[libvirt] dpdk/vpp and cross-version migration for vhost
by Michael S. Tsirkin
Hi!
So it looks like we face a problem with cross-version
migration when using vhost. It's not new but became more
acute with the advent of vhost user.
For users to be able to migrate between different versions
of the hypervisor the interface exposed to guests
by hypervisor must stay unchanged.
The problem is that a qemu device is connected
to a backend in another process, so the interface
exposed to guests depends on the capabilities of that
process.
Specifically, for vhost user interface based on virtio, this includes
the "host features" bitmap that defines the interface, as well as more
host values such as the max ring size. Adding new features/changing
values to this interface is required to make progress, but on the other
hand we need ability to get the old host features to be compatible.
To solve this problem within qemu, qemu has a versioning system based on
a machine type concept which fundamentally is a version string, by
specifying that string one can get hardware compatible with a previous
qemu version. QEMU also reports the latest version and list of versions
supported so libvirt records the version at VM creation and then is
careful to use this machine version whenever it migrates a VM.
One might wonder how is this solved with a kernel vhost backend. The
answer is that it mostly isn't - instead an assumption is made, that
qemu versions are deployed together with the kernel - this is generally
true for downstreams. Thus whenever qemu gains a new feature, it is
already supported by the kernel as well. However, if one attempts
migration with a new qemu from a system with a new to old kernel, one
would get a failure.
In the world where we have multiple userspace backends, with some of
these supplied by ISVs, this seems non-realistic.
IMO we need to support vhost backend versioning, ideally
in a way that will also work for vhost kernel backends.
So I'd like to get some input from both backend and management
developers on what a good solution would look like.
If we want to emulate the qemu solution, this involves adding the
concept of interface versions to dpdk. For example, dpdk could supply a
file (or utility printing?) with list of versions: latest and versions
supported. libvirt could read that and
- store latest version at vm creation
- pass it around with the vm
- pass it to qemu
>From here, qemu could pass this over the vhost-user channel,
thus making sure it's initialized with the correct
compatible interface.
As version here is an opaque string for libvirt and qemu,
anything can be used - but I suggest either a list
of values defining the interface, e.g.
any_layout=on,max_ring=256
or a version including the name and vendor of the backend,
e.g. "org.dpdk.v4.5.6".
Note that typically the list of supported versions can only be
extended, not shrunk. Also, if the host/guest interface
does not change, don't change the current version as
this just creates work for everyone.
Thoughts? Would this work well for management? dpdk? vpp?
Thanks!
--
MST
8 years, 3 months
[libvirt] [PATCH 0/2] Fix some storage vol formatting issues (code and doc)
by John Ferlan
While reviewing another set of changes I tripped across this.
John Ferlan (2):
conf: Format the "nocow" field.
docs: Adjust pool/vol target output
docs/formatstorage.html.in | 16 ++++++++--------
src/conf/storage_conf.c | 2 ++
tests/storagevolxml2xmlout/vol-qcow2-nocow.xml | 1 +
tests/storagevolxml2xmltest.c | 1 +
4 files changed, 12 insertions(+), 8 deletions(-)
--
2.7.4
8 years, 3 months
[libvirt] [PATCH] virsh: maxvcpus: Always fall back to the old command if domain caps fail
by Peter Krempa
1ec22be5 added code that detects the maximum cpu count according to
domain capabilities. The code fell back to the old command only if the
API was not supported. If the API fails for other reasons the command
would fail. There's no point in not trying the old API in such case.
https://bugzilla.redhat.com/show_bug.cgi?id=1402690
---
tools/virsh-host.c | 3 ---
1 file changed, 3 deletions(-)
diff --git a/tools/virsh-host.c b/tools/virsh-host.c
index 2fd3686..24ebde2 100644
--- a/tools/virsh-host.c
+++ b/tools/virsh-host.c
@@ -623,9 +623,6 @@ cmdMaxvcpus(vshControl *ctl, const vshCmd *cmd)
ignore_value(virXPathInt("string(./vcpu[1]/@max)", ctxt, &vcpus));
} else {
- if (last_error && last_error->code != VIR_ERR_NO_SUPPORT)
- goto cleanup;
-
vshResetLibvirtError();
}
--
2.10.2
8 years, 3 months
[libvirt] [PATCH 1/2] vz: set PVMT_DONT_CREATE_DISK migration flag
by Pavel Glushchak
This flag tells backend not to create instance
disks making behavior the same as in qemu driver.
Disk files have to be created beforehand on target
host manually or by upper management layer i.e.
OpenStack Nova.
Signed-off-by: Pavel Glushchak <pglushchak(a)virtuozzo.com>
---
src/vz/vz_sdk.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/vz/vz_sdk.c b/src/vz/vz_sdk.c
index d13c86a..4cd988a 100644
--- a/src/vz/vz_sdk.c
+++ b/src/vz/vz_sdk.c
@@ -4740,7 +4740,7 @@ int prlsdkSwitchToSnapshot(virDomainObjPtr dom, const char *uuid, bool paused)
* connection to dispatcher
*/
-#define PRLSDK_MIGRATION_FLAGS (PSL_HIGH_SECURITY)
+#define PRLSDK_MIGRATION_FLAGS (PSL_HIGH_SECURITY | PVMT_DONT_CREATE_DISK)
int prlsdkMigrate(virDomainObjPtr dom, virURIPtr uri,
const unsigned char *session_uuid,
--
2.7.4
8 years, 3 months
[libvirt] [PATCH 0/3] qemu: bugfixes for websocket port in graphics
by Nikolay Shirokovskiy
Patch 1 is a preparation for patch 2.
Patch 3 particularly fixes next 2 cases:
== A. Can not restore domain with autoconfigured websocket.
domain 1 and 2 have autoconfigured websocket.
1. domain 1 is started then, saved
2. domain 2 is started
3. domain 1 restoration is failed:
error: internal error: qemu unexpectedly closed the monitor: 2016-11-21T10:23:11.356687Z
qemu-kvm: -vnc 0.0.0.0:2,websocket=5700: Failed to start VNC server on `(null)':
Failed to bind socket: Address already in use
== B. Can not migrate domain with autoconfigured websocket.
domain 1 on host A, domain 2 on host B, both have autoconfigured websocket
1. domain 1 started, domain 2 started
2. domain 1 migration to host B is failed with the above error.
Nikolay Shirokovskiy (3):
qemu: refactor: use switch for enum in qemuProcessGraphicsReservePorts
qemu: mark user defined websocket as used
qemu: fix xml dump of autogenerated websocket
src/conf/domain_conf.c | 5 ++++-
src/conf/domain_conf.h | 1 +
src/qemu/qemu_process.c | 45 ++++++++++++++++++++++++++++++++++++---------
3 files changed, 41 insertions(+), 10 deletions(-)
--
1.8.3.1
8 years, 3 months
[libvirt] [PATCH] util: Fix build with clang
by Martin Kletzander
When compiling with clang on Linux, it complains that "passing an
object that undergoes default argument promotion to 'va_start' has
undefined behavior". That is true according to the C standard,
although I couldn't find any mention about enum->int promotion (even
though it's a sensible one). The only fix I came up with was changing
the layer parameter to int so that it does not undergo any default
argument promotion.
Signed-off-by: Martin Kletzander <mkletzan(a)redhat.com>
---
Even though this is a build-breaker, I don't see anyone (or CI)
complaining about it, and the fix doesn't lok that nice, so...
src/util/virfirewall.c | 4 ++--
src/util/virfirewall.h | 4 ++--
2 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/src/util/virfirewall.c b/src/util/virfirewall.c
index 3f976186a671..da4602aa422b 100644
--- a/src/util/virfirewall.c
+++ b/src/util/virfirewall.c
@@ -417,7 +417,7 @@ virFirewallAddRuleFullV(virFirewallPtr firewall,
*/
virFirewallRulePtr
virFirewallAddRule(virFirewallPtr firewall,
- virFirewallLayer layer,
+ int layer,
...)
{
virFirewallRulePtr rule;
@@ -452,7 +452,7 @@ virFirewallAddRule(virFirewallPtr firewall,
* Returns the new rule
*/
virFirewallRulePtr virFirewallAddRuleFull(virFirewallPtr firewall,
- virFirewallLayer layer,
+ int layer,
bool ignoreErrors,
virFirewallQueryCallback cb,
void *opaque,
diff --git a/src/util/virfirewall.h b/src/util/virfirewall.h
index dbf397537fa8..51608b648633 100644
--- a/src/util/virfirewall.h
+++ b/src/util/virfirewall.h
@@ -45,7 +45,7 @@ virFirewallPtr virFirewallNew(void);
void virFirewallFree(virFirewallPtr firewall);
virFirewallRulePtr virFirewallAddRule(virFirewallPtr firewall,
- virFirewallLayer layer,
+ int layer,
...)
ATTRIBUTE_SENTINEL;
@@ -54,7 +54,7 @@ typedef int (*virFirewallQueryCallback)(virFirewallPtr firewall,
void *opaque);
virFirewallRulePtr virFirewallAddRuleFull(virFirewallPtr firewall,
- virFirewallLayer layer,
+ int layer,
bool ignoreErrors,
virFirewallQueryCallback cb,
void *opaque,
--
2.11.0
8 years, 3 months
[libvirt] [PATCH] virFirewallAddRule: fix build with clang
by Ján Tomko
clang complains:
util/virfirewall.c:425:20: error: passing an object that undergoes
default argument promotion to 'va_start' has undefined behavior
[-Werror,-Wvarargs]
__builtin_va_start(args, layer);
Introduce a macro wrapper that puts a NULL pointer after the enum
argument.
---
Previous attempts exchanged the arguments:
https://www.redhat.com/archives/libvir-list/2016-June/msg02173.html
or changed the type to int:
https://www.redhat.com/archives/libvir-list/2016-December/msg00379.html
src/libvirt_private.syms | 2 +-
src/util/virfirewall.c | 9 +++++----
src/util/virfirewall.h | 11 ++++++++---
3 files changed, 14 insertions(+), 8 deletions(-)
diff --git a/src/libvirt_private.syms b/src/libvirt_private.syms
index 867acdb..4c92dba 100644
--- a/src/libvirt_private.syms
+++ b/src/libvirt_private.syms
@@ -1616,8 +1616,8 @@ virFindFileInPath;
# util/virfirewall.h
-virFirewallAddRule;
virFirewallAddRuleFull;
+virFirewallAddRuleInt;
virFirewallApply;
virFirewallFree;
virFirewallNew;
diff --git a/src/util/virfirewall.c b/src/util/virfirewall.c
index 3f97618..c71a274 100644
--- a/src/util/virfirewall.c
+++ b/src/util/virfirewall.c
@@ -416,13 +416,14 @@ virFirewallAddRuleFullV(virFirewallPtr firewall,
* Returns the new rule
*/
virFirewallRulePtr
-virFirewallAddRule(virFirewallPtr firewall,
- virFirewallLayer layer,
- ...)
+virFirewallAddRuleInt(virFirewallPtr firewall,
+ virFirewallLayer layer,
+ void *null,
+ ...)
{
virFirewallRulePtr rule;
va_list args;
- va_start(args, layer);
+ va_start(args, null);
rule = virFirewallAddRuleFullV(firewall, layer, false, NULL, NULL, args);
va_end(args);
return rule;
diff --git a/src/util/virfirewall.h b/src/util/virfirewall.h
index dbf3975..f28a421 100644
--- a/src/util/virfirewall.h
+++ b/src/util/virfirewall.h
@@ -44,11 +44,16 @@ virFirewallPtr virFirewallNew(void);
void virFirewallFree(virFirewallPtr firewall);
-virFirewallRulePtr virFirewallAddRule(virFirewallPtr firewall,
- virFirewallLayer layer,
- ...)
+virFirewallRulePtr virFirewallAddRuleInt(virFirewallPtr firewall,
+ virFirewallLayer layer,
+ void *null,
+ ...)
ATTRIBUTE_SENTINEL;
+/* work around passing an enum to va_start */
+# define virFirewallAddRule(firewall, layer, ...) \
+ virFirewallAddRuleInt(firewall, layer, NULL, __VA_ARGS__)
+
typedef int (*virFirewallQueryCallback)(virFirewallPtr firewall,
const char *const *lines,
void *opaque);
--
2.7.3
8 years, 3 months
[libvirt] RFC: add recreate option to domain events conf
by Nikolay Shirokovskiy
Hi, all.
Does it make sense to anybody else that rebooting and resetting
a persistent domain from outside or from inside should bring upon
pending configuration changes? For this purpose we can add another
option to on_reboot and other events, say 'recreate'. From technical
POV qemu has enough capabilities, namely -no-reboot option.
Nikolay
8 years, 3 months
[libvirt] [PATCH v3 0/4] Gathering network interface statistics with openvswitch
by Mehdi Abaakouk
This new version removes the virstat.c from locale configuration.
It also adds a new commit to autodected the ifname of the ovs vhostuser.
Mehdi Abaakouk (4):
Gathering vhostuser interface stats with ovs
virstat: fix signature of virstat helper
Move virstat.c code to virnetdevtap.c
domain_conf: autodetect vhostuser ifname
po/POTFILES.in | 1 -
src/Makefile.am | 1 -
src/conf/domain_conf.c | 10 +++
src/libvirt_private.syms | 5 +-
src/libxl/libxl_driver.c | 4 +-
src/lxc/lxc_driver.c | 3 +-
src/openvz/openvz_driver.c | 4 +-
src/qemu/qemu_driver.c | 31 +++++--
src/uml/uml_driver.c | 1 -
src/util/virnetdevopenvswitch.c | 106 ++++++++++++++++++++++++
src/util/virnetdevopenvswitch.h | 4 +
src/util/virnetdevtap.c | 143 ++++++++++++++++++++++++++++++++
src/util/virnetdevtap.h | 3 +
src/util/virstats.c | 178 ----------------------------------------
src/util/virstats.h | 31 -------
src/xen/xen_hypervisor.c | 4 +-
16 files changed, 298 insertions(+), 231 deletions(-)
delete mode 100644 src/util/virstats.c
delete mode 100644 src/util/virstats.h
--
2.10.2
8 years, 3 months
[libvirt] [PATCH 0/2] Couple of trivial fixes
by Michal Privoznik
Obviously my gcc do not warn me about some uninitialized or unused variables.
But fortunately we have freebsd build in the CI.
Both patches pushed under trivial rule.
Michal Privoznik (2):
qemuGetDomainHugepagePath: Initialize @ret
vsh: Mark some function arguments as unused
src/qemu/qemu_conf.c | 2 +-
tools/vsh.c | 10 +++++++---
2 files changed, 8 insertions(+), 4 deletions(-)
--
2.11.0
8 years, 3 months
