Re: [PATCH 0/4] x86: Restrict KVM-induced symbol exports to KVM
Peter Zijlstra <[email protected]>
| Newsgroups | org.kernel.vger.linux-sgx,dev.linux.lists.linux-coco,org.kernel.vger.kvm,org.kernel.vger.linux-kernel,org.kernel.vger.linux-perf-users |
|---|---|
| Message-ID | <[email protected]> |
On Wed, Nov 12, 2025 at 09:39:40AM -0800, Sean Christopherson wrote:
> Leverage and extend KVM's macro shenanigans to export symbols for KVM if
> and only if kvm{,-amd,intel}.ko is being built as a module, and only for
> the KVM modules that are being built.
>
> Note, this approach isn't 100% precise, as exports that are only strictly
> necessary for one of KVM's modules will get exported for all KVM modules.
> But I don't see any value in being super precise as it's not like kvm.ko is
> any more trustworthy tha kvm-{amd,intel}.ko (and it's easy to circumvent
> "for module" exports by abusing module names (in out-of-tree code)). And
> maintaining precise exports would likely be a nightmare (as would writing
> the macros to get the exports right).
>
> Patches 1-3 drop superfluous exports that I found while digging around for
> KVM-only exports.
>
> Sean Christopherson (4):
> x86/bugs: Drop unnecessary export of "x86_spec_ctrl_base"
> x86/mtrr: Drop unnecessary export of "mtrr_state"
> x86/mm: Drop unnecessary export of "ptdump_walk_pgd_level_debugfs"
> x86: Restrict KVM-induced symbol exports to KVM modules where
> obvious/possible
Nice!
Acked-by: Peter Zijlstra (Intel) <[email protected]>