virCapabilitiesAddGuestDomain() takes an optional binary
name: this is intended for cases where a certain domain
type can't use the default one registered for the guest
architecture, but has to use a special binary instead.

The current code, however, will pass 'binary' again when
'kvmbin' is not defined, which is unnecessary as 'binary'
has been registered as default earlier, and will result
in capabilities output such as

  <emulator>/usr/bin/qemu-system-x86_64</emulator>
  <domain type='qemu'/>
  <domain type='kvm'>
    <emulator>/usr/bin/qemu-system-x86_64</emulator>
  </domain>

with the second <emulator> element providing no additional
information.

Change it so that, when 'kvmbin' is not defined, NULL is
passed and so the default emulator will be used instead.
Signed-off-by: NAndrea Bolognani <abologna@redhat.com>

The function performing the checks, rather than its callers,
should contain comments explaining the rationale behind said
checks.
Signed-off-by: NAndrea Bolognani <abologna@redhat.com>

https://bugzilla.redhat.com/show_bug.cgi?id=1630164

Since 2a13a0a1 we are querying the vhostuser's interface name
when building qemu command line. However, we forgot to do so on
hotplug.
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NAndrea Bolognani <abologna@redhat.com>

Signed-off-by: NFabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

Signed-off-by: NFabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

The `if(!list || list->type != VIR_CONF_LIST)` check couldn't be
written in a 100% similar way. Instead, we're just checking whether
`virConfGetValueStringList() <= 0` and creating a new function to:
- return -1 in case virConfGetValueStringList fails either due to some
  allocation failure or when traversing the list;
- resetting the last error and return 0 otherwise;

Taking this approach we can have the behaviour with the new code as
close as possible to the old one.
Signed-off-by: NFabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

This change actually changes the behaviour of xenConfigGetString() as
now it returns a newly-allocated string.

Unfortunately, there's not much that can be done in order to avoid that
and all the callers have to be changed in order to avoid leaking the
return value.

Also, as a side-effect of the change above, the function now takes a
"char **" argument instead of a "const char **" one.
Signed-off-by: NFabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

Signed-off-by: NFabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

Signed-off-by: NFabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

Commit 12093f1f used %ld instead of %zd for a size_t.
Signed-off-by: NJohn Ferlan <jferlan@redhat.com>

There seems to be no need to add the ignore_value wrapper or
caste with (void) to the unlink() calls, so let's just remove
them. I assume at one point in time Coverity complained. So,
let's just be consistent - those that care to check the return
status can and those that don't can just have the naked unlink.
Signed-off-by: NJohn Ferlan <jferlan@redhat.com>
Reviewed-by: NErik Skultety <eskultet@redhat.com>

This patch is introducing cache monitor(CMT) to cache and
memory bandwidth monitor(MBM) for monitoring CPU memory
bandwidth.

The host capability of the two monitors is also introduced
in this patch.

For CMT, the host capability is shown like:
  <host>
  ...
    <cache>
      <bank id='0' level='3' type='both' size='15' unit='MiB' cpus='0-5'>
        <control granularity='768' min='1536' unit='KiB' type='both' maxAllocs='4'/>
      </bank>
      <monitor level='3' 'reuseThreshold'='270336' maxMonitors='176'>
        <feature name='llc_occupancy'/>
      </monitor>
    </cache>
    ...
  </host>

For MBM, the capability is shown like this:
  <host>
    ...
    <memory_bandwidth>
      <node id='1' cpus='6-11'>
        <control granularity='10' min ='10' maxAllocs='4'/>
      </node>
      <monitor maxMonitors='176'>
        <feature name='mbm_total_bytes'/>
        <feature name='mbm_local_bytes'/>
      </monitor>
    </memory_bandwidth>
    ...
  </host>
Signed-off-by: NWang Huaqiang <huaqiang.wang@intel.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

Move memory bandwidth capability nodes into one data structure,
this allows us to add a monitor for memory bandwidth.
Signed-off-by: NWang Huaqiang <huaqiang.wang@intel.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

Move all cache banks into one data structure, this allows
us to add other cache component, such as cache monitor.
Signed-off-by: NWang Huaqiang <huaqiang.wang@intel.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

This patch introduces the resource monitor and creates the interface
for getting host capability of resource monitor from the system resource
control file system.

The resource monitor takes the role of RDT monitoring group and could be
used to monitor the resource consumption information, such as the last
level cache occupancy and the utilization of memory bandwidth.
Signed-off-by: NWang Huaqiang <huaqiang.wang@intel.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

https://bugzilla.redhat.com/show_bug.cgi?id=1614283

Save the error from the refresh failure because the stopPool
processing may overwrite the error or even worse clear it
due to calling an external libvirt API that resets the last
error such as is the case with the SCSI pool which may call
virGetConnectNodeDev (see commit decaeb28) in order to
process deleting an NPIV vport.
Signed-off-by: NJohn Ferlan <jferlan@redhat.com>
ACKed-by: NMichal Privoznik <mprivozn@redhat.com>

Create a common pool refresh failure handling method as the
same code is repeated multiple times.
Signed-off-by: NJohn Ferlan <jferlan@redhat.com>
ACKed-by: NMichal Privoznik <mprivozn@redhat.com>

Rather than duplicate the error code, let's create an error
label to keep code common.
Signed-off-by: NJohn Ferlan <jferlan@redhat.com>
ACKed-by: NMichal Privoznik <mprivozn@redhat.com>

Alter the code path to remove the need to to go cleanup and thus
remove the label completely.
Signed-off-by: NJohn Ferlan <jferlan@redhat.com>
ACKed-by: NMichal Privoznik <mprivozn@redhat.com>

If the virStoragePoolRefresh fails and we call stopPool, the
code neglected to clean up the state file leading to the next
libvirtd restart attempting to start the pool. For a transient
pool this could make it unexpectedly reappear.
Signed-off-by: NJohn Ferlan <jferlan@redhat.com>
ACKed-by: NMichal Privoznik <mprivozn@redhat.com>

https://bugzilla.redhat.com/show_bug.cgi?id=1607202

It's essentially stated in the nwfilterBindingDelete that we
will allow the admin to shoot themselves in the foot by deleting
the nwfilter binding which then allows them to undefine the
nwfilter that is in use for the running guest...

However, by allowing this we cause a problem for libvirtd
restart reconnect processing which would then try to recreate
the missing binding attempting to use the deleted filter
resulting in an error and thus shutting the guest down.

So rather than keep adding virDomainConfNWFilterInstantiate
flags to "ignore" specific error conditions, modify the logic
to ignore, but VIR_WARN errors other than ignoreExists. This
will at least allow the guest to not shutdown for only nwfilter
binding errors that we can now perhaps recover from since we
have the binding create/delete capability.
Signed-off-by: NJohn Ferlan <jferlan@redhat.com>
ACKed-by: NMichal Privoznik <mprivozn@redhat.com>

All of the ones being removed are pulled in by internal.h. The only
exception is sanlock which expects the application to include <stdint.h>
before sanlock's headers, because sanlock prototypes use fixed width
int, but they don't include stdint.h themselves, so we have to leave
that one in place.
Signed-off-by: NErik Skultety <eskultet@redhat.com>
Acked-by: NMichal Privoznik <mprivozn@redhat.com>

It doesn't really make sense for us to have stdlib.h and string.h but
not stdio.h in the internal.h header.
Signed-off-by: NErik Skultety <eskultet@redhat.com>
Acked-by: NMichal Privoznik <mprivozn@redhat.com>

So, when trying to add some secret util sources, we referenced them with
a non-existent symbol.
Signed-off-by: NErik Skultety <eskultet@redhat.com>

Introduced in QEMU by commit v3.0.0-156-g8a11c62da9.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

QEMU commits:

    e37a5c7fa4 (v2.12.0)
        i386: Add Intel Processor Trace feature support

    c2f193b538 (v2.7.0)
        target-i386: Add support for UMIP and RDPID CPUID bits

    aff9e6e46a (v2.12.0)
        x86/cpu: Enable new SSE/AVX/AVX512 cpu features

    f77543772d (v2.9.0)
        x86: add AVX512_VPOPCNTDQ features

    5131dc433d (v3.1.0)
        i386: Add CPUID bit for PCONFIG

    59a80a19ca (v3.1.0)
        i386: Add CPUID bit for WBNOINVD
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

When restoring a domain from a compressed image, we launch an
intermediate process for decompressing the saved data. If QEMU fails to
load the data for some reason, we force close the stdin/stdout file
descriptors of the intermediate process and wait for it to die. However,
virCommandWait can report various errors which would overwrite the real
error from QEMU. Thus instead of getting something useful:

    internal error: process exited while connecting to monitor:
    2018-09-17T15:17:29.998910Z qemu-system-x86_64: can't apply global
    Skylake-Client-x86_64-cpu.osxsave=off: Property '.osxsave' not found

we could get an irrelevant error message:

    internal error: Child process (lzop -dc --ignore-warn) unexpected
    fatal signal 13
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

Signed-off-by: NJiri Denemark <jdenemar@redhat.com>
Reviewed-by: NPavel Hrdina <phrdina@redhat.com>

Lock all the paths we want to relabel to mutually exclude other
libvirt daemons.

The only hitch here is that directories can't be locked.
Therefore, when relabeling a directory do not lock it (this
happens only when setting up some domain private paths anyway,
e.g. huge pages directory).
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

So far the whole transaction handling is done
virSecuritySELinuxSetFileconHelper(). This needs to change for
the sake of security label remembering and locking. Otherwise we
would be locking a path when only appending it to transaction
list and not when actually relabelling it.
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

Firstly, the following code pattern is harder to follow:

  if (func() < 0) {
      error();
  } else {
      /* success */
  }

We should put 'goto cleanup' into the error branch and move the
else branch one level up.
Secondly, 'rc' should really be named 'ret' because it holds
return value of the function. Not some intermediate value.
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

This label is used in both successful and error paths. Therefore
it should be named 'cleanup' and not 'err'.
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

Lock all the paths we want to relabel to mutually exclude other
libvirt daemons.

The only hitch here is that directories can't be locked.
Therefore, when relabeling a directory do not lock it (this
happens only when setting up some domain private paths anyway,
e.g. huge pages directory).
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

Firstly, the message that says we're setting uid:gid shouldn't be
called from virSecurityDACSetOwnershipInternal() because
virSecurityDACRestoreFileLabelInternal() is calling it too.
Secondly, there are places between us reporting label restore and
us actually doing it where we can quit. Don't say we're doing
something until we are actually about to do it.
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

So far the whole transaction handling is done
virSecurityDACSetOwnershipInternal(). This needs to change for
the sake of security label remembering and locking. Otherwise we
would be locking a path when only appending it to transaction
list and not when actually relabeling it.
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

Two new APIs are added so that security driver can lock and
unlock paths it wishes to touch. These APIs are not for other
drivers to call but security drivers (DAC and SELinux). That is
the reason these APIs are not exposed through our
libvirt_private.syms file.

Three interesting things happen in this commit. The first is the
global @lockManagerMutex. Unfortunately, this has to exist so that
there is only one thread talking to virtlockd at a time. If there
were more threads and one of them closed the connection
prematurely, it would cause virtlockd killing libvirtd. Instead
of complicated code that would handle that, let's have a mutex
and keep the code simple.

The second interesting thing is keeping connection open between
lock and unlock API calls. This is achieved by duplicating client
FD and keeping it open until unlock is called. This trick is used
by regular disk content locking code when the FD is leaked to
qemu.

Finally, the third thing is polling implemented at client side.
Since virtlockd has only one thread that handles locking
requests, all it can do is either acquire lock or error out.
Therefore, the polling has to be implemented in client. The
polling is capped at 60 second timeout, which should be plenty
since the metadata lock is held only for a fraction of a second.
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

Now that we know what metadata lock manager user wishes to use we
can load it when initializing security driver. This is achieved
by adding new argument to virSecurityManagerNewDriver() and
subsequently to all functions that end up calling it.

The cfg.mk change is needed in order to allow lock_manager.h
inclusion in security driver without 'syntax-check' complaining.
This is safe thing to do as locking APIs will always exist (it's
only backend implementation that changes). However, instead of
allowing the include for all other drivers (like cpu, network,
and so on) allow it only for security driver. This will still
trigger the error if including from other drivers.
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

This config option allows users to set and enable lock manager
for domain metadata. The lock manager is going to be used by
security drivers to serialize each other when changing a file
ownership or changing the SELinux label. The only supported lock
manager is 'lockd' for now.
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

In some cases we might want to not load the lock driver config.
Alter virLockManagerPluginNew() and the lock drivers to cope with
this fact.
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

Soon there will be a virtlockd client that wants to either lock
all the resources or none (in order to avoid virtlockd killing
the client on connection close). Because on the RPC layer we can
only acquire one resource at a time, we have to perform a
rollback once we hit a resource that can't be acquired.
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>