> ill actually be an older R1 of the PE2950 (older > > chipset, older dual core Xeons w/ Hyperthreading -- hopefully VMX cpu > > extensions) > > > > The metric for success will be a functional HVM DomU of RHEL5/x86_64, > > FreeBSD 6.3/amd64, and WinXP. Otherwise I'll have to spend another year > > using VMWare to simulate machine in our lab -- which would really > > suck :) I don't think that its going to work on our 3.1.x. Probably a 3.2.x- only fix. There is actually specific code in the 3.2 branch which checks for the presents of the variable being declared in the development environment: xen/arch/x86/hvm/vmx/vmx.c: else if ( v->arch.hvm_vmx.vmxassist_enabled ) xen/arch/x86/hvm/vmx/vmx.c:#else /* !defined(VMXASSIST) */ xen/arch/x86/hvm/vmx/Makefile:ifneq ($(vmxassist),y) xen/arch/x86/hvm/vmx/x86_64/exits.S:#ifndef VMXASSIST xen/arch/x86/hvm/vmx/x86_64/exits.S:#ifndef VMXASSIST xen/arch/x86/hvm/vpic.c:#ifdef VMXASSIST xen/arch/x86/hvm/vpic.c: current->arch.hvm_vmx.vmxassist_enabled; xen/arch/x86/x86_64/asm-offsets.c:#ifndef VMXASSIST This type of code doesn't exist in the 3.1. --------------------------------------- bash-3.2# strings /usr/pkg/libexec/qemu-dm|grep -i vmx (HVMXen) HVMXEN- bash-3.2# gzcat /xen.gz |strings |grep -i vmx vmx_intr_assist vmx_vmexit_handler vmx_do_invlpg vmx_set_cr0 vmx_world_restore vmx_io_instruction vmx_do_msr_read vmx_do_msr_write vmx_restore_guest_msrs vmx_vmcs_restore vmx.c VMXON failed VMX disabled by Feature Control MSR. VMX: MSR intercept bitmap enabled <G><0>%s:%d:d%d vmx_vmcs restore failed! [HVM:%d.%d] <%s> switching to vmxassist. use phys table <G><0>%s:%d:d%d Failed to transfer to vmxassist [HVM:%d.%d] <%s> Transfering control to vmxassist %%eip 0x%lx tvmx bash-3.2# strings /usr/pkg/lib/xen/boot/hvmloader |grep -i vmx %05x-%05x: VMXAssist Loading VMXAssist ... VMX go ... vmxassist returned VMXAssist (%s) > > I'm going to go with "vmxassist=n" would be a MAKE_ENV+= argument in > > sysutils/xentools3-hvm/Makefile and sysutils/xentools3/Makefile > > And sysutils/xenkernel3/Makefile - I suspect vmxassist is mostly a kernel > thing. > -- Brian A. Seklecki <bseklecki%collaborativefusion.com@localhost> Collaborative Fusion, Inc.
Description: PNG image