vz: keep subscription to performance events thru domain lifetime
The approach of subscribing on first stat API call and then waiting
for receiving of performance event from sdk to process the call originates
in times when every vz libvirt connections spawns its own sdk connection.
Thus without this waiting virsh stat call would return empty stats. Now
with single sdk connection this scheme is unnecessary complicated.
This patch subscribes to performance events on first domain appearence
and unsubscribe on its removing.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com> Signed-off-by: Maxim Nestratov <mnestratov@virtuozzo.com>
We have the code for attaching redirdevs for ages now.
Unfortunately, our monitor code that handles talking to the qemu
process was missing a little piece of code that actually enabled
the feature.
BTW: it really is called "type" on the monitor, even though it's
called "name" on the cmd line. Don't ask.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
virt-login-shell: add ability to join the container cgroups
Prior to joining the namespaces of the container, move the
process into the containers' cgroups, so that the shell that
is subsequently launched is under the container resource
constraints.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
virt-login-shell: add ability to auto-detect shell from container
Currently the shell must be looked up from the config setting in
/etc/libvirt/virt-login-shell.conf. This is inflexible if there
are containers where different users need different shells. Add
add a new 'auto-shell' config parameter which instructs us to
query the containers' /etc/passwd for the shell to be exec'd.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The virt-login-shell environment will be initialized with
an arbitrary number of environment variables determined
by the SSH daemon and PAM configuration. Most of these are
not relevant inside the container, and at best they are
noise and at worst they'll break apps. For example if
XDG_RUNTIME_DIR is leaked to the container, it'll break
any apps using it, since the directory it points to is
only visible to the host OS filesystem, not the container
FS.
Use clearenv() to blank out everything and then set known
good values for PATH, SHELL, USER, LOGNAME HOME and TERM.
Everything else is left up to the login shell to initialize.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
virt-login-shell: change way we request a login shell
Currently we request a login shell by passing the -l argument
to the shell. This is either hardcoded, or required to be
specified by the user in the virt-login-shell.conf file.
The standard way for login programs to request a shell run
as a login shell is to modify the argv passed to execve()
so that argv[0] contains the relative shell filename
prefixed with a zero. eg instead of doing
virt-login-shell: honour the -c option to launch commands
The virt-login-shell program is supposed to look like a
regular shell to clients. Login services like sshd
expect the shell to accept a '-c cmdstring' argument to
specify a command to launch instead of presenting an
interactive prompt.
We can implement this by simply passing the '-c cmdstring'
data straight through to the real shell we use. This does
not open any security holes, since the command is not run
until we're inside the container namespaces. This allows
scp to work for users with virt-login-shell.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
virsh: make lxc-enter-namespace also join the cgroups
Extend the lxc-enter-namespace command so that it joins the
containers' cgroups before starting new namespaces. This
ensures that the commands run have the normal resource
limits applied
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
nodeinfo: remove FreeBSD specific code for getting memory
The GNULIB physmem module already provides support for
the FreeBSD platform, so there's no reason to re-implement
FreeBSD portability code in libvirt. If there are bugs in
the GNULIB code, we should fix GNULIB rather than workaround
it in libvirt.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
nodeinfo: make nodeGetInfo() call nodeGetMemory for memory size
The nodeGetInfo() method currently has its own code for getting
memory size in KB, that basically just re-invents what nodeGetMemory
already does. Remove it and just call nodeGetMemory, converting its
result from bytes to KB, allowing removal of more platform specific
conditional code.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Nearly all the methods in the nodeinfo file are given a
'const char *sysfs_prefix' parameter to override the
default sysfs path (/sys/devices/system). Every single
caller passes in NULL for this, except one use in the
unit tests. Furthermore this parameter is totally
Linux-specific, when the APIs are intended to be cross
platform portable.
This removes the sysfs_prefix parameter and instead gives
a new method linuxNodeInfoSetSysFSSystemPath for use by
the test suite.
For two of the methods this hardcodes use of the constant
SYSFS_SYSTEM_PATH, since the test suite does not need to
override the path for thos methods.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
If you want to set block device I/O tuning values that end with '_max'
and there is nothing else set, libvirt emits an error. In particular:
error: internal error: Unexpected error
That's an unknown error. That is because *_max values depend on their
respective non-_max values. QEMU even says that in the error message
sent as a response to the monitor command:
Pavel Hrdina [Wed, 18 May 2016 06:33:28 +0000 (08:33 +0200)]
spice: introduce listen type none
This new listen type is currently supported only by spice graphics.
It's introduced to make it easier and clearer specify to not listen
anywhere in order to start a guest with OpenGL support.
The old way to do this was set spice graphics autoport='no' and don't
specify any ports. The new way is to use <listen type='none'/>. In
order to be able to migrate to old libvirt the migratable XML will be
generated without the listen element and with autoport='no'. Also the
old configuration will be automatically converted to the this listen
type.
Pavel Hrdina [Wed, 8 Jun 2016 13:18:25 +0000 (15:18 +0200)]
vnc: add support for listen type 'socket'
VNC graphics already supports sockets but only via 'socket' attribute.
This patch coverts that attribute into listen type 'socket'.
For backward compatibility we need to handle listen type 'socket' and 'socket'
attribute properly to support old XMLs and new XMLs. If both are provided they
have to match, if only one of them is provided we need to be able to parse that
configuration too.
To not break migration back to old libvirt if the socket is provided by user we
need to generate migratable XML without the listen element and use only 'socket'
attribute.
Pavel Hrdina [Wed, 8 Jun 2016 11:24:52 +0000 (13:24 +0200)]
vnc: rename socketAutogenerated to socketFromConfig
Even though it's auto-generated it's based on qemu.conf option and listen type
address already uses "fromConfig" to carry this information. Following commits
will convert the socket to listen element so this rename is required because
there will be also an option to get socket auto-generated independently on the
qemu.conf option.
qemu: Generate channel target paths on hotplug as well
Since commit 714080791778e3dfbd484ccb3953bffd820b8ba9, qemu agent
channel cannot be plugged in because we won't generate its path
automatically. Let's not only fix that, but also add tests for it so
next time it's checked for.
Until now, the only hot thing in this test was the name. That's because
we set the id to '-1' before every test. With this change, we test the
hotplug on live domains as the name suggests and as it should be.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Jiri Denemark [Thu, 2 Jun 2016 08:17:22 +0000 (10:17 +0200)]
cputest: Rename nehalem-force to penryn-force
The actual CPU model in the data files is Penryn which makes the file
name look rather strange. Well, one of them contains Nehalem, but that's
a bug which will be fixed soon.
Jiri Denemark [Wed, 1 Jun 2016 13:40:09 +0000 (15:40 +0200)]
cpu: Sort CPU map features on eax_in
As a side effect this changes the order of CPU features in XMLs
generated by libvirt, but that's not a big deal since the order there is
insignificant.
Jiri Denemark [Fri, 20 May 2016 08:59:13 +0000 (10:59 +0200)]
cpu_x86: Prepare for ecx_in CPUID parameter
CPUID instruction normally takes its parameter from EAX, but sometimes
ECX is used as an additional parameter. This patch prepares the x86 CPU
driver code for the new 'ecx_in' CPUID parameter.
Jiri Denemark [Wed, 1 Jun 2016 13:57:00 +0000 (15:57 +0200)]
tests: Add CPU detection tests
So far we only test CPUID -> CPU def conversion on artificial CPUID data
computed from another CPU def. This patch adds the infrastructure to
test this conversion on real data gathered from a host CPU and two
helper scripts for adding new test data:
- cpu-gather.sh runs cpuid tool and qemu-system-x86_64 to get CPUID data
from the host CPU; this is what users can be asked to run if they run
into an issue with host CPU detection in libvirt
- cpu-parse.sh takes the data generated by cpu-gather.sh and creates
data files for CPU detection tests
The CPUID data queried from QEMU will eventually switch to the format
used by query-host-cpu QMP command once QEMU implements it. Until then
we just spawn QEMU with -cpu host and query the guest CPU in QOM. They
should both provide the same CPUID results, but query-host-cpu does not
require any guest CPU to be created by QEMU.
Jiri Denemark [Tue, 7 Jun 2016 10:09:41 +0000 (12:09 +0200)]
cpu_x86: Refactor internal KVM features
The internal features are only used in explicit checks with
cpuHasFeature. Loading them into the CPU map is dangerous since the
features may accidentally be reported to users when decoding CPUID data.
Jiri Denemark [Tue, 7 Jun 2016 10:04:13 +0000 (12:04 +0200)]
cpu_ppc64: Avoid unnecessary pointer to virCPUppc64Data
virCPUData and struct ppc64_model structures contained a pointer to
virCPUppc64Data, which was not very nice since the real data were
accessible by yet another level of pointers from virCPUppc64Data.
Jiri Denemark [Tue, 7 Jun 2016 07:38:53 +0000 (09:38 +0200)]
cpu_x86: Avoid unnecessary pointers to virCPUx86Data
virCPUData, virCPUx86Feature, and virCPUx86Model all contained a pointer
to virCPUx86Data, which was not very nice since the real CPUID data were
accessible by yet another pointer from virCPUx86Data. Moreover, using
virCPUx86Data directly will make static definitions of internal CPU
features a bit easier.
Jiri Denemark [Mon, 6 Jun 2016 14:55:05 +0000 (16:55 +0200)]
tests: Create simple monitor in qemuMonitorTestNewFromFile
The current version uses the first JSON reply from the file as monitor
greeting. With the new parameter the caller can now request a simple
test monitor to be created, which uses an artificial greeting and uses
all JSON strings from the file as regular replies.
Jiri Denemark [Fri, 20 May 2016 07:48:21 +0000 (09:48 +0200)]
cpu_x86: Rename CPUID function to eax_in
CPUID instruction normally takes its parameter from EAX, but sometimes
ECX is used as an additional parameter. Let's rename 'function' to
'eax_in' in preparation for adding 'ecx_in'.
Jiri Denemark [Wed, 1 Jun 2016 08:55:36 +0000 (10:55 +0200)]
cpu_x86: Propagate vendor to guest's virCPUData
When computing CPU data for a given guest CPU we should set CPUID vendor
bits appropriately so that we don't lose the vendor when transforming
CPU data back to XML description.
projects / libvirt.git / log