Skip to contentSkip to conversation

optimize the downtime for vfio migration

PATCH Superseded Longpeng(Mike) 2 participants 18 messages Sep 20, 2021 – Oct 8, 2021

v3 of 6 Superseded Latest is v6 →

lore
Share & tools
lore
#0 lore raw

v3 cover optimize the downtime for vfio migration

Hi guys,

In vfio migration resume phase, the cost would increase if the vfio device has more unmasked vectors. We try to optimize it in this series.

You can see the commit message in PATCH 9 for details.

Patch 1-5 are simple cleanups and fixup. Patch 6-8 are the preparations for the optimization. Patch 9 optimizes the vfio msix setup path.

Changes v2->v3:
- fix two errors [Longpeng]

Changes v1->v2:
- fix several typos and grammatical errors [Alex, Philippe]
- split fixups and cleanups into separate patches [Alex, Philippe]
- introduce kvm_irqchip_add_deferred_msi_route to
minimize code changes [Alex]
- enable the optimization in msi setup path [Alex]

Longpeng (Mike) (9):
vfio: simplify the conditional statements in vfio_msi_enable
vfio: move re-enabling INTX out of the common helper
vfio: simplify the failure path in vfio_msi_enable
msix: simplify the conditional in msix_set/unset_vector_notifiers
msix: reset poll_notifier to NULL if fail to set notifiers
kvm: irqchip: extract kvm_irqchip_add_deferred_msi_route
vfio: add infrastructure to commit the deferred kvm routing Revert "vfio: Avoid disabling and enabling vectors repeatedly in VFIO
migration"
vfio: defer to commit kvm irq routing when enable msi/msix

 accel/kvm/kvm-all.c  |  15 +++++-
 hw/pci/msix.c        |   7 ++-
 hw/vfio/pci.c        | 126 ++++++++++++++++++++++++++++---------------
 hw/vfio/pci.h        |   1 +
 include/sysemu/kvm.h |   6 +++
 5 files changed, 106 insertions(+), 49 deletions(-)
Signature · 1 line
2.23.0
#1 lore raw

PATCH v3 1/9 vfio: simplify the conditional statements in vfio_msi_enable

It's unnecessary to test against the specific return value of VFIO_DEVICE_SET_IRQS, since any positive return is an error indicating the number of vectors we should retry with.

---
 hw/vfio/pci.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
hw/vfio/pci.c +2 -2
index e1ea1d8a23..f7a3a13fb0 100644
@@ -650,7 +650,7 @@ retry:
     if (ret) {
         if (ret < 0) {
             error_report("vfio: Error: Failed to setup MSI fds: %m");
-        } else if (ret != vdev->nr_vectors) {
+        } else {
             error_report("vfio: Error: Failed to enable %d "
                          "MSI vectors, retry with %d", vdev->nr_vectors, ret);
         }
@@ -668,7 +668,7 @@ retry:
         g_free(vdev->msi_vectors);
         vdev->msi_vectors = NULL;
 
-        if (ret > 0 && ret != vdev->nr_vectors) {
+        if (ret > 0) {
             vdev->nr_vectors = ret;
             goto retry;
         }
Signature · 1 line
2.23.0
#2 lore raw

PATCH v3 2/9 vfio: move re-enabling INTX out of the common helper

In reply to Longpeng(Mike)

Move re-enabling INTX out, and the callers should decide to re-enable it or not.

---
 hw/vfio/pci.c | 17 +++++++++++------
 1 file changed, 11 insertions(+), 6 deletions(-)
hw/vfio/pci.c +11 -6
index f7a3a13fb0..1e6797fd4b 100644
@@ -690,7 +690,6 @@ retry:
 
 static void vfio_msi_disable_common(VFIOPCIDevice *vdev)
 {
-    Error *err = NULL;
     int i;
 
     for (i = 0; i < vdev->nr_vectors; i++) {
@@ -709,15 +708,11 @@ static void vfio_msi_disable_common(VFIOPCIDevice *vdev)
     vdev->msi_vectors = NULL;
     vdev->nr_vectors = 0;
     vdev->interrupt = VFIO_INT_NONE;
-
-    vfio_intx_enable(vdev, &err);
-    if (err) {
-        error_reportf_err(err, VFIO_MSG_PREFIX, vdev->vbasedev.name);
-    }
 }
 
 static void vfio_msix_disable(VFIOPCIDevice *vdev)
 {
+    Error *err = NULL;
     int i;
 
     msix_unset_vector_notifiers(&vdev->pdev);
@@ -738,6 +733,10 @@ static void vfio_msix_disable(VFIOPCIDevice *vdev)
     }
 
     vfio_msi_disable_common(vdev);
+    vfio_intx_enable(vdev, &err);
+    if (err) {
+        error_reportf_err(err, VFIO_MSG_PREFIX, vdev->vbasedev.name);
+    }
 
     memset(vdev->msix->pending, 0,
            BITS_TO_LONGS(vdev->msix->entries) * sizeof(unsigned long));
@@ -747,8 +746,14 @@ static void vfio_msix_disable(VFIOPCIDevice *vdev)
 
 static void vfio_msi_disable(VFIOPCIDevice *vdev)
 {
+    Error *err = NULL;
+
     vfio_disable_irqindex(&vdev->vbasedev, VFIO_PCI_MSI_IRQ_INDEX);
     vfio_msi_disable_common(vdev);
+    vfio_intx_enable(vdev, &err);
+    if (err) {
+        error_reportf_err(err, VFIO_MSG_PREFIX, vdev->vbasedev.name);
+    }
 
     trace_vfio_msi_disable(vdev->vbasedev.name);
 }
Signature · 1 line
2.23.0
#3 lore raw

PATCH v3 3/9 vfio: simplify the failure path in vfio_msi_enable

In reply to Longpeng(Mike)

Use vfio_msi_disable_common to simplify the error handling in vfio_msi_enable.

---
 hw/vfio/pci.c | 16 ++--------------
 1 file changed, 2 insertions(+), 14 deletions(-)
hw/vfio/pci.c +2 -14
index 1e6797fd4b..8e97ca93cf 100644
@@ -47,6 +47,7 @@
 
 static void vfio_disable_interrupts(VFIOPCIDevice *vdev);
 static void vfio_mmap_set_enabled(VFIOPCIDevice *vdev, bool enabled);
+static void vfio_msi_disable_common(VFIOPCIDevice *vdev);
 
 /*
  * Disabling BAR mmaping can be slow, but toggling it around INTx can
@@ -655,24 +656,12 @@ retry:
                          "MSI vectors, retry with %d", vdev->nr_vectors, ret);
         }
 
-        for (i = 0; i < vdev->nr_vectors; i++) {
-            VFIOMSIVector *vector = &vdev->msi_vectors[i];
-            if (vector->virq >= 0) {
-                vfio_remove_kvm_msi_virq(vector);
-            }
-            qemu_set_fd_handler(event_notifier_get_fd(&vector->interrupt),
-                                NULL, NULL, NULL);
-            event_notifier_cleanup(&vector->interrupt);
-        }
-
-        g_free(vdev->msi_vectors);
-        vdev->msi_vectors = NULL;
+        vfio_msi_disable_common(vdev);
 
         if (ret > 0) {
             vdev->nr_vectors = ret;
             goto retry;
         }
-        vdev->nr_vectors = 0;
 
         /*
          * Failing to setup MSI doesn't really fall within any specification.
@@ -680,7 +669,6 @@ retry:
          * out to fall back to INTx for this device.
          */
         error_report("vfio: Error: Failed to enable MSI");
-        vdev->interrupt = VFIO_INT_NONE;
 
         return;
     }
Signature · 1 line
2.23.0
#4 lore raw

PATCH v3 4/9 msix: simplify the conditional in msix_set/unset_vector_notifiers

In reply to Longpeng(Mike)

'msix_function_masked' is synchronized with the device's config, we can use it to replace the complex conditional statementis in msix_set/unset_vector_notifiers.

---
 hw/pci/msix.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
hw/pci/msix.c +2 -4
index ae9331cd0b..67682289af 100644
@@ -592,8 +592,7 @@ int msix_set_vector_notifiers(PCIDevice *dev,
     dev->msix_vector_release_notifier = release_notifier;
     dev->msix_vector_poll_notifier = poll_notifier;
 
-    if ((dev->config[dev->msix_cap + MSIX_CONTROL_OFFSET] &
-        (MSIX_ENABLE_MASK | MSIX_MASKALL_MASK)) == MSIX_ENABLE_MASK) {
+    if (!dev->msix_function_masked) {
         for (vector = 0; vector < dev->msix_entries_nr; vector++) {
             ret = msix_set_notifier_for_vector(dev, vector);
             if (ret < 0) {
@@ -622,8 +621,7 @@ void msix_unset_vector_notifiers(PCIDevice *dev)
     assert(dev->msix_vector_use_notifier &&
            dev->msix_vector_release_notifier);
 
-    if ((dev->config[dev->msix_cap + MSIX_CONTROL_OFFSET] &
-        (MSIX_ENABLE_MASK | MSIX_MASKALL_MASK)) == MSIX_ENABLE_MASK) {
+    if (!dev->msix_function_masked) {
         for (vector = 0; vector < dev->msix_entries_nr; vector++) {
             msix_unset_notifier_for_vector(dev, vector);
         }
Signature · 1 line
2.23.0
#5 lore raw

Re: [PATCH v3 4/9] msix: simplify the conditional in msix_set/unset_vector_notifiers

Quoted 24 lines · On Tue, 21 Sep 2021 07:01:57 +0800 "Longpeng(Mike)" <[email protected]>…

'msix_function_masked' is synchronized with the device's config, we can use it to replace the complex conditional statementis in msix_set/unset_vector_notifiers.

---
 hw/pci/msix.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
hw/pci/msix.c +1 -2
index ae9331cd0b..67682289af 100644
@@ -592,8 +592,7 @@ int msix_set_vector_notifiers(PCIDevice *dev,
     dev->msix_vector_release_notifier = release_notifier;
     dev->msix_vector_poll_notifier = poll_notifier;
 
-    if ((dev->config[dev->msix_cap + MSIX_CONTROL_OFFSET] &
-        (MSIX_ENABLE_MASK | MSIX_MASKALL_MASK)) == MSIX_ENABLE_MASK) {
+    if (!dev->msix_function_masked) {
         for (vector = 0; vector < dev->msix_entries_nr; vector++) {
             ret = msix_set_notifier_for_vector(dev, vector);
             if (ret < 0) {

hw/pci/msix.c void msix_unset_vector_notifiers(PCIDevice *dev) · line 627

Quoted 9 lines
hw/pci/msix.c +1 -2
@@ -622,8 +621,7 @@ void msix_unset_vector_notifiers(PCIDevice *dev)
     assert(dev->msix_vector_use_notifier &&
            dev->msix_vector_release_notifier);
 
-    if ((dev->config[dev->msix_cap + MSIX_CONTROL_OFFSET] &
-        (MSIX_ENABLE_MASK | MSIX_MASKALL_MASK)) == MSIX_ENABLE_MASK) {
+    if (!dev->msix_function_masked) {
         for (vector = 0; vector < dev->msix_entries_nr; vector++) {
             msix_unset_notifier_for_vector(dev, vector);
         }

This appears to be a cleanup that's not required for the functionality of this series. I'd suggest proposing it separately. Same for the patch 5/9 in this series. If it makes a functional difference it should be described in the commit log. Thanks,

Alex

#6 lore raw

RE: [PATCH v3 4/9] msix: simplify the conditional in msix_set/unset_vector_notifiers

Quoted 43 lines

-----Original Message-----
From: Alex Williamson [mailto:[email protected]]
Sent: Saturday, October 2, 2021 7:04 AM
To: Longpeng (Mike, Cloud Infrastructure Service Product Dept.) <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; Gonglei (Arei) <[email protected]>; chenjiashang <[email protected]>
Subject: Re: [PATCH v3 4/9] msix: simplify the conditional in msix_set/unset_vector_notifiers

Quoted 27 lines · On Tue, 21 Sep 2021 07:01:57 +0800 "Longpeng(Mike)" <[email protected]>…

'msix_function_masked' is synchronized with the device's config, we can use it to replace the complex conditional statementis in msix_set/unset_vector_notifiers.

---
 hw/pci/msix.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
hw/pci/msix.c +1 -2
index ae9331cd0b..67682289af 100644
@@ -592,8 +592,7 @@ int msix_set_vector_notifiers(PCIDevice *dev,
     dev->msix_vector_release_notifier = release_notifier;
     dev->msix_vector_poll_notifier = poll_notifier;
 
-    if ((dev->config[dev->msix_cap + MSIX_CONTROL_OFFSET] &
-        (MSIX_ENABLE_MASK | MSIX_MASKALL_MASK)) == MSIX_ENABLE_MASK) {
+    if (!dev->msix_function_masked) {
         for (vector = 0; vector < dev->msix_entries_nr; vector++) {
             ret = msix_set_notifier_for_vector(dev, vector);
             if (ret < 0) {
@@ -622,8 +621,7 @@ void msix_unset_vector_notifiers(PCIDevice *dev)
     assert(dev->msix_vector_use_notifier &&
            dev->msix_vector_release_notifier);
 
Quoted 12 lines

hw/pci/msix.c void msix_unset_vector_notifiers(PCIDevice *dev) · line 627

Quoted 6 lines
hw/pci/msix.c +1 -2
-    if ((dev->config[dev->msix_cap + MSIX_CONTROL_OFFSET] &
-        (MSIX_ENABLE_MASK | MSIX_MASKALL_MASK)) == MSIX_ENABLE_MASK) {
+    if (!dev->msix_function_masked) {
         for (vector = 0; vector < dev->msix_entries_nr; vector++) {
             msix_unset_notifier_for_vector(dev, vector);
         }

This appears to be a cleanup that's not required for the functionality of this series. I'd suggest proposing it separately. Same for the patch 5/9 in this series. If it makes a functional difference it should be described in the commit log. Thanks,

OK, will remove these two patches in the v4, thanks.

Quoted 1 line

Alex

#7 lore raw

PATCH v3 5/9 msix: reset poll_notifier to NULL if fail to set notifiers

In reply to Longpeng(Mike)

'msix_vector_poll_notifier' should be reset to NULL in the error path in msix_set_vector_notifiers().

---
 hw/pci/msix.c | 1 +
 1 file changed, 1 insertion(+)
hw/pci/msix.c +1 -0
index 67682289af..805770942b 100644
@@ -611,6 +611,7 @@ undo:
     }
     dev->msix_vector_use_notifier = NULL;
     dev->msix_vector_release_notifier = NULL;
+    dev->msix_vector_poll_notifier = NULL;
     return ret;
 }
 
Signature · 1 line
2.23.0
#8 lore raw

PATCH v3 6/9 kvm: irqchip: extract kvm_irqchip_add_deferred_msi_route

In reply to Longpeng(Mike)

Extract a common helper that add MSI route for specific vector but does not commit immediately.

---
 accel/kvm/kvm-all.c  | 15 +++++++++++++--
 include/sysemu/kvm.h |  6 ++++++
 2 files changed, 19 insertions(+), 2 deletions(-)
accel/kvm/kvm-all.c +13 -2
index e5b10dd129..d603afc44c 100644
@@ -1950,7 +1950,7 @@ int kvm_irqchip_send_msi(KVMState *s, MSIMessage msg)
     return kvm_set_irq(s, route->kroute.gsi, 1);
 }
 
-int kvm_irqchip_add_msi_route(KVMState *s, int vector, PCIDevice *dev)
+int kvm_irqchip_add_deferred_msi_route(KVMState *s, int vector, PCIDevice *dev)
 {
     struct kvm_irq_routing_entry kroute = {};
     int virq;
@@ -1993,7 +1993,18 @@ int kvm_irqchip_add_msi_route(KVMState *s, int vector, PCIDevice *dev)
 
     kvm_add_routing_entry(s, &kroute);
     kvm_arch_add_msi_route_post(&kroute, vector, dev);
-    kvm_irqchip_commit_routes(s);
+
+    return virq;
+}
+
+int kvm_irqchip_add_msi_route(KVMState *s, int vector, PCIDevice *dev)
+{
+    int virq;
+
+    virq = kvm_irqchip_add_deferred_msi_route(s, vector, dev);
+    if (virq >= 0) {
+        kvm_irqchip_commit_routes(s);
+    }
 
     return virq;
 }
include/sysemu/kvm.h +6 -0
index a1ab1ee12d..8de0d9a715 100644
@@ -476,6 +476,12 @@ void kvm_init_cpu_signals(CPUState *cpu);
  * @return: virq (>=0) when success, errno (<0) when failed.
  */
 int kvm_irqchip_add_msi_route(KVMState *s, int vector, PCIDevice *dev);
+/**
+ * Add MSI route for specific vector but does not commit to KVM
+ * immediately
+ */
+int kvm_irqchip_add_deferred_msi_route(KVMState *s, int vector,
+                                       PCIDevice *dev);
 int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg,
                                  PCIDevice *dev);
 void kvm_irqchip_commit_routes(KVMState *s);
Signature · 1 line
2.23.0
#9 lore raw

PATCH v3 7/9 vfio: add infrastructure to commit the deferred kvm routing

In reply to Longpeng(Mike)

'defer_kvm_irq_routing' indicates whether we should defer to commit the kvm routing.

---
 hw/vfio/pci.c | 43 ++++++++++++++++++++++++++++++++++++++++++-
 hw/vfio/pci.h |  1 +
 2 files changed, 43 insertions(+), 1 deletion(-)
hw/vfio/pci.c +42 -1
index 8e97ca93cf..8fe238b11d 100644
@@ -423,12 +423,24 @@ static void vfio_add_kvm_msi_virq(VFIOPCIDevice *vdev, VFIOMSIVector *vector,
         return;
     }
 
-    virq = kvm_irqchip_add_msi_route(kvm_state, vector_n, &vdev->pdev);
+    virq = kvm_irqchip_add_deferred_msi_route(kvm_state, vector_n, &vdev->pdev);
     if (virq < 0) {
         event_notifier_cleanup(&vector->kvm_interrupt);
         return;
     }
 
+    if (vdev->defer_kvm_irq_routing) {
+        /*
+         * Hold the allocated virq in vector->virq temporarily, will
+         * reset it to -1 when we fail to add the corresponding irqfd
+         * in vfio_commit_kvm_msi_virq().
+         */
+        vector->virq = virq;
+        return;
+    }
+
+    kvm_irqchip_commit_routes(kvm_state);
+
     if (kvm_irqchip_add_irqfd_notifier_gsi(kvm_state, &vector->kvm_interrupt,
                                        NULL, virq) < 0) {
         kvm_irqchip_release_virq(kvm_state, virq);
@@ -567,6 +579,35 @@ static void vfio_msix_vector_release(PCIDevice *pdev, unsigned int nr)
     }
 }
 
+/* TODO: invoked when enclabe msi/msix vectors */
+static __attribute__((unused)) void vfio_commit_kvm_msi_virq(VFIOPCIDevice *vdev)
+{
+    int i;
+    VFIOMSIVector *vector;
+
+    if (!vdev->defer_kvm_irq_routing || !vdev->nr_vectors) {
+        return;
+    }
+
+    kvm_irqchip_commit_routes(kvm_state);
+
+    for (i = 0; i < vdev->nr_vectors; i++) {
+        vector = &vdev->msi_vectors[i];
+
+        if (!vector->use || vector->virq < 0) {
+            continue;
+        }
+
+        if (kvm_irqchip_add_irqfd_notifier_gsi(kvm_state,
+                                               &vector->kvm_interrupt,
+                                               NULL, vector->virq) < 0) {
+            kvm_irqchip_release_virq(kvm_state, vector->virq);
+            event_notifier_cleanup(&vector->kvm_interrupt);
+            vector->virq = -1;
+        }
+    }
+}
+
 static void vfio_msix_enable(VFIOPCIDevice *vdev)
 {
     PCIDevice *pdev = &vdev->pdev;
hw/vfio/pci.h +1 -0
index 64777516d1..d3c5177d37 100644
@@ -171,6 +171,7 @@ struct VFIOPCIDevice {
     bool no_kvm_ioeventfd;
     bool no_vfio_ioeventfd;
     bool enable_ramfb;
+    bool defer_kvm_irq_routing;
     VFIODisplay *dpy;
     Notifier irqchip_change_notifier;
 };
Signature · 1 line
2.23.0
#10 lore raw

Re: [PATCH v3 7/9] vfio: add infrastructure to commit the deferred kvm routing

Quoted 19 lines · On Tue, 21 Sep 2021 07:02:00 +0800 "Longpeng(Mike)" <[email protected]>…

'defer_kvm_irq_routing' indicates whether we should defer to commit the kvm routing.

---
 hw/vfio/pci.c | 43 ++++++++++++++++++++++++++++++++++++++++++-
 hw/vfio/pci.h |  1 +
 2 files changed, 43 insertions(+), 1 deletion(-)
hw/vfio/pci.c +1 -1
index 8e97ca93cf..8fe238b11d 100644
@@ -423,12 +423,24 @@ static void vfio_add_kvm_msi_virq(VFIOPCIDevice *vdev, VFIOMSIVector *vector,
         return;
     }
 
-    virq = kvm_irqchip_add_msi_route(kvm_state, vector_n, &vdev->pdev);
+    virq = kvm_irqchip_add_deferred_msi_route(kvm_state, vector_n, &vdev->pdev);

hw/vfio/pci.c static void vfio_add_kvm_msi_virq(VFIOPCIDevice *vdev, VFIOMSIVector *vector, · line 436

Quoted 10 lines
hw/vfio/pci.c +5 -0
     if (virq < 0) {
         event_notifier_cleanup(&vector->kvm_interrupt);
         return;
     }
 
+    if (vdev->defer_kvm_irq_routing) {
+        /*
+         * Hold the allocated virq in vector->virq temporarily, will
+         * reset it to -1 when we fail to add the corresponding irqfd
+         * in vfio_commit_kvm_msi_virq().

s/when/if/

hw/vfio/pci.c static void vfio_add_kvm_msi_virq(VFIOPCIDevice *vdev, VFIOMSIVector *vector, · line 438

Quoted 2 lines
hw/vfio/pci.c +2 -0
+         */
+        vector->virq = virq;

Do we need to make this unique to the deferred case or could we use vector->virq directly and fill it with -1 on all error paths like we do on a failure in vfio_commit_kvm_msi_virq()?

hw/vfio/pci.c static void vfio_msix_vector_release(PCIDevice *pdev, unsigned int nr) · line 582

Quoted 13 lines
hw/vfio/pci.c +6 -0
+        return;
+    }
+
+    kvm_irqchip_commit_routes(kvm_state);
+
     if (kvm_irqchip_add_irqfd_notifier_gsi(kvm_state, &vector->kvm_interrupt,
                                        NULL, virq) < 0) {
         kvm_irqchip_release_virq(kvm_state, virq);
@@ -567,6 +579,35 @@ static void vfio_msix_vector_release(PCIDevice *pdev, unsigned int nr)
     }
 }
 
+/* TODO: invoked when enclabe msi/msix vectors */

"enclabe"? Is this meant to be "enable"?

hw/vfio/pci.c static void vfio_msix_vector_release(PCIDevice *pdev, unsigned int nr) · line 583

Quoted 1 line
hw/vfio/pci.c +1 -0
+static __attribute__((unused)) void vfio_commit_kvm_msi_virq(VFIOPCIDevice *vdev)

I'd move this function, if not this entire change, to patch 9 rather than adding these attributes for an unused function. Thanks,

Alex

Quoted 42 lines
hw/vfio/pci.c +27 -0
+{
+    int i;
+    VFIOMSIVector *vector;
+
+    if (!vdev->defer_kvm_irq_routing || !vdev->nr_vectors) {
+        return;
+    }
+
+    kvm_irqchip_commit_routes(kvm_state);
+
+    for (i = 0; i < vdev->nr_vectors; i++) {
+        vector = &vdev->msi_vectors[i];
+
+        if (!vector->use || vector->virq < 0) {
+            continue;
+        }
+
+        if (kvm_irqchip_add_irqfd_notifier_gsi(kvm_state,
+                                               &vector->kvm_interrupt,
+                                               NULL, vector->virq) < 0) {
+            kvm_irqchip_release_virq(kvm_state, vector->virq);
+            event_notifier_cleanup(&vector->kvm_interrupt);
+            vector->virq = -1;
+        }
+    }
+}
+
 static void vfio_msix_enable(VFIOPCIDevice *vdev)
 {
     PCIDevice *pdev = &vdev->pdev;
hw/vfio/pci.h +1 -0
index 64777516d1..d3c5177d37 100644
@@ -171,6 +171,7 @@ struct VFIOPCIDevice {
     bool no_kvm_ioeventfd;
     bool no_vfio_ioeventfd;
     bool enable_ramfb;
+    bool defer_kvm_irq_routing;
     VFIODisplay *dpy;
     Notifier irqchip_change_notifier;
 };
#11 lore raw

RE: [PATCH v3 7/9] vfio: add infrastructure to commit the deferred kvm routing

Quoted 42 lines

-----Original Message-----
From: Alex Williamson [mailto:[email protected]]
Sent: Saturday, October 2, 2021 7:05 AM
To: Longpeng (Mike, Cloud Infrastructure Service Product Dept.) <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; Gonglei (Arei) <[email protected]>; chenjiashang <[email protected]>
Subject: Re: [PATCH v3 7/9] vfio: add infrastructure to commit the deferred kvm routing

hw/vfio/pci.c static void vfio_add_kvm_msi_virq(VFIOPCIDevice *vdev,

Quoted 14 lines · On Tue, 21 Sep 2021 07:02:00 +0800 "Longpeng(Mike)" <[email protected]>…

'defer_kvm_irq_routing' indicates whether we should defer to commit the kvm routing.

---
 hw/vfio/pci.c | 43 ++++++++++++++++++++++++++++++++++++++++++-
 hw/vfio/pci.h |  1 +
 2 files changed, 43 insertions(+), 1 deletion(-)
hw/vfio/pci.c +0 -0
index 8e97ca93cf..8fe238b11d 100644
@@ -423,12 +423,24 @@ static void vfio_add_kvm_msi_virq(VFIOPCIDevice *vdev,

VFIOMSIVector *vector,

hw/vfio/pci.c static void vfio_add_kvm_msi_virq(VFIOPCIDevice *vdev, VFIOMSIVector *vector,

Quoted 5 lines
hw/vfio/pci.c +1 -1
         return;
     }
 
-    virq = kvm_irqchip_add_msi_route(kvm_state, vector_n, &vdev->pdev);
+    virq = kvm_irqchip_add_deferred_msi_route(kvm_state, vector_n,

&vdev->pdev);

Quoted 5 lines
hw/vfio/pci.c +0 -0
     if (virq < 0) {
         event_notifier_cleanup(&vector->kvm_interrupt);
         return;
     }
 
Quoted 8 lines

hw/vfio/pci.c static void vfio_add_kvm_msi_virq(VFIOPCIDevice *vdev, VFIOMSIVector *vector, · line 436

Quoted 5 lines
hw/vfio/pci.c +5 -0
+    if (vdev->defer_kvm_irq_routing) {
+        /*
+         * Hold the allocated virq in vector->virq temporarily, will
+         * reset it to -1 when we fail to add the corresponding irqfd
+         * in vfio_commit_kvm_msi_virq().

s/when/if/

OK, thanks.

Quoted 7 lines

hw/vfio/pci.c static void vfio_add_kvm_msi_virq(VFIOPCIDevice *vdev, VFIOMSIVector *vector, · line 438

Quoted 2 lines
hw/vfio/pci.c +2 -0
+         */
+        vector->virq = virq;

Do we need to make this unique to the deferred case or could we use vector->virq directly and fill it with -1 on all error paths like we do on a failure in vfio_commit_kvm_msi_virq()?

OK, I will use vector->irq directly, it looks neater.

Quoted 12 lines

hw/vfio/pci.c static void vfio_add_kvm_msi_virq(VFIOPCIDevice *vdev, VFIOMSIVector *vector,

Quoted 7 lines
hw/vfio/pci.c +5 -0
+        return;
+    }
+
+    kvm_irqchip_commit_routes(kvm_state);
+
     if (kvm_irqchip_add_irqfd_notifier_gsi(kvm_state,

&vector->kvm_interrupt,

hw/vfio/pci.c static void vfio_msix_vector_release(PCIDevice *pdev,

Quoted 3 lines
hw/vfio/pci.c +0 -0
                                        NULL, virq) < 0) {
         kvm_irqchip_release_virq(kvm_state, virq);
@@ -567,6 +579,35 @@ static void vfio_msix_vector_release(PCIDevice *pdev,

unsigned int nr)

Quoted 7 lines

hw/vfio/pci.c static void vfio_msix_vector_release(PCIDevice *pdev, unsigned int nr) · line 582

Quoted 4 lines
hw/vfio/pci.c +1 -0
     }
 }
 
+/* TODO: invoked when enclabe msi/msix vectors */

"enclabe"? Is this meant to be "enable"?

Yes, it's a typo.

Quoted 6 lines

hw/vfio/pci.c static void vfio_msix_vector_release(PCIDevice *pdev, · line 583

Quoted 1 line
hw/vfio/pci.c +1 -0
+static __attribute__((unused)) void vfio_commit_kvm_msi_virq(VFIOPCIDevice

*vdev)

I'd move this function, if not this entire change, to patch 9 rather than adding these attributes for an unused function. Thanks,

OK. I think I should merge this patch into patch 9 entirely if we decide to move this function.

Quoted 44 lines

Alex

Quoted 42 lines
hw/vfio/pci.c +27 -0
+{
+    int i;
+    VFIOMSIVector *vector;
+
+    if (!vdev->defer_kvm_irq_routing || !vdev->nr_vectors) {
+        return;
+    }
+
+    kvm_irqchip_commit_routes(kvm_state);
+
+    for (i = 0; i < vdev->nr_vectors; i++) {
+        vector = &vdev->msi_vectors[i];
+
+        if (!vector->use || vector->virq < 0) {
+            continue;
+        }
+
+        if (kvm_irqchip_add_irqfd_notifier_gsi(kvm_state,
+                                               &vector->kvm_interrupt,
+                                               NULL, vector->virq) < 0) {
+            kvm_irqchip_release_virq(kvm_state, vector->virq);
+            event_notifier_cleanup(&vector->kvm_interrupt);
+            vector->virq = -1;
+        }
+    }
+}
+
 static void vfio_msix_enable(VFIOPCIDevice *vdev)
 {
     PCIDevice *pdev = &vdev->pdev;
hw/vfio/pci.h +1 -0
index 64777516d1..d3c5177d37 100644
@@ -171,6 +171,7 @@ struct VFIOPCIDevice {
     bool no_kvm_ioeventfd;
     bool no_vfio_ioeventfd;
     bool enable_ramfb;
+    bool defer_kvm_irq_routing;
     VFIODisplay *dpy;
     Notifier irqchip_change_notifier;
 };
#12 lore raw

PATCH v3 8/9 Revert "vfio: Avoid disabling and enabling vectors repeatedly in VFIO migration"

In reply to Longpeng(Mike)

Commit ecebe53fe993 ("vfio: Avoid disabling and enabling vectors repeatedly in VFIO migration") avoid inefficiently disabling and enabling vectors repeatedly and let the unmasked vectors to be enabled one by one.

But we want to batch multiple routes and defer the commit, and only commit once out side the loop of setting vector notifiers, so we cannot to enable the vectors one by one in the loop now.

Revert that commit and we will take another way in the next patch, it can not only avoid disabling/enabling vectors repeatedly, but also satisfy our requirement of defer to commit.

---
 hw/vfio/pci.c | 20 +++-----------------
 1 file changed, 3 insertions(+), 17 deletions(-)
hw/vfio/pci.c +3 -17
index 8fe238b11d..2de1cc5425 100644
@@ -610,9 +610,6 @@ static __attribute__((unused)) void vfio_commit_kvm_msi_virq(VFIOPCIDevice *vdev
 
 static void vfio_msix_enable(VFIOPCIDevice *vdev)
 {
-    PCIDevice *pdev = &vdev->pdev;
-    unsigned int nr, max_vec = 0;
-
     vfio_disable_interrupts(vdev);
 
     vdev->msi_vectors = g_new0(VFIOMSIVector, vdev->msix->entries);
@@ -631,22 +628,11 @@ static void vfio_msix_enable(VFIOPCIDevice *vdev)
      * triggering to userspace, then immediately release the vector, leaving
      * the physical device with no vectors enabled, but MSI-X enabled, just
      * like the guest view.
-     * If there are already unmasked vectors (in migration resume phase and
-     * some guest startups) which will be enabled soon, we can allocate all
-     * of them here to avoid inefficiently disabling and enabling vectors
-     * repeatedly later.
      */
-    if (!pdev->msix_function_masked) {
-        for (nr = 0; nr < msix_nr_vectors_allocated(pdev); nr++) {
-            if (!msix_is_masked(pdev, nr)) {
-                max_vec = nr;
-            }
-        }
-    }
-    vfio_msix_vector_do_use(pdev, max_vec, NULL, NULL);
-    vfio_msix_vector_release(pdev, max_vec);
+    vfio_msix_vector_do_use(&vdev->pdev, 0, NULL, NULL);
+    vfio_msix_vector_release(&vdev->pdev, 0);
 
-    if (msix_set_vector_notifiers(pdev, vfio_msix_vector_use,
+    if (msix_set_vector_notifiers(&vdev->pdev, vfio_msix_vector_use,
                                   vfio_msix_vector_release, NULL)) {
         error_report("vfio: msix_set_vector_notifiers failed");
     }
Signature · 1 line
2.23.0
#13 lore raw

Re: [PATCH v3 8/9] Revert "vfio: Avoid disabling and enabling vectors repeatedly in VFIO migration"

Quoted 2 lines · On Tue, 21 Sep 2021 07:02:01 +0800 "Longpeng(Mike)" <[email protected]>…

Commit ecebe53fe993 ("vfio: Avoid disabling and enabling vectors repeatedly in VFIO migration") avoid inefficiently disabling and

s/avoid/avoids/

Quoted 1 line

enabling vectors repeatedly and let the unmasked vectors to be

s/let/lets/ s/to//

Quoted 4 lines

enabled one by one.

But we want to batch multiple routes and defer the commit, and only commit once out side the loop of setting vector notifiers, so we

s/out side/outside/

Quoted 1 line

cannot to enable the vectors one by one in the loop now.

s/to//

Thanks,
Alex

Quoted 50 lines

Revert that commit and we will take another way in the next patch, it can not only avoid disabling/enabling vectors repeatedly, but also satisfy our requirement of defer to commit.

---
 hw/vfio/pci.c | 20 +++-----------------
 1 file changed, 3 insertions(+), 17 deletions(-)
hw/vfio/pci.c +3 -17
index 8fe238b11d..2de1cc5425 100644
@@ -610,9 +610,6 @@ static __attribute__((unused)) void vfio_commit_kvm_msi_virq(VFIOPCIDevice *vdev
 
 static void vfio_msix_enable(VFIOPCIDevice *vdev)
 {
-    PCIDevice *pdev = &vdev->pdev;
-    unsigned int nr, max_vec = 0;
-
     vfio_disable_interrupts(vdev);
 
     vdev->msi_vectors = g_new0(VFIOMSIVector, vdev->msix->entries);
@@ -631,22 +628,11 @@ static void vfio_msix_enable(VFIOPCIDevice *vdev)
      * triggering to userspace, then immediately release the vector, leaving
      * the physical device with no vectors enabled, but MSI-X enabled, just
      * like the guest view.
-     * If there are already unmasked vectors (in migration resume phase and
-     * some guest startups) which will be enabled soon, we can allocate all
-     * of them here to avoid inefficiently disabling and enabling vectors
-     * repeatedly later.
      */
-    if (!pdev->msix_function_masked) {
-        for (nr = 0; nr < msix_nr_vectors_allocated(pdev); nr++) {
-            if (!msix_is_masked(pdev, nr)) {
-                max_vec = nr;
-            }
-        }
-    }
-    vfio_msix_vector_do_use(pdev, max_vec, NULL, NULL);
-    vfio_msix_vector_release(pdev, max_vec);
+    vfio_msix_vector_do_use(&vdev->pdev, 0, NULL, NULL);
+    vfio_msix_vector_release(&vdev->pdev, 0);
 
-    if (msix_set_vector_notifiers(pdev, vfio_msix_vector_use,
+    if (msix_set_vector_notifiers(&vdev->pdev, vfio_msix_vector_use,
                                   vfio_msix_vector_release, NULL)) {
         error_report("vfio: msix_set_vector_notifiers failed");
     }
#14 lore raw

RE: [PATCH v3 8/9] Revert "vfio: Avoid disabling and enabling vectors repeatedly in VFIO migration"

Quoted 27 lines

-----Original Message-----
From: Alex Williamson [mailto:[email protected]]
Sent: Saturday, October 2, 2021 7:04 AM
To: Longpeng (Mike, Cloud Infrastructure Service Product Dept.) <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; Gonglei (Arei) <[email protected]>; chenjiashang <[email protected]>
Subject: Re: [PATCH v3 8/9] Revert "vfio: Avoid disabling and enabling vectors repeatedly in VFIO migration"

Quoted 2 lines · On Tue, 21 Sep 2021 07:02:01 +0800 "Longpeng(Mike)" <[email protected]>…

Commit ecebe53fe993 ("vfio: Avoid disabling and enabling vectors repeatedly in VFIO migration") avoid inefficiently disabling and

s/avoid/avoids/

Quoted 1 line

enabling vectors repeatedly and let the unmasked vectors to be

s/let/lets/ s/to//

Quoted 2 lines

enabled one by one.

Quoted 9 lines
Quoted 2 lines

But we want to batch multiple routes and defer the commit, and only commit once out side the loop of setting vector notifiers, so we

s/out side/outside/

Quoted 1 line

cannot to enable the vectors one by one in the loop now.

s/to//

Thanks. All the typos and grammatical errors you pointed out will be fixed in v4.

Quoted 54 lines

Thanks,
Alex

hw/vfio/pci.c static __attribute__((unused)) void

Quoted 15 lines

Revert that commit and we will take another way in the next patch, it can not only avoid disabling/enabling vectors repeatedly, but also satisfy our requirement of defer to commit.

---
 hw/vfio/pci.c | 20 +++-----------------
 1 file changed, 3 insertions(+), 17 deletions(-)
hw/vfio/pci.c +0 -0
index 8fe238b11d..2de1cc5425 100644
@@ -610,9 +610,6 @@ static __attribute__((unused)) void

vfio_commit_kvm_msi_virq(VFIOPCIDevice *vdev

Quoted 35 lines
hw/vfio/pci.c +3 -17
 static void vfio_msix_enable(VFIOPCIDevice *vdev)
 {
-    PCIDevice *pdev = &vdev->pdev;
-    unsigned int nr, max_vec = 0;
-
     vfio_disable_interrupts(vdev);
 
     vdev->msi_vectors = g_new0(VFIOMSIVector, vdev->msix->entries);
@@ -631,22 +628,11 @@ static void vfio_msix_enable(VFIOPCIDevice *vdev)
      * triggering to userspace, then immediately release the vector, leaving
      * the physical device with no vectors enabled, but MSI-X enabled, just
      * like the guest view.
-     * If there are already unmasked vectors (in migration resume phase and
-     * some guest startups) which will be enabled soon, we can allocate all
-     * of them here to avoid inefficiently disabling and enabling vectors
-     * repeatedly later.
      */
-    if (!pdev->msix_function_masked) {
-        for (nr = 0; nr < msix_nr_vectors_allocated(pdev); nr++) {
-            if (!msix_is_masked(pdev, nr)) {
-                max_vec = nr;
-            }
-        }
-    }
-    vfio_msix_vector_do_use(pdev, max_vec, NULL, NULL);
-    vfio_msix_vector_release(pdev, max_vec);
+    vfio_msix_vector_do_use(&vdev->pdev, 0, NULL, NULL);
+    vfio_msix_vector_release(&vdev->pdev, 0);
 
-    if (msix_set_vector_notifiers(pdev, vfio_msix_vector_use,
+    if (msix_set_vector_notifiers(&vdev->pdev, vfio_msix_vector_use,
                                   vfio_msix_vector_release, NULL)) {
         error_report("vfio: msix_set_vector_notifiers failed");
     }
#15 lore raw

PATCH v3 9/9 vfio: defer to commit kvm irq routing when enable msi/msix

In reply to Longpeng(Mike)

In migration resume phase, all unmasked msix vectors need to be setup when load the VF state. However, the setup operation would take longer if the VM has more VFs and each VF has more unmasked vectors.

The hot spot is kvm_irqchip_commit_routes, it'll scan and update all irqfds that already assigned each invocation, so more vectors means need more time to process them.

vfio_pci_load_config
vfio_msix_enable

    msix_set_vector_notifiers
      for (vector = 0; vector < dev->msix_entries_nr; vector++) {
        vfio_msix_vector_do_use
          vfio_add_kvm_msi_virq
            kvm_irqchip_commit_routes <-- expensive
      }

We can reduce the cost by only commit once outside the loop. The routes is cached in kvm_state, we commit them first and then bind irqfd for each vector.

The test VM has 128 vcpus and 8 VF (each one has 65 vectors), we measure the cost of the vfio_msix_enable for each VF, and we can see 90+% costs can be reduce.

VF Count of irqfds[*] Original With this patch

1st           65            8               2
2nd           130           15              2
3rd           195           22              2
4th           260           24              3
5th           325           36              2
6th           390           44              3
7th           455           51              3
8th           520           58              4
Total                       258ms           21ms

[*] Count of irqfds
How many irqfds that already assigned and need to process in this round.

The optimition can be applied to msi type too.

---
 hw/vfio/pci.c | 36 ++++++++++++++++++++++++++++--------
 1 file changed, 28 insertions(+), 8 deletions(-)
hw/vfio/pci.c +28 -8
index 2de1cc5425..b26129bddf 100644
@@ -513,11 +513,13 @@ static int vfio_msix_vector_do_use(PCIDevice *pdev, unsigned int nr,
      * increase them as needed.
      */
     if (vdev->nr_vectors < nr + 1) {
-        vfio_disable_irqindex(&vdev->vbasedev, VFIO_PCI_MSIX_IRQ_INDEX);
         vdev->nr_vectors = nr + 1;
-        ret = vfio_enable_vectors(vdev, true);
-        if (ret) {
-            error_report("vfio: failed to enable vectors, %d", ret);
+        if (!vdev->defer_kvm_irq_routing) {
+            vfio_disable_irqindex(&vdev->vbasedev, VFIO_PCI_MSIX_IRQ_INDEX);
+            ret = vfio_enable_vectors(vdev, true);
+            if (ret) {
+                error_report("vfio: failed to enable vectors, %d", ret);
+            }
         }
     } else {
         Error *err = NULL;
@@ -579,8 +581,7 @@ static void vfio_msix_vector_release(PCIDevice *pdev, unsigned int nr)
     }
 }
 
-/* TODO: invoked when enclabe msi/msix vectors */
-static __attribute__((unused)) void vfio_commit_kvm_msi_virq(VFIOPCIDevice *vdev)
+static void vfio_commit_kvm_msi_virq(VFIOPCIDevice *vdev)
 {
     int i;
     VFIOMSIVector *vector;
@@ -610,6 +611,9 @@ static __attribute__((unused)) void vfio_commit_kvm_msi_virq(VFIOPCIDevice *vdev
 
 static void vfio_msix_enable(VFIOPCIDevice *vdev)
 {
+    PCIDevice *pdev = &vdev->pdev;
+    int ret;
+
     vfio_disable_interrupts(vdev);
 
     vdev->msi_vectors = g_new0(VFIOMSIVector, vdev->msix->entries);
@@ -632,11 +636,22 @@ static void vfio_msix_enable(VFIOPCIDevice *vdev)
     vfio_msix_vector_do_use(&vdev->pdev, 0, NULL, NULL);
     vfio_msix_vector_release(&vdev->pdev, 0);
 
-    if (msix_set_vector_notifiers(&vdev->pdev, vfio_msix_vector_use,
-                                  vfio_msix_vector_release, NULL)) {
+    vdev->defer_kvm_irq_routing = true;
+
+    ret = msix_set_vector_notifiers(&vdev->pdev, vfio_msix_vector_use,
+                                    vfio_msix_vector_release, NULL);
+    if (ret < 0) {
         error_report("vfio: msix_set_vector_notifiers failed");
+    } else if (!pdev->msix_function_masked) {
+        vfio_commit_kvm_msi_virq(vdev);
+        vfio_disable_irqindex(&vdev->vbasedev, VFIO_PCI_MSIX_IRQ_INDEX);
+        ret = vfio_enable_vectors(vdev, true);
+        if (ret) {
+            error_report("vfio: failed to enable vectors, %d", ret);
+        }
     }
 
+    vdev->defer_kvm_irq_routing = false;
     trace_vfio_msix_enable(vdev->vbasedev.name);
 }
 
@@ -645,6 +660,7 @@ static void vfio_msi_enable(VFIOPCIDevice *vdev)
     int ret, i;
 
     vfio_disable_interrupts(vdev);
+    vdev->defer_kvm_irq_routing = true;
 
     vdev->nr_vectors = msi_nr_vectors_allocated(&vdev->pdev);
 retry:
@@ -671,6 +687,8 @@ retry:
         vfio_add_kvm_msi_virq(vdev, vector, i, false);
     }
 
+    vfio_commit_kvm_msi_virq(vdev);
+
     /* Set interrupt type prior to possible interrupts */
     vdev->interrupt = VFIO_INT_MSI;
 
@@ -697,9 +715,11 @@ retry:
          */
         error_report("vfio: Error: Failed to enable MSI");
 
+        vdev->defer_kvm_irq_routing = false;
         return;
     }
 
+    vdev->defer_kvm_irq_routing = false;
     trace_vfio_msi_enable(vdev->vbasedev.name, vdev->nr_vectors);
 }
 
Signature · 1 line
2.23.0
#16 lore raw

Re: [PATCH v3 9/9] vfio: defer to commit kvm irq routing when enable msi/msix

Quoted 2 lines · On Tue, 21 Sep 2021 07:02:02 +0800 "Longpeng(Mike)" <[email protected]>…

In migration resume phase, all unmasked msix vectors need to be setup when load the VF state. However, the setup operation would

s/load/loading/

Quoted 5 lines

take longer if the VM has more VFs and each VF has more unmasked vectors.

The hot spot is kvm_irqchip_commit_routes, it'll scan and update all irqfds that already assigned each invocation, so more vectors

s/that/that are/

Quoted 4 lines

means need more time to process them.

vfio_pci_load_config
vfio_msix_enable

Quoted 8 lines
    msix_set_vector_notifiers
      for (vector = 0; vector < dev->msix_entries_nr; vector++) {
        vfio_msix_vector_do_use
          vfio_add_kvm_msi_virq
            kvm_irqchip_commit_routes <-- expensive
      }

We can reduce the cost by only commit once outside the loop. The

s/commit/committing/

Quoted 1 line

routes is cached in kvm_state, we commit them first and then bind

s/is/are/

Quoted 18 lines

irqfd for each vector.

The test VM has 128 vcpus and 8 VF (each one has 65 vectors), we measure the cost of the vfio_msix_enable for each VF, and we can see 90+% costs can be reduce.

VF Count of irqfds[*] Original With this patch

1st           65            8               2
2nd           130           15              2
3rd           195           22              2
4th           260           24              3
5th           325           36              2
6th           390           44              3
7th           455           51              3
8th           520           58              4
Total                       258ms           21ms
Quoted 5 lines

[*] Count of irqfds
How many irqfds that already assigned and need to process in this round.

The optimition can be applied to msi type too.

s/optimition/optimization/

Quoted 49 lines
---
 hw/vfio/pci.c | 36 ++++++++++++++++++++++++++++--------
 1 file changed, 28 insertions(+), 8 deletions(-)
hw/vfio/pci.c +10 -6
index 2de1cc5425..b26129bddf 100644
@@ -513,11 +513,13 @@ static int vfio_msix_vector_do_use(PCIDevice *pdev, unsigned int nr,
      * increase them as needed.
      */
     if (vdev->nr_vectors < nr + 1) {
-        vfio_disable_irqindex(&vdev->vbasedev, VFIO_PCI_MSIX_IRQ_INDEX);
         vdev->nr_vectors = nr + 1;
-        ret = vfio_enable_vectors(vdev, true);
-        if (ret) {
-            error_report("vfio: failed to enable vectors, %d", ret);
+        if (!vdev->defer_kvm_irq_routing) {
+            vfio_disable_irqindex(&vdev->vbasedev, VFIO_PCI_MSIX_IRQ_INDEX);
+            ret = vfio_enable_vectors(vdev, true);
+            if (ret) {
+                error_report("vfio: failed to enable vectors, %d", ret);
+            }
         }
     } else {
         Error *err = NULL;
@@ -579,8 +581,7 @@ static void vfio_msix_vector_release(PCIDevice *pdev, unsigned int nr)
     }
 }
 
-/* TODO: invoked when enclabe msi/msix vectors */
-static __attribute__((unused)) void vfio_commit_kvm_msi_virq(VFIOPCIDevice *vdev)
+static void vfio_commit_kvm_msi_virq(VFIOPCIDevice *vdev)
 {
     int i;
     VFIOMSIVector *vector;
@@ -610,6 +611,9 @@ static __attribute__((unused)) void vfio_commit_kvm_msi_virq(VFIOPCIDevice *vdev
 
 static void vfio_msix_enable(VFIOPCIDevice *vdev)
 {
+    PCIDevice *pdev = &vdev->pdev;
+    int ret;
+
     vfio_disable_interrupts(vdev);
 
     vdev->msi_vectors = g_new0(VFIOMSIVector, vdev->msix->entries);

hw/vfio/pci.c static void vfio_msix_enable(VFIOPCIDevice *vdev) · line 636

Quoted 5 lines
hw/vfio/pci.c +0 -2
@@ -632,11 +636,22 @@ static void vfio_msix_enable(VFIOPCIDevice *vdev)
     vfio_msix_vector_do_use(&vdev->pdev, 0, NULL, NULL);
     vfio_msix_vector_release(&vdev->pdev, 0);
 
-    if (msix_set_vector_notifiers(&vdev->pdev, vfio_msix_vector_use,
-                                  vfio_msix_vector_release, NULL)) {

A comment would be useful here, maybe something like:

/*
* Setting vector notifiers triggers synchronous vector-use
* callbacks for each active vector. Deferring to commit the KVM
* routes once rather than per vector provides a substantial
* performance improvement.
*/

hw/vfio/pci.c static void vfio_msix_enable(VFIOPCIDevice *vdev) · line 647

Quoted 9 lines
hw/vfio/pci.c +8 -0
+    vdev->defer_kvm_irq_routing = true;
+
+    ret = msix_set_vector_notifiers(&vdev->pdev, vfio_msix_vector_use,
+                                    vfio_msix_vector_release, NULL);
+    if (ret < 0) {
         error_report("vfio: msix_set_vector_notifiers failed");
+    } else if (!pdev->msix_function_masked) {
+        vfio_commit_kvm_msi_virq(vdev);
+        vfio_disable_irqindex(&vdev->vbasedev, VFIO_PCI_MSIX_IRQ_INDEX);

Couldn't we also optimize the do_use/release on vector 0 above to avoid this gratuitous disable here? We only want to make sure MSIX is always enabled on the device when we exit this function, so maybe that code becomes an "else" branch below?

Quoted 28 lines
hw/vfio/pci.c +8 -0
+        ret = vfio_enable_vectors(vdev, true);
+        if (ret) {
+            error_report("vfio: failed to enable vectors, %d", ret);
+        }
     }
 
+    vdev->defer_kvm_irq_routing = false;
     trace_vfio_msix_enable(vdev->vbasedev.name);
 }
 
@@ -645,6 +660,7 @@ static void vfio_msi_enable(VFIOPCIDevice *vdev)
     int ret, i;
 
     vfio_disable_interrupts(vdev);
+    vdev->defer_kvm_irq_routing = true;
 
     vdev->nr_vectors = msi_nr_vectors_allocated(&vdev->pdev);
 retry:
@@ -671,6 +687,8 @@ retry:
         vfio_add_kvm_msi_virq(vdev, vector, i, false);
     }
 
+    vfio_commit_kvm_msi_virq(vdev);
+
     /* Set interrupt type prior to possible interrupts */
     vdev->interrupt = VFIO_INT_MSI;
 

hw/vfio/pci.c retry: · line 722

Quoted 8 lines
hw/vfio/pci.c +2 -0
@@ -697,9 +715,11 @@ retry:
          */
         error_report("vfio: Error: Failed to enable MSI");
 
+        vdev->defer_kvm_irq_routing = false;
         return;
     }
 
+    vdev->defer_kvm_irq_routing = false;

Why wouldn't we clear the flag in vfio_commit_kvm_msi_virq()? It almost feels like there should be a vfio_prepare_kvm_msi_virq_batch() that enables the flag and an unconditional vfio_commit_kvm_msi_virq_batch() that clears the flag and decides if further work is necessary. Thanks,

Alex

Quoted 3 lines
hw/vfio/pci.c +0 -0
     trace_vfio_msi_enable(vdev->vbasedev.name, vdev->nr_vectors);
 }
 
#17 lore raw

RE: [PATCH v3 9/9] vfio: defer to commit kvm irq routing when enable msi/msix

Quoted 40 lines

-----Original Message-----
From: Alex Williamson [mailto:[email protected]]
Sent: Saturday, October 2, 2021 7:05 AM
To: Longpeng (Mike, Cloud Infrastructure Service Product Dept.) <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; Gonglei (Arei) <[email protected]>; chenjiashang <[email protected]>
Subject: Re: [PATCH v3 9/9] vfio: defer to commit kvm irq routing when enable msi/msix

Quoted 2 lines · On Tue, 21 Sep 2021 07:02:02 +0800 "Longpeng(Mike)" <[email protected]>…

In migration resume phase, all unmasked msix vectors need to be setup when load the VF state. However, the setup operation would

s/load/loading/

Quoted 5 lines

take longer if the VM has more VFs and each VF has more unmasked vectors.

The hot spot is kvm_irqchip_commit_routes, it'll scan and update all irqfds that already assigned each invocation, so more vectors

s/that/that are/

Quoted 11 lines

means need more time to process them.

vfio_pci_load_config
vfio_msix_enable

    msix_set_vector_notifiers
      for (vector = 0; vector < dev->msix_entries_nr; vector++) {
        vfio_msix_vector_do_use
          vfio_add_kvm_msi_virq
            kvm_irqchip_commit_routes <-- expensive
      }
Quoted 4 lines
Quoted 1 line

We can reduce the cost by only commit once outside the loop. The

s/commit/committing/

OK, will fix in the next version, thanks.

Quoted 4 lines
Quoted 1 line

routes is cached in kvm_state, we commit them first and then bind

s/is/are/

OK.

Quoted 18 lines
Quoted 18 lines

irqfd for each vector.

The test VM has 128 vcpus and 8 VF (each one has 65 vectors), we measure the cost of the vfio_msix_enable for each VF, and we can see 90+% costs can be reduce.

VF Count of irqfds[*] Original With this patch

1st           65            8               2
2nd           130           15              2
3rd           195           22              2
4th           260           24              3
5th           325           36              2
6th           390           44              3
7th           455           51              3
8th           520           58              4
Total                       258ms           21ms
Quoted 8 lines
Quoted 5 lines

[*] Count of irqfds
How many irqfds that already assigned and need to process in this round.

The optimition can be applied to msi type too.

s/optimition/optimization/

OK, thanks.

Quoted 62 lines

hw/vfio/pci.c static int vfio_msix_vector_do_use(PCIDevice *pdev,

Quoted 11 lines
---
 hw/vfio/pci.c | 36 ++++++++++++++++++++++++++++--------
 1 file changed, 28 insertions(+), 8 deletions(-)
hw/vfio/pci.c +0 -0
index 2de1cc5425..b26129bddf 100644
@@ -513,11 +513,13 @@ static int vfio_msix_vector_do_use(PCIDevice *pdev,

unsigned int nr,

hw/vfio/pci.c static int vfio_msix_vector_do_use(PCIDevice *pdev, unsigned int nr,

Quoted 10 lines
hw/vfio/pci.c +2 -4
      * increase them as needed.
      */
     if (vdev->nr_vectors < nr + 1) {
-        vfio_disable_irqindex(&vdev->vbasedev, VFIO_PCI_MSIX_IRQ_INDEX);
         vdev->nr_vectors = nr + 1;
-        ret = vfio_enable_vectors(vdev, true);
-        if (ret) {
-            error_report("vfio: failed to enable vectors, %d", ret);
+        if (!vdev->defer_kvm_irq_routing) {
+            vfio_disable_irqindex(&vdev->vbasedev,

VFIO_PCI_MSIX_IRQ_INDEX);

hw/vfio/pci.c static void vfio_msix_vector_release(PCIDevice *pdev,

Quoted 8 lines
hw/vfio/pci.c +4 -0
+            ret = vfio_enable_vectors(vdev, true);
+            if (ret) {
+                error_report("vfio: failed to enable vectors, %d", ret);
+            }
         }
     } else {
         Error *err = NULL;
@@ -579,8 +581,7 @@ static void vfio_msix_vector_release(PCIDevice *pdev,

unsigned int nr)

hw/vfio/pci.c static void vfio_msix_vector_release(PCIDevice *pdev, unsigned int nr)

Quoted 5 lines
hw/vfio/pci.c +0 -2
     }
 }
 
-/* TODO: invoked when enclabe msi/msix vectors */
-static __attribute__((unused)) void vfio_commit_kvm_msi_virq(VFIOPCIDevice

*vdev)

hw/vfio/pci.c static __attribute__((unused)) void

Quoted 5 lines
hw/vfio/pci.c +1 -0
+static void vfio_commit_kvm_msi_virq(VFIOPCIDevice *vdev)
 {
     int i;
     VFIOMSIVector *vector;
@@ -610,6 +611,9 @@ static __attribute__((unused)) void

vfio_commit_kvm_msi_virq(VFIOPCIDevice *vdev

Quoted 10 lines
hw/vfio/pci.c +3 -0
 static void vfio_msix_enable(VFIOPCIDevice *vdev)
 {
+    PCIDevice *pdev = &vdev->pdev;
+    int ret;
+
     vfio_disable_interrupts(vdev);
 
     vdev->msi_vectors = g_new0(VFIOMSIVector, vdev->msix->entries);

hw/vfio/pci.c static void vfio_msix_enable(VFIOPCIDevice *vdev) · line 636

Quoted 5 lines
hw/vfio/pci.c +0 -2
@@ -632,11 +636,22 @@ static void vfio_msix_enable(VFIOPCIDevice *vdev)
     vfio_msix_vector_do_use(&vdev->pdev, 0, NULL, NULL);
     vfio_msix_vector_release(&vdev->pdev, 0);
 
-    if (msix_set_vector_notifiers(&vdev->pdev, vfio_msix_vector_use,
-                                  vfio_msix_vector_release, NULL)) {

A comment would be useful here, maybe something like:

Quoted 7 lines

/*
* Setting vector notifiers triggers synchronous vector-use
* callbacks for each active vector. Deferring to commit the KVM
* routes once rather than per vector provides a substantial
* performance improvement.
*/

Will add in the next version.

Quoted 10 lines

hw/vfio/pci.c static void vfio_msix_enable(VFIOPCIDevice *vdev) · line 647

Quoted 9 lines
hw/vfio/pci.c +8 -0
+    vdev->defer_kvm_irq_routing = true;
+
+    ret = msix_set_vector_notifiers(&vdev->pdev, vfio_msix_vector_use,
+                                    vfio_msix_vector_release, NULL);
+    if (ret < 0) {
         error_report("vfio: msix_set_vector_notifiers failed");
+    } else if (!pdev->msix_function_masked) {
+        vfio_commit_kvm_msi_virq(vdev);
+        vfio_disable_irqindex(&vdev->vbasedev, VFIO_PCI_MSIX_IRQ_INDEX);
Quoted 2 lines

Couldn't we also optimize the do_use/release on vector 0 above to avoid this gratuitous disable here? We only want to make sure MSIX is always

The disable here seems can be removed directly, because we already disable interrupts at the beginning of the vfio_msix_enable() ?

Quoted 3 lines

enabled on the device when we exit this function, so maybe that code becomes an "else" branch below?

Do you mean something like:

if (ret < 0) {
    ....
} else if (!pdev->msix_function_masked) {
    ....
    ret = vfio_enable_vectors(vdev, true);
    ....
} else {
    /* do_use/release on vector 0 */
}

We'll get '-EINVAL' if invoke vfio_enable_vectors with vdev->nr_vectors=0, this cannot happen before but it can now in this way. So maybe the "else if" conditional expression should be convert to "!pdev->msix_function_masked && vdev->nr_vectors" ?

Quoted 48 lines
Quoted 28 lines
hw/vfio/pci.c +8 -0
+        ret = vfio_enable_vectors(vdev, true);
+        if (ret) {
+            error_report("vfio: failed to enable vectors, %d", ret);
+        }
     }
 
+    vdev->defer_kvm_irq_routing = false;
     trace_vfio_msix_enable(vdev->vbasedev.name);
 }
 
@@ -645,6 +660,7 @@ static void vfio_msi_enable(VFIOPCIDevice *vdev)
     int ret, i;
 
     vfio_disable_interrupts(vdev);
+    vdev->defer_kvm_irq_routing = true;
 
     vdev->nr_vectors = msi_nr_vectors_allocated(&vdev->pdev);
 retry:
@@ -671,6 +687,8 @@ retry:
         vfio_add_kvm_msi_virq(vdev, vector, i, false);
     }
 
+    vfio_commit_kvm_msi_virq(vdev);
+
     /* Set interrupt type prior to possible interrupts */
     vdev->interrupt = VFIO_INT_MSI;
 

hw/vfio/pci.c retry: · line 722

Quoted 8 lines
hw/vfio/pci.c +2 -0
@@ -697,9 +715,11 @@ retry:
          */
         error_report("vfio: Error: Failed to enable MSI");
 
+        vdev->defer_kvm_irq_routing = false;
         return;
     }
 
+    vdev->defer_kvm_irq_routing = false;

Why wouldn't we clear the flag in vfio_commit_kvm_msi_virq()? It almost feels like there should be a vfio_prepare_kvm_msi_virq_batch() that enables the flag and an unconditional vfio_commit_kvm_msi_virq_batch() that clears the flag and decides if further work is necessary. Thanks,

Alex

Quoted 3 lines
hw/vfio/pci.c +0 -0
     trace_vfio_msi_enable(vdev->vbasedev.name, vdev->nr_vectors);
 }