n Thu, Feb 21, 2013 at 10:00 AM, James White <james.white.2012@xxxxxxxxxxx> wrote: > >> Subject: Re: signal_pending(current) in kvm_arch_vcpu_ioctl_run() is always true >> >> On Wed, Feb 20, 2013 at 2:45 AM, James White <james.white.2012@xxxxxxxxxxx> wrote: >> > >> > >> > I have ported KVM/QEMU on ARM to a 3.4 kernel but have merged with https://github.com/virtualopensystems/linux-kvm-arm.git branch kvm-arm-v17-vgic-timers. >> > When I try to run up a VM, the kernel runs endlessly a loop wihtout entering guest code in kvm_arch_vcpu_ioctl_run() as signal_pending(current) is always true. Any idea how a scenario like this could happen? I'd really appreciate any pointers. >> > >> signal_pending() is not a kvm/arm specific function, and I doubt that >> it's buggy, so my best guess is that you're constantly sending a signal >> to the process, for whatever reason. >> >> You can block certain signals during vcpu execution using the >> KVM_SET_SIGNAL_MASK ioctl, or you can prevent these signals to be sent >> to your process in the first case. >> >> If you're using the proper QEMU versions and you backported the patches >> correctly, then I have no idea what's going on. >> > > Thanks for your reply. Could behavior like this also be due to the vgic not operating properly? Would a pending interrupt cause the same issues? > A pending interrupt in the vgic is a very different thing from a signal pending on the vcpu process. The only thing that *could* come close, would be if your emulated device uses a software timer or something like that, but then you should be aware of that already, and we're not seeing this behavior with qemu devices in general. -Christoffer _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
