detail of my implementation: > > (1) The code is from > git://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git > <https://email-cn.huawei.com/owa/redir.aspx?C=bb801de808eb45e6a9a2c865fd4007cb&URL=http%3a%2f%2fgit.kernel.org%2fpub%2fscm%2flinux%2fkernel%2fgit%2fmaz%2farm-platforms.git> Which branch? > (2) The FastModels is FastModels_atlas_0.8, and the model is Atlasx4 As mentioned in the post you quoted, this has only been tested on the VE model and the Foundation model. I suggest you start by using one of those. > (3) Vexpress-v2p-aarch64.dts for host and guest(guest use the same > kernel Image as host). This DT doesn't mean a thing for the guest. Kvmtool is providing its own DT to the guest anyway. > (4) I use kvm tool to boot guest, the command line is: > > #lkvm run –kernel Image -c 4 -m 512 –initrd filesystem.cpio.gz –dump-dtb > vexpress-v2p-aarch64.dtb -p “rdinit=/linuxrc console=ttyAMA0 mem=512M > root=/dev/ram” Well, kvmtool doesn't provide any non-virtio device. For a start, you should have "console=hvc0" instead of "console=ttyAMA0". You can also add "early_printk=smh". "mem=512" is superfluous. You don't need the --dump-dtb option unless you're debugging kvmtool. > What’s wrong with my steps or is something missing? Your reply and > suggestion will be very helpful. Thanks. My suggestion is that you start with a known, tested configuration/model (see the configuration I posted a few days ago in response to a similar question). M. > > 2012/12/19 Marc Zyngier <marc.zyngier@xxxxxxx <mailto:marc.zyngier@xxxxxxx>> > > As I'm about to go in turkey mode for a few days, I thought it would > be good to let people have a taste of KVM on arm64. At least for those > who are not following the same path... > > So this is the first public drop of KVM/arm64. It supports: > - arm and arm64 guests > - SMP host and guest > - 4kB and 64kB pages (both as a host and as a guest) > - Virtual GIC > - Architected timers > > It has been tested on ARMv8 VE model as well as the Foundation Model > (after reworking the DTS), using Will Deacon's excellent port of > kvmtool to arm64 for the platform emulation side (with virtio for > console, block device, networking...). We happily booted guests using > the Linaro OE filesystem (64bit) as well Debian and Ubuntu (32bit). > > It reuses a fair amount of the KVM/arm port source code and general > architecture, though there still some work to be done in order to > share more code. It currently relies on so much out of tree code that > I'm not even trying to post the patches, hence merely signaling the > existence of the port. > > The code sits in a number of (non-stable) branches at > git://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git > <http://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git> > > - kvm-arm64/pre-arm64-cleanup: Christoffer Dall's kvm-arm-master > branch plus a number of patches reorganizing some parts of the > KVM/arm code in order to reuse them on arm64. > > - kvm-arm64/psci: the above, plus the implementation of the Power > State Coordination Interface > > - kvm-arm64/soc-armv8-kvm: a merge of Catalin Marinas' soc-armv8-model > branch and the above. > > - kvm-arm64/soc-armv8-kvm-prereq: the above, plus a bunch of patches > adding the KVM prerequisites (including an architected timer rework > courtesy of Mark Rutland). > > - kvm-arm64/soc-armv8-kvm-arm64: The turkey itself. > > - kvm-arm64/kvmtool: The above, with stuffing, vegs and gravy. Enjoy. > > The soc-armv8-kvm-arm64 branch contains patches that are roughly split > by theme (world switch, MMU, exit handling...), but does not yet > reflect the way I intend to split them before review. > > The code is currently based on v3.7 + kvm-next. Don't try it if you're > not prepared to tinker with what is essentially experimental, fast > changing and probably broken code (though my tests seem to show that > it is solid enough. YMMV). > > What is missing: > - FP access from userspace > - COMPAT layer for 32bit userspace > > As always, comments are welcome, though I may be a bit slow to reply... > > M. > -- > Jazz is not dead. It just smells funny... > > > _______________________________________________ > kvmarm mailing list > kvmarm@xxxxxxxxxxxxxxxxxxxxx <mailto:kvmarm@xxxxxxxxxxxxxxxxxxxxx> > https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm > > -- Jazz is not dead. It just smells funny... _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
