On Thu, 2023-02-09 at 17:21 +0800, Chao Gao wrote: > The subject doesn't match what the patch does; intercepting > CR4.LAM_SUP isn't done by this patch. How about: > > Virtualize CR4.LAM_SUP All right, although I think this patch is all about intercepting CR4.LAM_SUP. Additional handling on CR4 bits intercepting in kvm/vmx_set_cr4() isn't always necessary. > > and in the changelog, Do you mean in cover letter? or in this patch's description here? > explain a bit why CR4.LAM_SUP isn't > pass-thru'd and why no change to kvm/vmx_set_cr4() is needed. OK. Existing kvm/vmx_set_cr4() can handle CR4.LAM_SUP well, no additional code need to be added. If we take a look at kvm/vmx_set_cr4() body, besides the ultimate goal of vmcs_writel(CR4_READ_SHADOW, cr4); vmcs_writel(GUEST_CR4, hw_cr4); other hunks are about constructing/adjust cr4/hw_cr4. Those are for the CR4 bits that has dependency on other features/system status (e.g. paging), while CR4.LAM_SUP doesn't. > > On Thu, Feb 09, 2023 at 10:40:14AM +0800, Robert Hoo wrote: > > Remove CR4.LAM_SUP (bit 28) from default CR4_RESERVED_BITS, while > > reserve > > s/(bit 28)// > > > it in __cr4_reserved_bits() by feature testing. > > > > Signed-off-by: Robert Hoo <robert.hu@xxxxxxxxxxxxxxx> > > Reviewed-by: Jingqi Liu <jingqi.liu@xxxxxxxxx>