| Message ID | 20211115234603.2908381-7-bgardon@google.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Currently disabling dirty logging with the TDP MMU is extremely slow. On a 96 vCPU / 96G VM it takes ~45 seconds to disable dirty logging with the TDP MMU, as opposed to ~3.5 seconds with the legacy MMU. This series optimizes TLB flushes and intro | expand |
On 11/16/21 00:45, Ben Gardon wrote: > The vCPU argument to mmu_try_to_unsync_pages is now only used to get a > pointer to the associated struct kvm, so pass in the kvm pointer from > the beginning to remove the need for a vCPU when calling the function. > > > Signed-off-by: Ben Gardon <bgardon@google.com> > --- > arch/x86/kvm/mmu/mmu.c | 16 ++++++++-------- > arch/x86/kvm/mmu/mmu_internal.h | 2 +- > arch/x86/kvm/mmu/spte.c | 2 +- > 3 files changed, 10 insertions(+), 10 deletions(-) > > diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c > index 7d0da79668c0..1e890509b93f 100644 > --- a/arch/x86/kvm/mmu/mmu.c > +++ b/arch/x86/kvm/mmu/mmu.c > @@ -2561,10 +2561,10 @@ static int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva) > return r; > } > > -static void kvm_unsync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp) > +static void kvm_unsync_page(struct kvm *kvm, struct kvm_mmu_page *sp) > { > trace_kvm_mmu_unsync_page(sp); > - ++vcpu->kvm->stat.mmu_unsync; > + ++kvm->stat.mmu_unsync; > sp->unsync = 1; > > kvm_mmu_mark_parents_unsync(sp); > @@ -2576,7 +2576,7 @@ static void kvm_unsync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp) > * were marked unsync (or if there is no shadow page), -EPERM if the SPTE must > * be write-protected. > */ > -int mmu_try_to_unsync_pages(struct kvm_vcpu *vcpu, struct kvm_memory_slot *slot, > +int mmu_try_to_unsync_pages(struct kvm *kvm, struct kvm_memory_slot *slot, > gfn_t gfn, bool can_unsync, bool prefetch) > { > struct kvm_mmu_page *sp; > @@ -2587,7 +2587,7 @@ int mmu_try_to_unsync_pages(struct kvm_vcpu *vcpu, struct kvm_memory_slot *slot, > * track machinery is used to write-protect upper-level shadow pages, > * i.e. this guards the role.level == 4K assertion below! > */ > - if (kvm_slot_page_track_is_active(vcpu->kvm, slot, gfn, KVM_PAGE_TRACK_WRITE)) > + if (kvm_slot_page_track_is_active(kvm, slot, gfn, KVM_PAGE_TRACK_WRITE)) > return -EPERM; > > /* > @@ -2596,7 +2596,7 @@ int mmu_try_to_unsync_pages(struct kvm_vcpu *vcpu, struct kvm_memory_slot *slot, > * that case, KVM must complete emulation of the guest TLB flush before > * allowing shadow pages to become unsync (writable by the guest). > */ > - for_each_gfn_indirect_valid_sp(vcpu->kvm, sp, gfn) { > + for_each_gfn_indirect_valid_sp(kvm, sp, gfn) { > if (!can_unsync) > return -EPERM; > > @@ -2615,7 +2615,7 @@ int mmu_try_to_unsync_pages(struct kvm_vcpu *vcpu, struct kvm_memory_slot *slot, > */ > if (!locked) { > locked = true; > - spin_lock(&vcpu->kvm->arch.mmu_unsync_pages_lock); > + spin_lock(&kvm->arch.mmu_unsync_pages_lock); > > /* > * Recheck after taking the spinlock, a different vCPU > @@ -2630,10 +2630,10 @@ int mmu_try_to_unsync_pages(struct kvm_vcpu *vcpu, struct kvm_memory_slot *slot, > } > > WARN_ON(sp->role.level != PG_LEVEL_4K); > - kvm_unsync_page(vcpu, sp); > + kvm_unsync_page(kvm, sp); > } > if (locked) > - spin_unlock(&vcpu->kvm->arch.mmu_unsync_pages_lock); > + spin_unlock(&kvm->arch.mmu_unsync_pages_lock); > > /* > * We need to ensure that the marking of unsync pages is visible > diff --git a/arch/x86/kvm/mmu/mmu_internal.h b/arch/x86/kvm/mmu/mmu_internal.h > index 52c6527b1a06..1073d10cce91 100644 > --- a/arch/x86/kvm/mmu/mmu_internal.h > +++ b/arch/x86/kvm/mmu/mmu_internal.h > @@ -118,7 +118,7 @@ static inline bool kvm_vcpu_ad_need_write_protect(struct kvm_vcpu *vcpu) > kvm_x86_ops.cpu_dirty_log_size; > } > > -int mmu_try_to_unsync_pages(struct kvm_vcpu *vcpu, struct kvm_memory_slot *slot, > +int mmu_try_to_unsync_pages(struct kvm *kvm, struct kvm_memory_slot *slot, > gfn_t gfn, bool can_unsync, bool prefetch); > > void kvm_mmu_gfn_disallow_lpage(const struct kvm_memory_slot *slot, gfn_t gfn); > diff --git a/arch/x86/kvm/mmu/spte.c b/arch/x86/kvm/mmu/spte.c > index 75c666d3e7f1..b7271daa06c5 100644 > --- a/arch/x86/kvm/mmu/spte.c > +++ b/arch/x86/kvm/mmu/spte.c > @@ -160,7 +160,7 @@ bool make_spte(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp, > * e.g. it's write-tracked (upper-level SPs) or has one or more > * shadow pages and unsync'ing pages is not allowed. > */ > - if (mmu_try_to_unsync_pages(vcpu, slot, gfn, can_unsync, prefetch)) { > + if (mmu_try_to_unsync_pages(vcpu->kvm, slot, gfn, can_unsync, prefetch)) { > pgprintk("%s: found shadow page for %llx, marking ro\n", > __func__, gfn); > wrprot = true; > Queued, thanks. Paolo
diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c index 7d0da79668c0..1e890509b93f 100644 --- a/arch/x86/kvm/mmu/mmu.c +++ b/arch/x86/kvm/mmu/mmu.c @@ -2561,10 +2561,10 @@ static int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva) return r; } -static void kvm_unsync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp) +static void kvm_unsync_page(struct kvm *kvm, struct kvm_mmu_page *sp) { trace_kvm_mmu_unsync_page(sp); - ++vcpu->kvm->stat.mmu_unsync; + ++kvm->stat.mmu_unsync; sp->unsync = 1; kvm_mmu_mark_parents_unsync(sp); @@ -2576,7 +2576,7 @@ static void kvm_unsync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp) * were marked unsync (or if there is no shadow page), -EPERM if the SPTE must * be write-protected. */ -int mmu_try_to_unsync_pages(struct kvm_vcpu *vcpu, struct kvm_memory_slot *slot, +int mmu_try_to_unsync_pages(struct kvm *kvm, struct kvm_memory_slot *slot, gfn_t gfn, bool can_unsync, bool prefetch) { struct kvm_mmu_page *sp; @@ -2587,7 +2587,7 @@ int mmu_try_to_unsync_pages(struct kvm_vcpu *vcpu, struct kvm_memory_slot *slot, * track machinery is used to write-protect upper-level shadow pages, * i.e. this guards the role.level == 4K assertion below! */ - if (kvm_slot_page_track_is_active(vcpu->kvm, slot, gfn, KVM_PAGE_TRACK_WRITE)) + if (kvm_slot_page_track_is_active(kvm, slot, gfn, KVM_PAGE_TRACK_WRITE)) return -EPERM; /* @@ -2596,7 +2596,7 @@ int mmu_try_to_unsync_pages(struct kvm_vcpu *vcpu, struct kvm_memory_slot *slot, * that case, KVM must complete emulation of the guest TLB flush before * allowing shadow pages to become unsync (writable by the guest). */ - for_each_gfn_indirect_valid_sp(vcpu->kvm, sp, gfn) { + for_each_gfn_indirect_valid_sp(kvm, sp, gfn) { if (!can_unsync) return -EPERM; @@ -2615,7 +2615,7 @@ int mmu_try_to_unsync_pages(struct kvm_vcpu *vcpu, struct kvm_memory_slot *slot, */ if (!locked) { locked = true; - spin_lock(&vcpu->kvm->arch.mmu_unsync_pages_lock); + spin_lock(&kvm->arch.mmu_unsync_pages_lock); /* * Recheck after taking the spinlock, a different vCPU @@ -2630,10 +2630,10 @@ int mmu_try_to_unsync_pages(struct kvm_vcpu *vcpu, struct kvm_memory_slot *slot, } WARN_ON(sp->role.level != PG_LEVEL_4K); - kvm_unsync_page(vcpu, sp); + kvm_unsync_page(kvm, sp); } if (locked) - spin_unlock(&vcpu->kvm->arch.mmu_unsync_pages_lock); + spin_unlock(&kvm->arch.mmu_unsync_pages_lock); /* * We need to ensure that the marking of unsync pages is visible diff --git a/arch/x86/kvm/mmu/mmu_internal.h b/arch/x86/kvm/mmu/mmu_internal.h index 52c6527b1a06..1073d10cce91 100644 --- a/arch/x86/kvm/mmu/mmu_internal.h +++ b/arch/x86/kvm/mmu/mmu_internal.h @@ -118,7 +118,7 @@ static inline bool kvm_vcpu_ad_need_write_protect(struct kvm_vcpu *vcpu) kvm_x86_ops.cpu_dirty_log_size; } -int mmu_try_to_unsync_pages(struct kvm_vcpu *vcpu, struct kvm_memory_slot *slot, +int mmu_try_to_unsync_pages(struct kvm *kvm, struct kvm_memory_slot *slot, gfn_t gfn, bool can_unsync, bool prefetch); void kvm_mmu_gfn_disallow_lpage(const struct kvm_memory_slot *slot, gfn_t gfn); diff --git a/arch/x86/kvm/mmu/spte.c b/arch/x86/kvm/mmu/spte.c index 75c666d3e7f1..b7271daa06c5 100644 --- a/arch/x86/kvm/mmu/spte.c +++ b/arch/x86/kvm/mmu/spte.c @@ -160,7 +160,7 @@ bool make_spte(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp, * e.g. it's write-tracked (upper-level SPs) or has one or more * shadow pages and unsync'ing pages is not allowed. */ - if (mmu_try_to_unsync_pages(vcpu, slot, gfn, can_unsync, prefetch)) { + if (mmu_try_to_unsync_pages(vcpu->kvm, slot, gfn, can_unsync, prefetch)) { pgprintk("%s: found shadow page for %llx, marking ro\n", __func__, gfn); wrprot = true;
The vCPU argument to mmu_try_to_unsync_pages is now only used to get a pointer to the associated struct kvm, so pass in the kvm pointer from the beginning to remove the need for a vCPU when calling the function. Signed-off-by: Ben Gardon <bgardon@google.com> --- arch/x86/kvm/mmu/mmu.c | 16 ++++++++-------- arch/x86/kvm/mmu/mmu_internal.h | 2 +- arch/x86/kvm/mmu/spte.c | 2 +- 3 files changed, 10 insertions(+), 10 deletions(-)