qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Peter Xu <peterx@redhat.com>
To: qemu-devel@nongnu.org
Cc: imammedo@redhat.com, rth@twiddle.net, ehabkost@redhat.com,
	jasowang@redhat.com, marcel@redhat.com, mst@redhat.com,
	pbonzini@redhat.com, jan.kiszka@web.de, rkrcmar@redhat.com,
	alex.williamson@redhat.com, wexu@redhat.com,
	davidkiarie4@gmail.com, peterx@redhat.com
Subject: [Qemu-devel] [PATCH v10.2 24/26] kvm-irqchip: introduce kvm_irqchip_update_msi_route_no_commit
Date: Wed, 22 Jun 2016 11:42:45 +0800	[thread overview]
Message-ID: <1466566965-12809-1-git-send-email-peterx@redhat.com> (raw)
In-Reply-To: <1466495274-5011-25-git-send-email-peterx@redhat.com>

We do gsi route commit for each irqchip route update before. This is not
efficient if we are updating lots of routes in the same time. This patch
introduce a new "no_commit" version of the update function, which can be
used when we update multiple route entries in a sequence.

This change also requires that we export kvm_irqchip_commit_routes() to
public.

Signed-off-by: Peter Xu <peterx@redhat.com>
---

After a second thought, a better way to do this is to introduce
another new function, rather than modifying the old
kvm_iqrchip_udpate_msi_route(). By doing this, I can avoid touch other
part of codes, also I can keep add_msi_route() and update_msi_route()
aligned since both of them will contain one commit phase.

Please review this v10.2 instead of v10 for this patch, and will use
this one in future versions if np.

 hw/intc/ioapic.c     |  3 ++-
 include/sysemu/kvm.h | 10 +++++++++-
 kvm-all.c            | 19 +++++++++++++++----
 kvm-stub.c           |  4 ++++
 target-i386/kvm.c    |  5 +++--
 5 files changed, 33 insertions(+), 8 deletions(-)

diff --git a/hw/intc/ioapic.c b/hw/intc/ioapic.c
index 0c34e3e..931aeaf 100644
--- a/hw/intc/ioapic.c
+++ b/hw/intc/ioapic.c
@@ -192,7 +192,8 @@ static void ioapic_update_kvm_routes(IOAPICCommonState *s)
             ioapic_entry_parse(s->ioredtbl[i], &info);
             msg.address = info.addr;
             msg.data = info.data;
-            kvm_irqchip_update_msi_route(kvm_state, i, msg, NULL);
+            kvm_irqchip_update_msi_route_no_commit(kvm_state, i,
+                                                   msg, NULL);
         }
         kvm_irqchip_commit_routes(kvm_state);
     }
diff --git a/include/sysemu/kvm.h b/include/sysemu/kvm.h
index 0a16e0e..ba7e7f0 100644
--- a/include/sysemu/kvm.h
+++ b/include/sysemu/kvm.h
@@ -371,7 +371,6 @@ int kvm_set_irq(KVMState *s, int irq, int level);
 int kvm_irqchip_send_msi(KVMState *s, MSIMessage msg);
 
 void kvm_irqchip_add_irq_route(KVMState *s, int gsi, int irqchip, int pin);
-void kvm_irqchip_commit_routes(KVMState *s);
 
 void kvm_put_apic_state(DeviceState *d, struct kvm_lapic_state *kapic);
 void kvm_get_apic_state(DeviceState *d, struct kvm_lapic_state *kapic);
@@ -494,6 +493,15 @@ static inline void cpu_synchronize_post_init(CPUState *cpu)
 int kvm_irqchip_add_msi_route(KVMState *s, int vector, PCIDevice *dev);
 int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg,
                                  PCIDevice *dev);
+/*
+ * Same as kvm_irqchip_update_msi_route(), but need explicit
+ * kvm_irqchip_commit_routes() afterward. This is efficient when we
+ * need to update multiple MSI routes at the same time, to avoid
+ * unnecessary commits between updates.
+ */
+int kvm_irqchip_update_msi_route_no_commit(KVMState *s, int virq,
+                                           MSIMessage msg, PCIDevice *dev);
+void kvm_irqchip_commit_routes(KVMState *s);
 void kvm_irqchip_release_virq(KVMState *s, int virq);
 
 int kvm_irqchip_add_adapter_route(KVMState *s, AdapterInfo *adapter);
diff --git a/kvm-all.c b/kvm-all.c
index ca30a58..b896184 100644
--- a/kvm-all.c
+++ b/kvm-all.c
@@ -1094,8 +1094,6 @@ static int kvm_update_routing_entry(KVMState *s,
 
         *entry = *new_entry;
 
-        kvm_irqchip_commit_routes(s);
-
         return 0;
     }
 
@@ -1280,8 +1278,8 @@ int kvm_irqchip_add_msi_route(KVMState *s, int vector, PCIDevice *dev)
     return virq;
 }
 
-int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg,
-                                 PCIDevice *dev)
+int kvm_irqchip_update_msi_route_no_commit(KVMState *s, int virq,
+                                           MSIMessage msg, PCIDevice *dev)
 {
     struct kvm_irq_routing_entry kroute = {};
 
@@ -1306,6 +1304,19 @@ int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg,
     return kvm_update_routing_entry(s, &kroute);
 }
 
+int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg,
+                                 PCIDevice *dev)
+{
+    int ret;
+
+    ret = kvm_irqchip_update_msi_route_no_commit(s, virq, msg, dev);
+    if (ret >= 0) {
+        kvm_irqchip_commit_routes(s);
+    }
+
+    return ret;
+}
+
 static int kvm_irqchip_assign_irqfd(KVMState *s, int fd, int rfd, int virq,
                                     bool assign)
 {
diff --git a/kvm-stub.c b/kvm-stub.c
index 982e590..64e23f6 100644
--- a/kvm-stub.c
+++ b/kvm-stub.c
@@ -135,6 +135,10 @@ int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg,
     return -ENOSYS;
 }
 
+void kvm_irqchip_commit_routes(KVMState *s)
+{
+}
+
 int kvm_irqchip_add_adapter_route(KVMState *s, AdapterInfo *adapter)
 {
     return -ENOSYS;
diff --git a/target-i386/kvm.c b/target-i386/kvm.c
index f02ba0a..62aec24 100644
--- a/target-i386/kvm.c
+++ b/target-i386/kvm.c
@@ -3376,9 +3376,10 @@ static void kvm_update_msi_routes_all(void *private, bool global,
     QLIST_FOREACH(entry, &msi_route_list, list) {
         cnt++;
         msg = pci_get_msi_message(entry->dev, entry->vector);
-        kvm_irqchip_update_msi_route(kvm_state, entry->virq,
-                                     msg, entry->dev);
+        kvm_irqchip_update_msi_route_no_commit(kvm_state, entry->virq,
+                                               msg, entry->dev);
     }
+    kvm_irqchip_commit_routes(kvm_state);
     trace_kvm_x86_update_msi_routes(cnt);
 }
 
-- 
2.4.11

  reply	other threads:[~2016-06-22  3:43 UTC|newest]

Thread overview: 63+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-06-21  7:47 [Qemu-devel] [PATCH v10 00/26] IOMMU: Enable interrupt remapping for Intel IOMMU Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 01/26] x86-iommu: introduce parent class Peter Xu
2016-06-24  7:10   ` [Qemu-devel] [PATCH v10 27/26] intel_iommu: disallow kernel-irqchip=on with IR Peter Xu
2016-06-24  9:20     ` Peter Xu
2016-07-04 15:39       ` Michael S. Tsirkin
2016-07-05  3:51         ` Peter Xu
2016-07-11 10:17     ` David Kiarie
2016-07-11 12:08       ` Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 02/26] x86-iommu: provide x86_iommu_get_default Peter Xu
2016-07-04 15:16   ` Michael S. Tsirkin
2016-07-05  5:11     ` Peter Xu
2016-07-04 15:17   ` Michael S. Tsirkin
2016-07-05  5:12     ` Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 03/26] x86-iommu: q35: generalize find_add_as() Peter Xu
2016-07-04 15:16   ` Michael S. Tsirkin
2016-07-04 16:08     ` Paolo Bonzini
2016-07-04 16:35       ` Michael S. Tsirkin
2016-07-04 16:40         ` Paolo Bonzini
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 04/26] x86-iommu: introduce "intremap" property Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 05/26] acpi: enable INTR for DMAR report structure Peter Xu
2016-07-04 15:14   ` Michael S. Tsirkin
2016-07-05  6:39     ` Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 06/26] intel_iommu: allow queued invalidation for IR Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 07/26] intel_iommu: set IR bit for ECAP register Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 08/26] acpi: add DMAR scope definition for root IOAPIC Peter Xu
2016-07-04 15:22   ` Michael S. Tsirkin
2016-07-05  7:30     ` Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 09/26] intel_iommu: define interrupt remap table addr register Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 10/26] intel_iommu: handle interrupt remap enable Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 11/26] intel_iommu: define several structs for IOMMU IR Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 12/26] intel_iommu: add IR translation faults defines Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 13/26] intel_iommu: Add support for PCI MSI remap Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 14/26] q35: ioapic: add support for emulated IOAPIC IR Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 15/26] ioapic: introduce ioapic_entry_parse() helper Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 16/26] intel_iommu: add support for split irqchip Peter Xu
2016-06-25  8:08   ` Jan Kiszka
2016-06-25 13:18     ` Peter Xu
2016-06-25 15:18       ` Jan Kiszka
2016-06-26  1:48         ` Peter Xu
2016-06-26 13:27           ` Jan Kiszka
2016-06-28  6:10             ` Michael S. Tsirkin
2016-06-28  7:25             ` Peter Xu
2017-01-03  6:15             ` Peter Xu
2017-01-04 10:33               ` Jan Kiszka
2017-01-05  2:21                 ` Peter Xu
2016-07-04 14:32   ` Paolo Bonzini
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 17/26] x86-iommu: introduce IEC notifiers Peter Xu
2016-07-04 14:22   ` Paolo Bonzini
2016-07-05  7:32     ` Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 18/26] ioapic: register IOMMU IEC notifier for ioapic Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 19/26] intel_iommu: Add support for Extended Interrupt Mode Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 20/26] intel_iommu: add SID validation for IR Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 21/26] kvm-irqchip: simplify kvm_irqchip_add_msi_route Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 22/26] kvm-irqchip: i386: add hook for add/remove virq Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 23/26] kvm-irqchip: x86: add msi route notify fn Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 24/26] kvm-irqchip: do explicit commit when update irq Peter Xu
2016-06-22  3:42   ` Peter Xu [this message]
2016-07-04 14:23   ` Paolo Bonzini
2016-07-05  7:35     ` Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 25/26] intel_iommu: support all masks in interrupt entry cache invalidation Peter Xu
2016-06-21  7:47 ` [Qemu-devel] [PATCH v10 26/26] kvm-all: add trace events for kvm irqchip ops Peter Xu
2016-07-04 14:33 ` [Qemu-devel] [PATCH v10 00/26] IOMMU: Enable interrupt remapping for Intel IOMMU Paolo Bonzini
2016-07-04 16:39 ` Michael S. Tsirkin

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1466566965-12809-1-git-send-email-peterx@redhat.com \
    --to=peterx@redhat.com \
    --cc=alex.williamson@redhat.com \
    --cc=davidkiarie4@gmail.com \
    --cc=ehabkost@redhat.com \
    --cc=imammedo@redhat.com \
    --cc=jan.kiszka@web.de \
    --cc=jasowang@redhat.com \
    --cc=marcel@redhat.com \
    --cc=mst@redhat.com \
    --cc=pbonzini@redhat.com \
    --cc=qemu-devel@nongnu.org \
    --cc=rkrcmar@redhat.com \
    --cc=rth@twiddle.net \
    --cc=wexu@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).