* [Patch v2 1/2] KVM: remember position in kvm->vcpus array
2019-11-07 12:53 [PATCH v2 0/2] KVM: deliver IOAPIC scan request only to the target vCPUs Nitesh Narayan Lal
@ 2019-11-07 12:53 ` Nitesh Narayan Lal
2019-11-07 12:53 ` [Patch v2 2/2] KVM: x86: deliver KVM IOAPIC scan request to target vCPUs Nitesh Narayan Lal
1 sibling, 0 replies; 5+ messages in thread
From: Nitesh Narayan Lal @ 2019-11-07 12:53 UTC (permalink / raw)
To: kvm, linux-kernel, pbonzini, mtosatti, rkrcmar, vkuznets,
sean.j.christopherson, wanpengli, jmattson, joro
From: Radim Krčmář <rkrcmar@redhat.com>
Fetching an index for any vcpu in kvm->vcpus array by traversing
the entire array everytime is costly.
This patch remembers the position of each vcpu in kvm->vcpus array
by storing it in vcpus_idx under kvm_vcpu structure.
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Nitesh Narayan Lal <nitesh@redhat.com>
---
include/linux/kvm_host.h | 11 +++--------
virt/kvm/kvm_main.c | 5 +++--
2 files changed, 6 insertions(+), 10 deletions(-)
diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index 719fc3e..22068fe 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -266,7 +266,8 @@ struct kvm_vcpu {
struct preempt_notifier preempt_notifier;
#endif
int cpu;
- int vcpu_id;
+ int vcpu_id; /* id given by userspace at creation */
+ int vcpu_idx; /* index in kvm->vcpus array */
int srcu_idx;
int mode;
u64 requests;
@@ -571,13 +572,7 @@ static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
static inline int kvm_vcpu_get_idx(struct kvm_vcpu *vcpu)
{
- struct kvm_vcpu *tmp;
- int idx;
-
- kvm_for_each_vcpu(idx, tmp, vcpu->kvm)
- if (tmp == vcpu)
- return idx;
- BUG();
+ return vcpu->vcpu_idx;
}
#define kvm_for_each_memslot(memslot, slots) \
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index 67ef3f2..72132fc 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -2673,7 +2673,8 @@ static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
goto unlock_vcpu_destroy;
}
- BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]);
+ vcpu->vcpu_idx = atomic_read(&kvm->online_vcpus);
+ BUG_ON(kvm->vcpus[vcpu->vcpu_idx]);
/* Now it's all set up, let userspace reach it */
kvm_get_kvm(kvm);
@@ -2683,7 +2684,7 @@ static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
goto unlock_vcpu_destroy;
}
- kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu;
+ kvm->vcpus[vcpu->vcpu_idx] = vcpu;
/*
* Pairs with smp_rmb() in kvm_get_vcpu. Write kvm->vcpus
--
1.8.3.1
^ permalink raw reply related [flat|nested] 5+ messages in thread* [Patch v2 2/2] KVM: x86: deliver KVM IOAPIC scan request to target vCPUs
2019-11-07 12:53 [PATCH v2 0/2] KVM: deliver IOAPIC scan request only to the target vCPUs Nitesh Narayan Lal
2019-11-07 12:53 ` [Patch v2 1/2] KVM: remember position in kvm->vcpus array Nitesh Narayan Lal
@ 2019-11-07 12:53 ` Nitesh Narayan Lal
2019-11-15 10:26 ` Paolo Bonzini
1 sibling, 1 reply; 5+ messages in thread
From: Nitesh Narayan Lal @ 2019-11-07 12:53 UTC (permalink / raw)
To: kvm, linux-kernel, pbonzini, mtosatti, rkrcmar, vkuznets,
sean.j.christopherson, wanpengli, jmattson, joro
In IOAPIC fixed delivery mode instead of flushing the scan
requests to all vCPUs, we should only send the requests to
vCPUs specified within the destination field.
This patch introduces kvm_get_dest_vcpus_mask() API which
retrieves an array of target vCPUs by using
kvm_apic_map_get_dest_lapic() and then based on the
vcpus_idx, it sets the bit in a bitmap. However, if the above
fails kvm_get_dest_vcpus_mask() finds the target vCPUs by
traversing all available vCPUs. Followed by setting the
bits in the bitmap.
If we had different vCPUs in the previous request for the
same redirection table entry then bits corresponding to
these vCPUs are also set. This to done to keep
ioapic_handled_vectors synchronized.
This bitmap is then eventually passed on to
kvm_make_vcpus_request_mask() to generate a masked request
only for the target vCPUs.
This would enable us to reduce the latency overhead on isolated
vCPUs caused by the IPI to process due to KVM_REQ_IOAPIC_SCAN.
Suggested-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Nitesh Narayan Lal <nitesh@redhat.com>
---
arch/x86/include/asm/kvm_host.h | 2 ++
arch/x86/kvm/ioapic.c | 33 +++++++++++++++++++++++++++++--
arch/x86/kvm/lapic.c | 44 +++++++++++++++++++++++++++++++++++++++++
arch/x86/kvm/lapic.h | 3 +++
arch/x86/kvm/x86.c | 14 +++++++++++++
include/linux/kvm_host.h | 2 ++
6 files changed, 96 insertions(+), 2 deletions(-)
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 24d6598..b2aca6d 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -1571,6 +1571,8 @@ int kvm_pv_send_ipi(struct kvm *kvm, unsigned long ipi_bitmap_low,
void kvm_make_mclock_inprogress_request(struct kvm *kvm);
void kvm_make_scan_ioapic_request(struct kvm *kvm);
+void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
+ unsigned long *vcpu_bitmap);
void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
struct kvm_async_pf *work);
diff --git a/arch/x86/kvm/ioapic.c b/arch/x86/kvm/ioapic.c
index d859ae8..c8d0a83 100644
--- a/arch/x86/kvm/ioapic.c
+++ b/arch/x86/kvm/ioapic.c
@@ -271,8 +271,9 @@ static void ioapic_write_indirect(struct kvm_ioapic *ioapic, u32 val)
{
unsigned index;
bool mask_before, mask_after;
- int old_remote_irr, old_delivery_status;
union kvm_ioapic_redirect_entry *e;
+ unsigned long vcpu_bitmap;
+ int old_remote_irr, old_delivery_status, old_dest_id, old_dest_mode;
switch (ioapic->ioregsel) {
case IOAPIC_REG_VERSION:
@@ -296,6 +297,8 @@ static void ioapic_write_indirect(struct kvm_ioapic *ioapic, u32 val)
/* Preserve read-only fields */
old_remote_irr = e->fields.remote_irr;
old_delivery_status = e->fields.delivery_status;
+ old_dest_id = e->fields.dest_id;
+ old_dest_mode = e->fields.dest_mode;
if (ioapic->ioregsel & 1) {
e->bits &= 0xffffffff;
e->bits |= (u64) val << 32;
@@ -321,7 +324,33 @@ static void ioapic_write_indirect(struct kvm_ioapic *ioapic, u32 val)
if (e->fields.trig_mode == IOAPIC_LEVEL_TRIG
&& ioapic->irr & (1 << index))
ioapic_service(ioapic, index, false);
- kvm_make_scan_ioapic_request(ioapic->kvm);
+ if (e->fields.delivery_mode == APIC_DM_FIXED) {
+ struct kvm_lapic_irq irq;
+
+ irq.shorthand = 0;
+ irq.vector = e->fields.vector;
+ irq.delivery_mode = e->fields.delivery_mode << 8;
+ irq.dest_id = e->fields.dest_id;
+ irq.dest_mode = e->fields.dest_mode;
+ kvm_get_dest_vcpus_mask(ioapic->kvm, &irq,
+ &vcpu_bitmap);
+ if (old_dest_mode != e->fields.dest_mode ||
+ old_dest_id != e->fields.dest_id) {
+ /*
+ * Update vcpu_bitmap with vcpus specified in
+ * the previous request as well. This is done to
+ * keep ioapic_handled_vectors synchronized.
+ */
+ irq.dest_id = old_dest_id;
+ irq.dest_mode = old_dest_mode;
+ kvm_get_dest_vcpus_mask(ioapic->kvm, &irq,
+ &vcpu_bitmap);
+ }
+ kvm_make_scan_ioapic_request_mask(ioapic->kvm,
+ &vcpu_bitmap);
+ } else {
+ kvm_make_scan_ioapic_request(ioapic->kvm);
+ }
break;
}
}
diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
index b29d00b..411d675 100644
--- a/arch/x86/kvm/lapic.c
+++ b/arch/x86/kvm/lapic.c
@@ -1124,6 +1124,50 @@ static int __apic_accept_irq(struct kvm_lapic *apic, int delivery_mode,
return result;
}
+/*
+ * This routine identifies the destination vcpus mask meant to receive the
+ * IOAPIC interrupts. It either uses kvm_apic_map_get_dest_lapic() to find
+ * out the destination vcpus array and set the bitmap or it traverses to
+ * each available vcpu to identify the same.
+ */
+void kvm_get_dest_vcpus_mask(struct kvm *kvm, struct kvm_lapic_irq *irq,
+ unsigned long *vcpu_bitmap)
+{
+ struct kvm_lapic **dest_vcpu = NULL;
+ struct kvm_lapic *src = NULL;
+ struct kvm_apic_map *map;
+ struct kvm_vcpu *vcpu;
+ unsigned long bitmap;
+ int i, vcpu_idx;
+ bool ret;
+
+ rcu_read_lock();
+ map = rcu_dereference(kvm->arch.apic_map);
+
+ ret = kvm_apic_map_get_dest_lapic(kvm, &src, irq, map, &dest_vcpu,
+ &bitmap);
+ if (ret) {
+ for_each_set_bit(i, &bitmap, 16) {
+ if (!dest_vcpu[i])
+ continue;
+ vcpu_idx = dest_vcpu[i]->vcpu->vcpu_idx;
+ __set_bit(vcpu_idx, vcpu_bitmap);
+ }
+ } else {
+ kvm_for_each_vcpu(i, vcpu, kvm) {
+ if (!kvm_apic_present(vcpu))
+ continue;
+ if (!kvm_apic_match_dest(vcpu, NULL,
+ irq->delivery_mode,
+ irq->dest_id,
+ irq->dest_mode))
+ continue;
+ __set_bit(i, vcpu_bitmap);
+ }
+ }
+ rcu_read_unlock();
+}
+
int kvm_apic_compare_prio(struct kvm_vcpu *vcpu1, struct kvm_vcpu *vcpu2)
{
return vcpu1->arch.apic_arb_prio - vcpu2->arch.apic_arb_prio;
diff --git a/arch/x86/kvm/lapic.h b/arch/x86/kvm/lapic.h
index 1f501485..49b0c6c 100644
--- a/arch/x86/kvm/lapic.h
+++ b/arch/x86/kvm/lapic.h
@@ -226,6 +226,9 @@ static inline int kvm_lapic_latched_init(struct kvm_vcpu *vcpu)
void kvm_wait_lapic_expire(struct kvm_vcpu *vcpu);
+void kvm_get_dest_vcpus_mask(struct kvm *kvm, struct kvm_lapic_irq *irq,
+ unsigned long *vcpu_bitmap);
+
bool kvm_intr_is_single_vcpu_fast(struct kvm *kvm, struct kvm_lapic_irq *irq,
struct kvm_vcpu **dest_vcpu);
int kvm_vector_to_index(u32 vector, u32 dest_vcpus,
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index ff395f8..445e1c7 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -7838,6 +7838,20 @@ static void process_smi(struct kvm_vcpu *vcpu)
kvm_make_request(KVM_REQ_EVENT, vcpu);
}
+void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
+ unsigned long *vcpu_bitmap)
+{
+ cpumask_var_t cpus;
+ bool called;
+
+ zalloc_cpumask_var(&cpus, GFP_ATOMIC);
+
+ called = kvm_make_vcpus_request_mask(kvm, KVM_REQ_SCAN_IOAPIC,
+ vcpu_bitmap, cpus);
+
+ free_cpumask_var(cpus);
+}
+
void kvm_make_scan_ioapic_request(struct kvm *kvm)
{
kvm_make_all_cpus_request(kvm, KVM_REQ_SCAN_IOAPIC);
diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index 22068fe..09bccd8 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -786,6 +786,8 @@ int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
unsigned long *vcpu_bitmap, cpumask_var_t tmp);
bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
+bool kvm_make_cpus_request_mask(struct kvm *kvm, unsigned int req,
+ unsigned long *vcpu_bitmap);
long kvm_arch_dev_ioctl(struct file *filp,
unsigned int ioctl, unsigned long arg);
--
1.8.3.1
^ permalink raw reply related [flat|nested] 5+ messages in thread