qemu-discuss
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Qemu-discuss] Problem About Multi-Queue Interface


From: Reno Gan
Subject: [Qemu-discuss] Problem About Multi-Queue Interface
Date: Mon, 06 Jan 2014 11:22:59 +0800
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:24.0) Gecko/20100101 Thunderbird/24.2.0

Hi, all,

I set up a testing env for MQ function as reference of http://www.linux-kvm.org/page/Multiqueue

The setup process is fine, however when I check /proc/interrupts, I got this:

 44:          0          0          0          0   PCI-MSI-edge      virtio0-config
 45:          0          0          0          0   PCI-MSI-edge      virtio0-virtqueues
 46:          0          0          0          0   PCI-MSI-edge      virtio1-config
 47:          0          0          0          0   PCI-MSI-edge      virtio1-input.0
 48:          0          0          0          0   PCI-MSI-edge      virtio1-output.0

There is only one IRQ number 45 for "virtio0-virtqueues".  I was hoping that there are multiple "virtio0-virtqueues" with multiple IRQ numbers, for example, 50 for "virtio0-virtqueues.0", 51 for "virtio0-virtqueues.1", etc.

Is this correct or do i miss something?

The below is my test env:

I use libvirt to facilitate the process:
    -  libvirt:   1.2.0
    -  qemu:  1.7.0
    -  host:   ubuntu 12.04.3 with kernel 3.8.0-34 
    -  guest:  ubuntu 13.10 with kernel 3.11.0-14

Libvirt xml segment:
    <interface type='bridge'>
        <source bridge='br1'/>
        <model type='virtio'/>
        <driver name='vhost' queues='8'/>
    </interface>

The generated qemu command line is :
     -netdev tap,fds=17:18:19:20:21:22:23:24,id=hostnet0,vhost=on,vhostfds=25:26:27:28:29:30:31:32 -device virtio-net-pci,mq=on,netdev=hostnet0,id=net0,mac=52:54:00:a3:9e:52,bus=pci.0,addr=0x3

In guest, I enabled multi-queue by command:
    ethtool -L eth0 combined 8

I do noticed that eight rx&tx queues are created in /sys/class/net/eth0/queues from rx-0 to rx-7.

Best Regards,
Reno


reply via email to

[Prev in Thread] Current Thread [Next in Thread]