Hi, On Thu, Nov 26, 2009 at 02:23:09PM +0100, Alexander Graf wrote: > While S390x was one of the first targets that were supported by KVM it always > lacked qemu system emulation support. > > In order to change that sad fact, I figured I'd just take on the task myself, > taking kuli (http://www.ibm.com/developerworks/linux/linux390/kuli.html), > Documentation/s390/kvm.txt and lguest as starting points to write a clean, > qemu'ish S390x Virtio machine emulation. > > So now you can run Linux VMs on Linux on z/VM on LPAR on zSeries! > > This is the resulting code. Please comment on things you like and also on the > ones you don't :-). > > The patchset is based on Uli's S390x userspace emulation patches. There's not > really that much shared functionality, but I didn't want to reimplement the > configure wheels. So make sure to have his patches applied first. > > Please only run S390x system virtualization using io threads. S390x has very > few intercepts and even less that actually reach userspace. So you'll end up > with a non-responsive VM without a dedicated IO thread. > > Use: $ ./configure --target-list=s390x-softmmu --enable-io-thread > > Currently only -kernel, -initrd and friends work. Booting from a real dasd > device is not supported. > > Also to actually run this code you need a patch for an ugly bug in the kernel > module that Carsten sent a patch to on the ML. > > I verified that the userspace actually works on a 2.6.27 (SLES11) kernel, so > if it doesn't work on current git, please tell me! I'm too afraid that I might > end up in a 3270 terminal to update the kernel on my z/VM instance :-). > > ##### WARNING ##### > > Apparently the qemu review process has scared off yet another developer for > now. > > So I'm sending this patchset without relying on Uli's set but rather > implemented small stubs for TCG host and TCG guest support, so we can run > on KVM. He promised me to send it again "soon" - whenever that is. > > These stubs don't serve any purpose but to get KVM running, so you get no x86 > on S390X or S390X on x86 for now! >
This patchset basically looks ok. I have some minor comments, a lot of them being related to coding style, but they should be easy to solve, and should not prevent inclusion in 0.12.0 release. Please find comments in the individual patches. -- Aurelien Jarno GPG: 1024D/F1BCDB73 aurel...@aurel32.net http://www.aurel32.net
