| Message ID | 20230626182016.4127366-6-mizhang@google.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Update document description for kvm_mmu_page and kvm_mmu_page_role | expand |
On Mon, Jun 26, 2023, Mingwei Zhang wrote: > Add the description for mmu_valid_gen into kvm_mmu_page description. > mmu_valid_gen is used in shadow MMU for fast zapping. Update the doc to > reflect that. > > Signed-off-by: Mingwei Zhang <mizhang@google.com> > Reviewed-by: Kai Huang <kai.huang@intel.com> > --- > Documentation/virt/kvm/x86/mmu.rst | 4 ++++ > 1 file changed, 4 insertions(+) > > diff --git a/Documentation/virt/kvm/x86/mmu.rst b/Documentation/virt/kvm/x86/mmu.rst > index 97d695207e11..cc4bd190c93d 100644 > --- a/Documentation/virt/kvm/x86/mmu.rst > +++ b/Documentation/virt/kvm/x86/mmu.rst > @@ -208,6 +208,10 @@ Shadow pages contain the following information: > The page is not backed by a guest page table, but its first entry > points to one. This is set if NPT uses 5-level page tables (host > CR4.LA57=1) and is shadowing L1's 4-level NPT (L1 CR4.LA57=1). > + mmu_valid_gen: > + Used by comparing against kvm->arch.mmu_valid_gen to check whether the This needs to explain what the generation is, and where it comes from. The MMU generation of this page, used to effect a "fast" zap of all MMU pages across all roots. To zap all pages in all roots without blocking vCPUs, e.g. when deleting a memslot, KVM updates the per-VM valid MMU generation to mark all existing pages and roots invalid/obsolete. Obsolete pages can't be used, e.g. vCPUs must load a new, valid root before re-entering the guest. The MMU generation is only ever '0' or '1', as slots_lock must be held until all obsolete pages are zapped and freed, i.e. there is exactly one valid generation and (at most) one invalid generation. Note, the TDP MMU doesn't use mmu_gen as non-root TDP MMU pages are reachable only from their owning root, whereas all pages for shadow MMUs are reachable via the hash map. The TDP MMU uses role.invalid to track obsolete roots. And then big bonus points if you add Page Role ========= to explain the purpose of the role, and how/when it's used in the shadow MMU versus the TDP MMU. The shadow MMU's use of a hash map is a fundemental aspect that really should be documented here. > + shadow page is obsolete thus a convenient variable for fast zapping. > + Note that TDP MMU does not use mmu_valid_gen. > gfn: > Either the guest page table containing the translations shadowed by this > page, or the base page frame for linear translations. See role.direct. > -- > 2.41.0.162.gfafddb0af9-goog >
On Tue, Jun 27, 2023, Sean Christopherson wrote: > On Mon, Jun 26, 2023, Mingwei Zhang wrote: > > Add the description for mmu_valid_gen into kvm_mmu_page description. > > mmu_valid_gen is used in shadow MMU for fast zapping. Update the doc to > > reflect that. > > > > Signed-off-by: Mingwei Zhang <mizhang@google.com> > > Reviewed-by: Kai Huang <kai.huang@intel.com> > > --- > > Documentation/virt/kvm/x86/mmu.rst | 4 ++++ > > 1 file changed, 4 insertions(+) > > > > diff --git a/Documentation/virt/kvm/x86/mmu.rst b/Documentation/virt/kvm/x86/mmu.rst > > index 97d695207e11..cc4bd190c93d 100644 > > --- a/Documentation/virt/kvm/x86/mmu.rst > > +++ b/Documentation/virt/kvm/x86/mmu.rst > > @@ -208,6 +208,10 @@ Shadow pages contain the following information: > > The page is not backed by a guest page table, but its first entry > > points to one. This is set if NPT uses 5-level page tables (host > > CR4.LA57=1) and is shadowing L1's 4-level NPT (L1 CR4.LA57=1). > > + mmu_valid_gen: > > + Used by comparing against kvm->arch.mmu_valid_gen to check whether the > > This needs to explain what the generation is, and where it comes from. > > The MMU generation of this page, used to effect a "fast" zap of all MMU pages > across all roots. To zap all pages in all roots without blocking vCPUs, e.g. > when deleting a memslot, KVM updates the per-VM valid MMU generation to mark > all existing pages and roots invalid/obsolete. Obsolete pages can't be used, > e.g. vCPUs must load a new, valid root before re-entering the guest. > > The MMU generation is only ever '0' or '1', as slots_lock must be held until > all obsolete pages are zapped and freed, i.e. there is exactly one valid > generation and (at most) one invalid generation. > > Note, the TDP MMU doesn't use mmu_gen as non-root TDP MMU pages are reachable > only from their owning root, whereas all pages for shadow MMUs are reachable > via the hash map. The TDP MMU uses role.invalid to track obsolete roots. Sean, thanks for the detailed explanation. I will pick the most of the content and get into the next version. > > And then big bonus points if you add > > Page Role > ========= > > to explain the purpose of the role, and how/when it's used in the shadow MMU versus > the TDP MMU. The shadow MMU's use of a hash map is a fundemental aspect that really > should be documented here. > > > + shadow page is obsolete thus a convenient variable for fast zapping. > > + Note that TDP MMU does not use mmu_valid_gen. > > gfn: > > Either the guest page table containing the translations shadowed by this > > page, or the base page frame for linear translations. See role.direct. > > -- > > 2.41.0.162.gfafddb0af9-goog > >
diff --git a/Documentation/virt/kvm/x86/mmu.rst b/Documentation/virt/kvm/x86/mmu.rst index 97d695207e11..cc4bd190c93d 100644 --- a/Documentation/virt/kvm/x86/mmu.rst +++ b/Documentation/virt/kvm/x86/mmu.rst @@ -208,6 +208,10 @@ Shadow pages contain the following information: The page is not backed by a guest page table, but its first entry points to one. This is set if NPT uses 5-level page tables (host CR4.LA57=1) and is shadowing L1's 4-level NPT (L1 CR4.LA57=1). + mmu_valid_gen: + Used by comparing against kvm->arch.mmu_valid_gen to check whether the + shadow page is obsolete thus a convenient variable for fast zapping. + Note that TDP MMU does not use mmu_valid_gen. gfn: Either the guest page table containing the translations shadowed by this page, or the base page frame for linear translations. See role.direct.