* src/qemu/qemu_conf.c: change the error message to refer to
  'cgroup_controllers' instead of 'cgroup_device_acl'

We should always be using virGetHostname in place of
gethostname; thus add in a new syntax-check rule to make
sure no new uses creep in.
Signed-off-by: NChris Lalancette <clalance@redhat.com>

We can slightly tighten up the regex's used to detect the use of
nonreentrant functions.  We can also check src/util/virterror.c
by modifying a comment; I think it's worth it to get the additional
coverage.
Signed-off-by: NChris Lalancette <clalance@redhat.com>

Signed-off-by: NChris Lalancette <clalance@redhat.com>

xenUnifiedDomainEventRegister() calls out to
virDomainEventCallbackListAdd(), which increments the reference
count on the connection.  That is fine, but then
xenUnifiedDomainEventRegister() increments the usage count again,
leading to a usage count leak.  Remove the increment in the xen
register, and the UnrefConnect in the xen unregister.
Signed-off-by: NChris Lalancette <clalance@redhat.com>

If no matching device was found (cap == NULL) then no strdup() call
was made and *wwnn and *wwpn are untouched. Checking them for NULL
in this situation may result in reporting an false-positive OOM error
because *wwnn and *wwpn may be initialized to NULL by the caller.

Only check *wwnn and *wwpn for NULL if a matching device was found
(cap != NULL) and thus strdup() was called.

* src/conf/node_device_conf.c: only report an OOM error if there
  really is one

virXPathNodeSet() could return -1 when doing an evaluation failure
due to xmlXPathEval() from libxml2 behaviour.
* src/util/xml.c: make sure we always return 0 unless the returned
  XPath type is of the wrong type (meaning the query passed didn't
  evaluate to a node set and code must be fixed)

This adds simple support for configuring a guest with a QEMU/KVM
virtual hardware watchdog device.

https://bugzilla.redhat.com/show_bug.cgi?id=528575
virsh -c lxc:/// autostart vm1
was crashing the daemon

* src/lxc/lxc_conf.h src/lxc/lxc_conf.c: initialize the driver
  autostartDir to avoid a NULL reference and implement autostart for LXC

* src/Makefile.am: add selinux CFLAGS and linker flags to security driver

Currently MAC address configuration of container veth is just ignored.
This patch implements the missing feature.

* src/lxc/veth.c, src/lxc/veth.h: add setMacAddr
* src/lxc/lxc_driver.c: set macaddr of container veth if specified

These will be used by the test driver, so move them to a shareable space.

Certain error paths won't unlock the node device object.

Most of the hash iterators need to modify either payload of
data args. The const annotation prevents this.

* src/util/hash.h, src/util/hash.c: Remove const-ness from
  virHashForEach/Iterator
* src/xen/xm_internal.c: Remove bogus casts

* src/libvirt.c: remove reverences to non existent dconn parameter
  when using that entry point.

A cgroup file returns integer value terminated with '\n' and remaining
it has sometimes harmful effects, for example it leads virStrToLong_ull
to fail.
* src/util/cgroup.c: strip out terminating \n when reading a value

If the the qemu and kvm binaries are the same, we don't include machine
types in the kvm domain info.

However, the code which refreshes the machine types info from the
previous capabilities structure first looks at the kvm domain's info,
finds it matches and then copies the empty machine types list over
for the top-level qemu domain.

That doesn't make sense, we shouldn't copy an empty machin types list.

* src/qemu/qemu_conf.c: qemudGetOldMachinesFromInfo(): don't copy an
  empty machine types list.

* src/util/buf.c: if virBufferEscapeString was called on a buffer that
  had 0 bytes of space, a size of -1 will be passed to snprintf, resulting
  in a segmentation fault, this preallocate some space.

* src/conf/storage_conf.c src/conf/storage_conf.h: extend the enums
  and values
* docs/schemas/storagepool.rng: add to the list of storage pool type
  formats

* src/lxc/lxc_driver.c: without PATH, the controller will definitely
  fail to call ip command

Normally, when you migrate a domain from host A to host B,
the domain on host A remains defined but shutoff and the domain
on host B remains running but is a "transient".  Add a new
flag to virDomainMigrate() to allow the original domain to be
undefined on source host A, and a new flag to virDomainMigrate() to
allow the new domain to be persisted on the destination host B.
Signed-off-by: NChris Lalancette <clalance@redhat.com>

When specifying bridge delay via network XML define, we were looking for
the 'delay' attribute, but would dump the value as 'forwardDelay'. Have
the output match the expected input (and schema).

The fread_file_lim() function uses fread() but never handles
EINTR results, causing unexpected failures when reading QEMU
help arg info. It was unneccessarily using FILE * instead
of plain UNIX file handles, which prevented use of saferead()

* src/util/util.c: Switch fread_file_lim over to use saferead
  instead of fread, remove FILE * use, and rename

We were missing a closing tag, so the XML wasn't proper.

$ sudo virsh pool-start idontexist
10:58:18.716: warning : processCallDispatchReply:7612 : Method call error
error: failed to get pool 'idontexist'
error: Storage pool not found: no pool with matching name 'idontexist'

That warning doesn't server much purpose being printed via a virsh call. So
remove the message.

This matches the behavior of domains, storage, etc. Also, fix up some
whitespace issues in the area.

The logic for running the decompression programs was broken in
commit f2387093, so that for
non-raw formats the decompression program was never run, and
for raw formats, it tried to exec an argv[] with initial NULL
in the program name.

* src/qemu/qemu_driver.c: Fix logic in runing decompression program

In lxcVmStart we first should not reuse rc for another purpose
than the return value.

* src/lxc/lxc_driver.c: don't reuse rc for local procedure calls

If one has e.g.

  <guest>
    <os_type>hvm</os_type>
    <arch name='x86_64'>
      <wordsize>64</wordsize>
      <emulator>/usr/bin/qemu-system-x86_64</emulator>
      <machine>pc-0.11</machine>
      <machine canonical='pc-0.11'>pc</machine>
      <machine>pc-0.10</machine>
      <machine>isapc</machine>
      <domain type='qemu'>
      </domain>
      <domain type='kvm'>
        <emulator>/usr/bin/kvm</emulator>
        <machine>pc</machine>
        <machine>isapc</machine>
      </domain>
    </arch>
  </guest>

and start a guest with:

  <domain type='kvm'>
    ...
    <os>
      <type arch='x86_64'>hvm</type>
      ...
    </os>
  </domain>

then the default machine type should be 'pc' and not 'pc-0.11'

Issue was reported by Anton Protopopov.

* src/capabilities.[ch]: pass the domain type to
  virCapabilitiesDefaultGuestArch() and use it to look up the default
  machine type from a specific guest domain if needed.

* src/conf/domain_conf.c, src/xen/xm_internal.c: update

* tests/qemuxml2argvdata/qemuxml2argv-machine-aliases2.xml: update
  the domain type to 'kvm' and remove the machine type to check
  that the default gets looked up correctly

Introduces several new public API options for migration

 - VIR_MIGRATE_PEER2PEER: With this flag the client only
   invokes the virDomainMigratePerform method, expecting
   the source host driver to do whatever is required to
   complete the entire migration process.
 - VIR_MIGRATE_TUNNELLED: With this flag the actual data
   for migration will be tunnelled over the libvirtd RPC
   channel. This requires that VIR_MIGRATE_PEER2PEER is
   also set.
 - virDomainMigrateToURI: This is variant of the existing
   virDomainMigrate method which does not require any
   virConnectPtr for the destination host. Given suitable
   driver support, this allows for all the same modes as
   virDomainMigrate()

The URI for VIR_MIGRATE_PEER2PEER must be a valid libvirt
URI. For non-p2p migration a hypervisor specific migration
URI is used.

virDomainMigrateToURI without a PEER2PEER flag is only
support for Xen currently, and it involves XenD talking
directly to XenD, no libvirtd involved at all.

* include/libvirt/libvirt.h.in: Add VIR_MIGRATE_PEER2PEER
  flag for migration
* src/libvirt_internal.h: Add feature flags for peer to
  peer migration (VIR_FEATURE_MIGRATE_P2P) and direct
  migration (VIR_MIGRATE_PEER2PEER mode)
* src/libvirt.c: Implement support for VIR_MIGRATE_PEER2PEER
  and virDomainMigrateToURI APIs.
* src/xen/xen_driver.c: Advertise support for DIRECT migration
* src/xen/xend_internal.c: Add TODO item for p2p migration
* src/libvirt_public.syms: Export virDomainMigrateToURI
  method
* src/qemu/qemu_driver.c: Add support for PEER2PEER and
  migration, and adapt TUNNELLED migration.
* tools/virsh.c: Add --p2p and --direct args and use the
  new virDomainMigrateToURI method where possible.

Re-arrange the doTunnelMigrate method putting all non-QEMU local
state setup steps first. This maximises chances of success before
then starting destination QEMU for receiving incoming migration.
Altogether this can reduce the number of goto cleanup labels to
something more managable.

* qemu/qemu_driver.c: Re-order steps in doTunnelMigrate

Simplify the doTunnelMigrate code by pulling out the code for
sending all tunnelled data into separate helper

* qemu/qemu_driver.c: introduce doTunnelSendAll() method

Simplify the doTunnelMigrate() method by pulling out the code
which opens/closes the virConnectPtr  object into a parent
method

* qemu/qemu_driver.c: Add doPeer2PeerMigrate which then calls
  doTunnelMigrate with dconn & dom_xml

virStreamAbort is needed when the caller wishes to terminate
the stream early, not when virStreamSend fails.

* qemu/qemu_driver.c: Fix calling of virStreamAbort during
  tunnelled migration

The code for tunnelled migration was added in a dedicated method,
but the native migration code is still inline in the top level
qemudDomainMigratePerform() API. Move the native code out into
a dedicated method too to make things more maintainable.

* src/qemu/qemu_driver.c: Pull code for performing a native
  QEMU migration out into separate method

The code for tunnelled migration wierdly required the app to pass
a  NULL 'dconn' parameter, only to have to use virConnectOpen
itself shortly thereafter to get a 'dconn' object. Remove this
bogus check & require the app to always pas 'dconn' as before

* src/libvirt.c: Require 'dconn' for virDomainMigrate calls again
  and remove call to virConnectOpen

Since virMigratePrepareTunnel() is used for migration over the
native libvirt connection, there is never any need to pass the
target URI to this method.

* daemon/remote.c, src/driver.h, src/libvirt.c, src/libvirt_internal.h,
  src/qemu/qemu_driver.c, src/remote/remote_driver.c,
  src/remote/remote_protocol.c, src/remote/remote_protocol.h,
  src/remote/remote_protocol.x: Remove 'uri_in' parameter from
  virMigratePrepareTunnel() method

Move the VIR_DRV_FEATURE* constants into libvirt_internal.h
since these flags are indicating whether  APIs in the
libvirt_internal.h file are supported by a driver

* src/driver.h: Remove VIR_DRV_FEATURE* constants
* src/libvirt_internal.h: Add VIR_DRV_FEATURE* constants, using
  an enum instead of #define
* src/internal.h: pull in libvirt_internal.h