| Message ID | 20210817002639.3856694-1-jingzhangos@google.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | KVM: stats: Add VM stat for remote tlb flush requests | expand |
On Tue, 17 Aug 2021 01:26:39 +0100, Jing Zhang <jingzhangos@google.com> wrote: > > Add a new stat that counts the number of times a remote TLB flush is > requested, regardless of whether it kicks vCPUs out of guest mode. This > allows us to look at how often flushes are initiated. nit: this is a very x86-centric view of things. On arm64, TLB invalidation is broadcast in HW, and does not interrupt the guest execution. > > Original-by: David Matlack <dmatlack@google.com> > Signed-off-by: Jing Zhang <jingzhangos@google.com> > --- > arch/arm64/kvm/mmu.c | 1 + > include/linux/kvm_host.h | 3 ++- > include/linux/kvm_types.h | 1 + > virt/kvm/kvm_main.c | 1 + > 4 files changed, 5 insertions(+), 1 deletion(-) > > diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c > index 0625bf2353c2..f5bb235bbb59 100644 > --- a/arch/arm64/kvm/mmu.c > +++ b/arch/arm64/kvm/mmu.c > @@ -80,6 +80,7 @@ static bool memslot_is_logging(struct kvm_memory_slot *memslot) > */ > void kvm_flush_remote_tlbs(struct kvm *kvm) > { > + ++kvm->stat.generic.remote_tlb_flush_requests; > kvm_call_hyp(__kvm_tlb_flush_vmid, &kvm->arch.mmu); > } We already have this queued for 5.15 [1]. Thanks, M. [1] https://lore.kernel.org/r/20210727103251.16561-1-pbonzini@redhat.com
Hi Marc, On Wed, Aug 18, 2021 at 1:11 AM Marc Zyngier <maz@kernel.org> wrote: > > On Tue, 17 Aug 2021 01:26:39 +0100, > Jing Zhang <jingzhangos@google.com> wrote: > > > > Add a new stat that counts the number of times a remote TLB flush is > > requested, regardless of whether it kicks vCPUs out of guest mode. This > > allows us to look at how often flushes are initiated. > > nit: this is a very x86-centric view of things. On arm64, TLB > invalidation is broadcast in HW, and does not interrupt the guest > execution. > Understood. Thanks. > > > > Original-by: David Matlack <dmatlack@google.com> > > Signed-off-by: Jing Zhang <jingzhangos@google.com> > > --- > > arch/arm64/kvm/mmu.c | 1 + > > include/linux/kvm_host.h | 3 ++- > > include/linux/kvm_types.h | 1 + > > virt/kvm/kvm_main.c | 1 + > > 4 files changed, 5 insertions(+), 1 deletion(-) > > > > diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c > > index 0625bf2353c2..f5bb235bbb59 100644 > > --- a/arch/arm64/kvm/mmu.c > > +++ b/arch/arm64/kvm/mmu.c > > @@ -80,6 +80,7 @@ static bool memslot_is_logging(struct kvm_memory_slot *memslot) > > */ > > void kvm_flush_remote_tlbs(struct kvm *kvm) > > { > > + ++kvm->stat.generic.remote_tlb_flush_requests; > > kvm_call_hyp(__kvm_tlb_flush_vmid, &kvm->arch.mmu); > > } > > We already have this queued for 5.15 [1]. I guess you mean the change for "remote_tlb_flush" has been queued, not "remote_tlb_flush_requests"? These two counters would have the same value for arm64, but not for others (at least x86). > > Thanks, > > M. > > [1] https://lore.kernel.org/r/20210727103251.16561-1-pbonzini@redhat.com > > -- > Without deviation from the norm, progress is not possible. Thanks, Jing
On Wed, 18 Aug 2021 17:15:36 +0100, Jing Zhang <jingzhangos@google.com> wrote: > > Hi Marc, > > On Wed, Aug 18, 2021 at 1:11 AM Marc Zyngier <maz@kernel.org> wrote: > > > > On Tue, 17 Aug 2021 01:26:39 +0100, > > Jing Zhang <jingzhangos@google.com> wrote: > > > > > > Add a new stat that counts the number of times a remote TLB flush is > > > requested, regardless of whether it kicks vCPUs out of guest mode. This > > > allows us to look at how often flushes are initiated. > > > > nit: this is a very x86-centric view of things. On arm64, TLB > > invalidation is broadcast in HW, and does not interrupt the guest > > execution. > > > Understood. Thanks. > > > > > > Original-by: David Matlack <dmatlack@google.com> > > > Signed-off-by: Jing Zhang <jingzhangos@google.com> > > > --- > > > arch/arm64/kvm/mmu.c | 1 + > > > include/linux/kvm_host.h | 3 ++- > > > include/linux/kvm_types.h | 1 + > > > virt/kvm/kvm_main.c | 1 + > > > 4 files changed, 5 insertions(+), 1 deletion(-) > > > > > > diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c > > > index 0625bf2353c2..f5bb235bbb59 100644 > > > --- a/arch/arm64/kvm/mmu.c > > > +++ b/arch/arm64/kvm/mmu.c > > > @@ -80,6 +80,7 @@ static bool memslot_is_logging(struct kvm_memory_slot *memslot) > > > */ > > > void kvm_flush_remote_tlbs(struct kvm *kvm) > > > { > > > + ++kvm->stat.generic.remote_tlb_flush_requests; > > > kvm_call_hyp(__kvm_tlb_flush_vmid, &kvm->arch.mmu); > > > } > > > > We already have this queued for 5.15 [1]. > I guess you mean the change for "remote_tlb_flush" has been queued, > not "remote_tlb_flush_requests"? Yes, this is what I meant, as I only quoted the bit that was redundant. > These two counters would have the same value for arm64, but not for > others (at least x86). Sure. It is just that this patch does two things at once, but doesn't advertise it. Thanks, M.
On 18/08/21 18:52, Marc Zyngier wrote: > Yes, this is what I meant, as I only quoted the bit that was redundant. > >> These two counters would have the same value for arm64, but not for >> others (at least x86). > Sure. It is just that this patch does two things at once, but doesn't > advertise it. All in all, Jing's new stat is more appropriate than the one I used. I'll fix the conflict when pulling from you. Paolo
diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c index 0625bf2353c2..f5bb235bbb59 100644 --- a/arch/arm64/kvm/mmu.c +++ b/arch/arm64/kvm/mmu.c @@ -80,6 +80,7 @@ static bool memslot_is_logging(struct kvm_memory_slot *memslot) */ void kvm_flush_remote_tlbs(struct kvm *kvm) { + ++kvm->stat.generic.remote_tlb_flush_requests; kvm_call_hyp(__kvm_tlb_flush_vmid, &kvm->arch.mmu); } diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h index d447b21cdd73..27cb69e564cb 100644 --- a/include/linux/kvm_host.h +++ b/include/linux/kvm_host.h @@ -1444,7 +1444,8 @@ struct _kvm_stats_desc { KVM_STATS_BASE_POW10, -9, sz) #define KVM_GENERIC_VM_STATS() \ - STATS_DESC_COUNTER(VM_GENERIC, remote_tlb_flush) + STATS_DESC_COUNTER(VM_GENERIC, remote_tlb_flush), \ + STATS_DESC_COUNTER(VM_GENERIC, remote_tlb_flush_requests) #define KVM_GENERIC_VCPU_STATS() \ STATS_DESC_COUNTER(VCPU_GENERIC, halt_successful_poll), \ diff --git a/include/linux/kvm_types.h b/include/linux/kvm_types.h index de7fb5f364d8..2237abb93ccd 100644 --- a/include/linux/kvm_types.h +++ b/include/linux/kvm_types.h @@ -80,6 +80,7 @@ struct kvm_mmu_memory_cache { struct kvm_vm_stat_generic { u64 remote_tlb_flush; + u64 remote_tlb_flush_requests; }; struct kvm_vcpu_stat_generic { diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c index 3e67c93ca403..26b77ab98a04 100644 --- a/virt/kvm/kvm_main.c +++ b/virt/kvm/kvm_main.c @@ -318,6 +318,7 @@ void kvm_flush_remote_tlbs(struct kvm *kvm) */ long dirty_count = smp_load_acquire(&kvm->tlbs_dirty); + ++kvm->stat.generic.remote_tlb_flush_requests; /* * We want to publish modifications to the page tables before reading * mode. Pairs with a memory barrier in arch-specific code.
Add a new stat that counts the number of times a remote TLB flush is requested, regardless of whether it kicks vCPUs out of guest mode. This allows us to look at how often flushes are initiated. Original-by: David Matlack <dmatlack@google.com> Signed-off-by: Jing Zhang <jingzhangos@google.com> --- arch/arm64/kvm/mmu.c | 1 + include/linux/kvm_host.h | 3 ++- include/linux/kvm_types.h | 1 + virt/kvm/kvm_main.c | 1 + 4 files changed, 5 insertions(+), 1 deletion(-) base-commit: a3e0b8bd99ab098514bde2434301fa6fde040da2