On Mon, Apr 15, 2013 at 05:42:27PM +0200, Kevin Wolf wrote: > Am 15.04.2013 um 17:38 hat Gleb Natapov geschrieben: > > On Mon, Apr 15, 2013 at 10:56:42AM +0200, Kevin Wolf wrote: > > > Am 14.04.2013 um 14:42 hat Gleb Natapov geschrieben: > > > > On Fri, Apr 12, 2013 at 01:14:47PM +0200, Kevin Wolf wrote: > > > > > This adds a test case that jumps into VM86 by iret-ing to a TSS and back > > > > > to Protected Mode using a task gate in the IDT. > > > > > > > > > > Signed-off-by: Kevin Wolf <kwolf@xxxxxxxxxx> > > > > Applied, thanks. Found a bug with it and emulate_invalid_guest_state=1 > > > > which is default. Are you running with emulate_invalid_guest_state=0? > > > > > > Not knowingly at least, I didn't specify any module options. I > > > guess I just have enable_unrestricted_guest == true, which makes > > > guest_state_valid() return true immediately. > > > > > Can you check in > > /sys/module/kvm_intel/parameters/emulate_invalid_guest_state and > > /sys/module/kvm_intel/parameters/unrestricted_guest? You shouldn't have > > failed entry problem with enable_unrestricted_guest == true either. > > $ cat /sys/module/kvm_intel/parameters/emulate_invalid_guest_state > Y > $ cat /sys/module/kvm_intel/parameters/unrestricted_guest > Y > > Why do you think that I wouldn't have the failed entry check with > unrestricted_guest == true? The task switch is always done in software > and if it leads to an invalid segment descriptor in the VMCS, then a > failed VM entry looks quite expected to me. > unrestricted_guest eliminates a lot of check on vcpu state during guest entry since now real mode, along with big real mode, are virtualized directly by the CPU. But you are right that with VM86 we hit the check that is still there: if vcpu is in the VM86 mode segment registers should be in a specific state. -- Gleb. -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
