[PATCH 2/2] examples: systemtap: Update to linux-6.3

[Peter Krempa]

On Mon, Mar 06, 2023 at 10:37:54 +0000, Daniel P. Berrangé wrote:
> On Mon, Mar 06, 2023 at 11:21:54AM +0100, Peter Krempa wrote:
> > The 'vmsa' struct was moved out of 'struct vcpu_svm' in linux commit:
> > 
> >   commit b67a4cc35c9f726999fa29880713ce72d4e39e8d
> >   Author: Peter Gonda <pgonda at google.com>
> >   Date:   Thu Oct 21 10:42:59 2021 -0700
> > 
> >       KVM: SEV: Refactor out sev_es_state struct
> > 
> >       Move SEV-ES vCPU metadata into new sev_es_state struct from vcpu_svm.
> > 
> > Also update the line reference for linux-6.3.
> > 
> > NB: I strongly considered removing the example as it's impossible to
> > keep in sync. With the warning added by previous commit I think we can
> > give it one more chance.
> 
> It sucks, but it sucks less than what every was doing before I
> suggested systemtap, which is to tell people to edit their
> kernel source to add printfs and then rebuild their kernel
> 
> Ideally the kernel would expose the pristine expected VMSA
> data in some sane interface like debugfs.
> 
> > 
> > Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2175598
> 
> In general such bugs should be closed WONTFIX. There is no
> expectation that this demo works on RHEL kernels. At most I
> would worry about upstream kernel, but even that's not very
> critical. This is just a demo of the approach you should take
> not a supported script we expect to work out of the box.

I agree. I'll close this one.

> 
> > Signed-off-by: Peter Krempa <pkrempa at redhat.com>
> > ---
> >  examples/systemtap/amd-sev-es-vmsa.stp | 6 +++---
> >  1 file changed, 3 insertions(+), 3 deletions(-)
> > 
> > diff --git a/examples/systemtap/amd-sev-es-vmsa.stp b/examples/systemtap/amd-sev-es-vmsa.stp
> > index 14bfb01c10..ab2f202681 100644
> > --- a/examples/systemtap/amd-sev-es-vmsa.stp
> > +++ b/examples/systemtap/amd-sev-es-vmsa.stp
> > @@ -46,7 +46,7 @@ function dump_vmsa(addr:long) {
> >  # is the one beween the call to clflush_cache_range(...) and the
> >  # call to sev_issue_cmd(kvm, SEV_CMD_LAUNCH_UPDATE...).
> >  #
> > -# Line 632 is correct for Linux v6.0
> > -probe module("kvm_amd").statement("__sev_launch_update_vmsa at arch/x86/kvm/svm/sev.c:632") {
> > -  dump_vmsa($svm->vmsa)
> > +# Line 635 is correct for Linux v6.3
> > +probe module("kvm_amd").statement("__sev_launch_update_vmsa at arch/x86/kvm/svm/sev.c:635") {
> > +  dump_vmsa($svm->sev_es->vmsa)

At least the change to this line IMO makes sense from upstream's POV
going forward as the code was refactored and thus the old example will
never work again with new kernels.