Re: [PATCH] tools lib traceevent: update KVM plugin

[Paolo Bonzini <pbonzini@redhat.com>]

On 20/10/2015 16:44, Steven Rostedt wrote:
> What happens if you run new perf on an older kernel. Is this new plugin
> going to be screwed up? Plugins should be backward compatible.

If you run new perf on older kernel, the new plugin will print the 
"role" field (see kvm_mmu_print_role) slightly incorrectly.  That said, 
the existing plugin was _also_ printing the role in a wildly wrong 
format, like 2.6.35 vintage; the glevels field was removed by commit 
5b7e0102ae74, "KVM: MMU: Replace role.glevels with role.cr4_pae", in 
April 2010.

Going forward it's really unlikely that the role will change apart from 
adding new bits.  These can be added to the plugin while keeping it 
backwards-compatible.  Addition to the role happen when you implement 
new virtual MMU features such as SMEP, SMAP or SMM.  That's once per year
or less.

> Is the plugin even still needed? I'm looking at some of the kvm events
> and they seem to be mostly self sufficient. What ones need a plugin
> today?

Yes, most of them are.  It's only needed for kvm_mmu_get_page, 
kvm_mmu_prepare_zap_page and kvm_emulate_insn.  The latter is only 
interesting if you install the disassembler library, and I wouldn't 
really care if it went away.

kvm_mmu_get_page and kvm_mmu_prepare_zap_page, however, have output like

    kvm_mmu_get_page: [FAILED TO PARSE] mmu_valid_gen=0x2 gfn=786432 role=1923 root_count=0 unsync=0 created=1

without the plugin vs.

    kvm_mmu_get_page: new sp gfn c0000 3 q0 direct rwx !pae !nxe !wp root 0 sync

with the plugin.

Paolo