On Wed, Aug 31, 2016 at 05:10:33PM +0200, Michal Privoznik wrote:
https://bugzilla.redhat.com/show_bug.cgi?id=1366989
QEMU added another virtio-net tunable [1]. It basically allows
users to set the size of RX virtio ring. But because virtio-net
uses two separate ring buffers to pass data from/to guest they
named it explicitly rx_queue_size. We should expose it in our XML
too.
1:
http://lists.nongnu.org/archive/html/qemu-devel/2016-08/msg02029.html
Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com>
---
docs/formatdomain.html.in | 16 ++++++++-
docs/schemas/domaincommon.rng | 5 +++
src/conf/domain_conf.c | 16 +++++++++
src/conf/domain_conf.h | 1 +
src/qemu/qemu_domain.c | 7 ++++
...ml2argv-net-virtio-rxqueuesize-invalid-size.xml | 29 +++++++++++++++
.../qemuxml2argv-net-virtio-rxqueuesize.xml | 29 +++++++++++++++
tests/qemuxml2argvtest.c | 1 +
.../qemuxml2xmlout-net-virtio-rxqueuesize.xml | 41 ++++++++++++++++++++++
tests/qemuxml2xmltest.c | 1 +
10 files changed, 145 insertions(+), 1 deletion(-)
create mode 100644
tests/qemuxml2argvdata/qemuxml2argv-net-virtio-rxqueuesize-invalid-size.xml
create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-net-virtio-rxqueuesize.xml
create mode 100644 tests/qemuxml2xmloutdata/qemuxml2xmlout-net-virtio-rxqueuesize.xml
diff --git a/docs/formatdomain.html.in b/docs/formatdomain.html.in
index 8c15a73..be19bb9 100644
--- a/docs/formatdomain.html.in
+++ b/docs/formatdomain.html.in
@@ -4650,7 +4650,7 @@ qemu-kvm -net nic,model=? /dev/null
<source network='default'/>
<target dev='vnet1'/>
<model type='virtio'/>
- <b><driver name='vhost' txmode='iothread'
ioeventfd='on' event_idx='off' queues='5'>
+ <b><driver name='vhost' txmode='iothread'
ioeventfd='on' event_idx='off' queues='5'
rx_queue_size='256'>
<host csum='off' gso='off' tso4='off'
tso6='off' ecn='off' ufo='off' mrg_rxbuf='off'/>
<guest csum='off' tso4='off' tso6='off'
ecn='off' ufo='off'/>
</driver>
@@ -4766,6 +4766,20 @@ qemu-kvm -net nic,model=? /dev/null
<span class="since">virtio-net since 1.0.6 (QEMU and KVM
only)</span>
<span class="since">vhost-user since 1.2.17 (QEMU and KVM
only)</span>
</dd>
+ <dt><code>rx_queue_size</code></dt>
+ <dd>
+ The optional <code>rx_queue_size</code> attribute controls
+ the size of virtio ring for each queue as described above.
+ The default value is hypervisor dependent and may change
+ across its releases. Moreover, some hypervisors may pose
+ some restrictions on actual value. For instance, latest
I wonder how long this "latest" will be true. Either say "latest as of
yyyy-mm-dd" or QEMU version or just leave the example out.
diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c
index 61f6dbb..05a1227 100644
--- a/src/conf/domain_conf.c
+++ b/src/conf/domain_conf.c
@@ -9517,6 +9519,17 @@ virDomainNetDefParseXML(virDomainXMLOptionPtr xmlopt,
if (q > 1)
def->driver.virtio.queues = q;
}
+ if (rx_queue_size) {
+ unsigned int q;
+ if (virStrToLong_uip(rx_queue_size, NULL, 10, &q) < 0) {
+ virReportError(VIR_ERR_XML_DETAIL,
+ _("'rx_queue_size' attribute must be positive
number: %s"),
+ rx_queue_size);
+ goto error;
+ }
+ if (q > 1)
+ def->driver.virtio.rx_queue_size = q;
Pointless condition, you can assign it unconditionally.
+ }
if ((str = virXPathString("string(./driver/host/@csum)", ctxt))) {
if ((val = virTristateSwitchTypeFromString(str)) <= 0) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
@@ -9697,6 +9710,7 @@ virDomainNetDefParseXML(virDomainXMLOptionPtr xmlopt,
VIR_FREE(ioeventfd);
VIR_FREE(event_idx);
VIR_FREE(queues);
+ VIR_FREE(rx_queue_size);
VIR_FREE(str);
VIR_FREE(filter);
VIR_FREE(type);
@@ -20890,6 +20904,8 @@ virDomainVirtioNetDriverFormat(char **outstr,
}
if (def->driver.virtio.queues)
virBufferAsprintf(&buf, "queues='%u' ",
def->driver.virtio.queues);
+ if (def->driver.virtio.rx_queue_size)
+ virBufferAsprintf(&buf, "rx_queue_size='%u' ",
def->driver.virtio.rx_queue_size);
Long line.
diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c
index c1fb771..29ade0e 100644
--- a/src/qemu/qemu_domain.c
+++ b/src/qemu/qemu_domain.c
@@ -2475,6 +2475,13 @@ qemuDomainDeviceDefValidate(const virDomainDeviceDef *dev,
"not supported by QEMU"));
goto cleanup;
}
+
+ if (STREQ_NULLABLE(net->model, "virtio") &&
+ net->driver.virtio.rx_queue_size &
(net->driver.virtio.rx_queue_size - 1)) {
+ virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
+ _("rx_queue_size has to be a power of two"));
+ goto cleanup;
+ }
No heck for the boundaries that you mentioned in the example? I would
expect that boundaries to be checked when they are written there. I
understand that it doesn't make sense since we can't know them and
adjusting them after every change in qemu is plain wrong. I guess I
just wanted to say I was surprised =)
ACK with those first three things fixed.
Martin