libvirt's qemu driver doesn't have direct access to the config on the
guest side of a network interface, and currently doesn't have any
method in place to even inform the guest of the desired config. In the
future, an unenforceable attempt to set the guest-side IP info could
be made by adding a static host entry to the appropriate dnsmasq
configuration (or changing the default dhcp client address on the qemu
commandline for type='user' interfaces), or enhancing the guest agent
to allow setting an IP address, but for now it can't have any effect,
and we don't want to give the illusion that it does.

To prevent the "disappearance" of any existing configs with ip
address/route info (due to parser failure), this check is added in the
newly implemented qemuDomainDeviceDefValidate(), which is only called
when a domain is defined or started, *not* when it is reread from disk
at libvirtd startup.

All the same information was already there, just in slightly different
places in the virDomainNetDef.

a.k.a. <hostdev mode='capabilities' type='net'>.

This replaces the existing nips, ips, nroutes, and routes with a
single virNetDevIPInfo, and simplifies the code by calling that
object's parse/format/clear functions instead of open coding.

There are currently two places in the domain where this combination is
used, and there is about to be another. This patch puts them together
for brevity and uniformity.

As with the newly-renamed virNetDevIPAddr and virNetDevIPRoute
objects, the new virNetDevIPInfo object will need to be accessed by a
utility function that calls low level Netlink functions (so we don't
want it to be in the conf directory) and will be called from multiple
hypervisor drivers (so it can't be in any hypervisor directory); the
most appropriate place is thus once again the util directory.

The parse and format functions are in conf/domain_conf.c because only
the domain XML (i.e. *not* the network XML) has this exact combination
of IP addresses plus routes. Note that virDomainNetIPInfoFormat() will
end up being the only caller to virDomainNetRoutesFormat() and
virDomainNetIPsFormat(), so it will just subsume those functions in a
later patch, but we can't do that until they are no longer called.

(It would have been nice to include the interface name within the
virNetDevIPInfo object (with a slight name change), but that can't
be done cleanly, because in each case the interface name is provided
in a different place in the XML relative to the routes and IP
addresses, so putting it in this object would actually make the code
more confused rather than simpler).

These functions all need to be called from a utility function that
must be located in the util directory, so we move them all into
util/virnetdevip.[ch] now that it exists.

Function and struct names were appropriately changed for the new
location, but all code is unchanged aside from motion and renaming.

This patch splits virnetdev.[ch] into multiple files, with the new
virnetdevip.[ch] containing all the functions related to setting and
retrieving IP-related info for a device (both addresses and routes).

When support for <interface type='ethernet'> was added in commit
9a4b705f back in 2010, it erroneously looked at <source dev='blah'/>
for a user-specified guest-side interface name. This was never
documented though. (that attribute already existed at the time in the
data.ethernet union member of virDomainNetDef, but apparently had no
practical use - it was only used as a storage place for a NetDef's
bridge name during qemuDomainXMLToNative(), but even then that was
never used for anything).

When support for similar guest-side device naming was added to the lxc
driver several years later, it was put in a new subelement <guest
dev='blah'/>.

In the intervening years, since there was no validation that
ethernet.dev was NULL in the other drivers that didn't actually use
it, innocent souls who were adding other features assuming they needed
to account for non-NULL ethernet.dev when really they didn't, so
little bits of the usual pointless cargo-cult code showed up.

This patch not only switches the openvz driver to use the documented
<guest dev='blah'/> notation for naming the guest-side device (just in
case anyone is still using the openvz driver), and logs an error if
anyone tries to set <source dev='blah'/> for a type='ethernet'
interface, it also removes the cargo-cult uses of ethernet.dev and
<source dev='blah'/>, and eliminates if from the RNG and from
virDomainNetDef.

NB: I decided on this course of action after mentioning the
inconsistency here:

  https://www.redhat.com/archives/libvir-list/2016-May/msg02038.html

and getting encouragement do eliminate it in a later IRC discussion
with danpb.

in qemuConnectDomainXMLToNative. This function was only accounting for
about 1/10 of all the allocated items in the NetDef prior to memseting
it to all 0's. On top of that, it was going to great pains to learn
the name of the bridge device, but then never doing anything useful
with it (just putting it into data.ethernet.dev, which is *never* used
when building a qemu commandline). (I think this again all started off
as code with good intentions, but it was never completed, and instead
was just Frankensteinically cargo-culted into the odd mish mash we
have today).

The resulting code is much simpler, produces exactly the same output,
and doesn't leak memory.

This patch removes the expanded and duplicated code that all sprung
out of two well-intentioned-but-useless settings of
net->data.(bridge|ethernet).ipaddr.

qemu has never supported even a single IP address in the interface
config, much less a list of them. All of the instances of "clearing
out the IP addresses" that are now in this function originated with
commit d8dbd6 "Basic domain XML conversions for Xen/QEMU drivers" in
May 2009, but even then the single "ipaddr" in the struct for
type='ethernet' and type='bridge' wasn't used in the qemu driver (only
in xen and openvz). Since then anyone who added a new interface type
also tacked on another unnecessary clearing of ipaddr, and when it was
made into a list of IPs (so far supported only by the LXC driver) this
simple setting was turned into a loop (well, multiple loops) to clear
them all.

We need to clear these out without freeing the object completely.

Commit c9a641 (first appearred in 1.2.12) added support for setting
the guest-side IP address of veth devices in lxc domains.
Unfortunately, it hardcoded the assumption that the proper prefix for
any IP address with no explicit prefix in the config should be "24";
that is only correct for class C IPv4 addresses, but not for any other
IPv4 address, nor for any IPv6 address.

The good news is that there is already a function in libvirt that will
determine the proper default prefix for any IP address. This patch
replaces the use of the ill-fated VIR_SOCKET_ADDR_DEFAULT_PREFIX with
calls to virSocketAddrGetIPPrefix().

lxcContainerRenameAndEnableInterfaces() isn't making a copy of the
interface's ifname_guest (into newname), it's just copying the pointer
to it. This means that when it later calls VIR_FREE(newname), it's
actually freeing up (and fortunately NULLing out, so at least we don't
try to access free'd memory) netDef->ifname_guest.

There are times when we don't have a netmask pointer to give to
virSocketAddrGetIPPrefix() (e.g. the IP addresses in domain interfaces
only have a prefix, no netmask), but it would have caused a segv if we
called it with NULL instead of a pointer to a netmask. This patch
qualifies the code that would use the netmask or address pointers to
check for NULL first.

Now that we can include <interface type='ethernet'> in tests, we could
almost test XML that has an <ip> element in an interface. Except that
the test fails when it tries to actually set the IP address for the
interface's tap device. This patch mocks virNetDevSetIPAddress() to
just return success.

Rearrange this function to be better organized and more correct:

* the error codes were changed from the incorrect INVALID_ARG to
  XML_ERROR

* prefix still isn't required, but if present it must be valid or an
  error will be logged.

* don't emit a debug log just because prefix is missing - this
  is valid.

* group everything related to setting prefix in one place rather than
  scattered through the function.

I'm tired of mistyping this all the time, so let's do it the same all
the time (similar to how we changed all "Pci" to "PCI" awhile back).

(NB: I've left alone some things in the esx and vbox drivers because
I'm unable to compile them and they weren't obviously *not* a part of
some API. I also didn't change a couple of variables named,
e.g. "somethingIptables", because they were derived from the name of
the "iptables" command)

These had been declared in conf/device_conf.h, but then used in
util/virnetdev.c, meaning that we had to #include conf/device_conf.h
in virnetdev.c (which we have for a long time said shouldn't be done.

This caused a bigger problem when I tried to #include util/virnetdev.h
in a file in src/conf (which is allowed) - for some reason the
"device_conf.h: File not found" error.

The solution is to move the data types and functions used in util
sources from conf to util. Some names were adjusted during the move
("virInterface" --> "virNetDevIf", and "VIR_INTERFACE" -->
"VIR_NETDEV_IF")

virNetDevLinkDump should have been in virnetlink.c, but that file
didn't exist yet when the function was created. It didn't really
matter until now - I found that having virnetlink.h included by
virnetdev.h caused build problems when trying to #include virnetdev.h
in a .c file in src/conf (due to missing directory in -I). Rather than
fix that to further institutionalize the incorrect placement of this
one function, this patch moves the function.

With respect to to the following thread
https://www.redhat.com/archives/libvir-list/2016-June/msg01822.html, until we
introduce a new rpm package '-libs' that would allow us to drop daemon's
dependency on the client package, distribute admin API related stuff within
the client package (since it's the best analogy to the virsh client).
Signed-off-by: NErik Skultety <eskultet@redhat.com>

Some of the examples make use of asprintf and strtol functions (to keep
things simple) which are prohibited to use within our code (enforced by
syntax-check). Therefore besides adding some examples, this patch also updates
cfg.mk to exclude examples directory from asprintf and strtol rules, as well as
updates .gitignore to exclude all the new admin binaries created in the
'examples' dir.
Signed-off-by: NErik Skultety <eskultet@redhat.com>

This patch enables admin socket creation in daemon's code, bumps the library
version in libvirt_admin_public.syms, and performs all necessary modifications
to our makefiles so that admin API can finally be included in the tarball,
and eventually become part of an rpm package (a patch later in this series).
Signed-off-by: NErik Skultety <eskultet@redhat.com>

We need this because apply graphics functions is used on
update too. Also in case of NULL address resolve it to default
instead of error.
Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>

Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>

Allocation will do this job. Also we don't use the explicit setting
in other places.
Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>

Move graphic device config to post parse. This way we
detect error on early stage and leverage checking on detach too.
Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>

Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>

First we need to always set value to vz sdk parameter so
we can leverage setting code for device updates. This patch
resolves tristate default to off implicitly. This is easier
then extract default value from vz sdk itself. First current
default is off too, second this approach is already taken
for 'net->linkstate'.

Second dump this option in domain xml.
Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>

Fix net->data usage accordingly to type field.
Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>

Current code that pass gateways to vz sdk is not suitable for
updates. If update has no gateways while we had them before
we need to pass "" for vz sdk gateways to reset old value.

The code definitely deserves its own function.

Drop checks that skip setting gateways if network address
is not set. Such a configuration is possible in vz sdk.
Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>

Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>

Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>

Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>

Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>

And reformat so that we don't have lengthy lines. Also simplify
some checks.
Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>

This is as easy as moving disks checks from domain post
parse callback to device post parse callback.
Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>

Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
Signed-off-by: NMaxim Nestratov <mnestratov@virtuozzo.com>

Attach/detach functions for disk/net are quite trivial and
typically call a few functions in begin/end edit frame. Having
in mind update function too adding configuring for another
device (like graphics) will introduce 3 trivial functions more.
Let's replace current approach by attach/detach functions for
device.
Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
Signed-off-by: NMaxim Nestratov <mnestratov@virtuozzo.com>

Current implementation works with hard disks only. This patch
adds support for any disk device (cdroms and hdds right now).
Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
Signed-off-by: NMaxim Nestratov <mnestratov@virtuozzo.com>

This code was added as a part of huge patch that moves driver
from working with prlctl to vz sdk so there is no good explanation
why this is done this way. The problem that it is not correct.
vz sdk cache mode parameter affects all domain disks while this hunk
resets its on every disk to a new value.
Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>

Some Intel processor families (e.g. the Intel Xeon processor E5 v3
family) introduced some PQos (Platform Qos) features, including CMT
(Cache Monitoring technology) and MBM (Memory Bandwidth Monitoring),
to monitor or control shared resource. This patch add them into x86
part of cpu_map.xml to be used for applications based on libvirt to
get cpu capabilities. For example, Nova in OpenStack schedules guests
based on the CPU features that the host has.
Signed-off-by: NQiaowei Ren <qiaowei.ren@intel.com>