When QEMU guest finishes its shutdown sequence, qemu stops virtual CPUs
and when started with -no-shutdown waits for us to kill it using
SGITERM. Since QEMU is flushing its internal buffers, some time may pass
before QEMU actually dies. We mistakenly used "paused" state (and
events) for this which is quite confusing since users may see a domain
going to pause while they expect it to shutdown. Since we already have
"shutdown" state with "the domain is being shut down" semantics, we
should use it for this state.

However, the state didn't have a corresponding event so I created one
and called its detail as VIR_DOMAIN_EVENT_SHUTDOWN_FINISHED (guest OS
finished its shutdown sequence) with the intent to add
VIR_DOMAIN_EVENT_SHUTDOWN_STARTED in the future if we have a
sufficiently capable guest agent that can notify us when guest OS starts
to shutdown.

Otherwise connections to older libvirt abort with:

$ virsh -c qemu+ssh://host.example.com/system list
error: invalid connection pointer in virDrvSupportsFeature
error: failed to connect to the hypervisor

Tested against 0.8.3 and 0.9.8-rc2.

https://bugzilla.redhat.com/show_bug.cgi?id=648855 mentioned a
misuse of 'an' where 'a' is proper; that has since been fixed,
but a search found other problems (some were a spelling error for
'and', while most were fixed by 'a').

* daemon/stream.c: Fix grammar.
* src/conf/domain_conf.c: Likewise.
* src/conf/domain_event.c: Likewise.
* src/esx/esx_driver.c: Likewise.
* src/esx/esx_vi.c: Likewise.
* src/rpc/virnetclient.c: Likewise.
* src/rpc/virnetserverprogram.c: Likewise.
* src/storage/storage_backend_fs.c: Likewise.
* src/util/conf.c: Likewise.
* src/util/dnsmasq.c: Likewise.
* src/util/iptables.c: Likewise.
* src/xen/xen_hypervisor.c: Likewise.
* src/xen/xend_internal.c: Likewise.
* src/xen/xs_internal.c: Likewise.
* tools/virsh.c: Likewise.

virBufferContentAndReset (intentionally) returns NULL for a buffer
with no content, but it is feasible to invoke a command with an
explicit empty string.

* src/util/command.c (virCommandAddEnvBuffer): Reject empty string.
(virCommandAddArgBuffer): Allow explicit empty argument.
* tests/commandtest.c (test9): Test it.
* tests/commanddata/test9.log: Adjust.

The RPC fixups needed on Linux are also needed on cygwin, and
worked without further tweaking to the list of fixups.  Also,
unlike BSD, Cygwin exports 'struct ifreq', but unlike Linux,
Cygwin lacks the ioctls that we were using 'struct ifreq' to
access.  This patch allows compilation under cygwin.

* src/rpc/genprotocol.pl: Also perform fixups on cygwin.
* src/util/virnetdev.c (HAVE_STRUCT_IFREQ): Also require AF_PACKET
definition.
* src/util/virnetdevbridge.c (virNetDevSetupControlFull): Only
compile if SIOCBRADDBR works.

The pathname for the pipe for tunnelled migration is unresolvable.  The
libvirt apparmor driver therefore refuses access, causing migration to
fail.  If we can't resolve the path, the worst that can happen is that
we should have given permission to the file but didn't.  Otherwise
(especially since this is a /proc/$$/fd/N file) the file is already open
and libvirt won't be refused access by apparmor anyway.

Also adjust virt-aa-helper to allow access to the
*.tunnelmigrate.dest.name files.

For more information, see https://launchpad.net/bugs/869553.
Signed-off-by: NSerge Hallyn <serge.hallyn@canonical.com>

Originaly, the code checked if another client is the queue and infered
ownership of the buck from that. Commit fa959500
added a separate variable to track the buck. That caused, that a new
call might enter claiming it has the buck, while another thread was
signalled to take the buck. This ends in two threads claiming they hold
the buck and entering poll(). This happens due to a race on waking up
threads on the client lock mutex.

This caused multi-threaded clients to hang, most prominently visible and
reproducible on python based clients, like virt-manager.

This patch causes threads, that have been signalled to take the buck to
re-check if buck is held by another thread.

Many of these were mentioned by Yuri Chornoivan in:
https://bugzilla.redhat.com/show_bug.cgi?id=669506

* src/esx/esx_vi.c (esxVI_WaitForTaskCompletion): Fix spelling.
* src/conf/netdev_vport_profile_conf.c
(virNetDevVPortProfileParse): Likewise.
* src/xen/xend_internal.c (xenDaemonDomainSetVcpusFlags):
Likewise.
* src/xen/xm_internal.c (xenXMDomainSetVcpusFlags): Likewise.
* src/esx/esx_util.c (esxUtil_ResolveHostname): Likewise.
* src/storage/storage_backend_fs.c
(virStorageBackendFileSystemBuild): Likewise.
* daemon/libvirtd.conf: Likewise.
* src/util/logging.c (virLogMessage): Likewise.
* src/uml/uml_conf.c (umlBuildCommandLineNet): Likewise.
* src/vmx/vmx.c (virVMXFormatEthernet): Likewise.

This ought to fix the build if you have net/if.h but do
not have struct ifreq

* configure.ac: Check for struct ifreq in net/if.h
* src/util/virnetdev.c: Conditionalize to avoid use of
  struct ifreq if it does not exist

probes.h can only be generated on Linux, and then only with dtrace
installed.  If it is part of the tarball, then either 'make dist'
will fail if you don't have that setup, or we would have to start
keeping probes.h in libvirt.git.  Since we only need it to be
generated when dtrace is in use, it's better to avoid shipping
it in the first place, and avoid tracking it in git.

Meanwhile, there is a build dependency - since the RPC code is
generated, it can be built early; but when dtrace is enabled, we
must ensure probes.h is built even earlier.  Commit 1afcfbdd tried
to fix this, but did so in a way that added probes.h into the
tarball, and broke VPATH as well.  Commit ecbca767 fixed VPATH,
but didn't fix the more fundamental problem.  This patch solves
the issue by adding a dependency instead.

Tested with 'make dist' in a clean VPATH builds, for both
'./configure --without-dtrace' and './configure --with-dtrace';
all configurations were able to correctly build a tarball, and
the dtrace configuration no longer sticks probes.h in the tarball.

* src/Makefile.am (REMOTE_DRIVER_GENERATED): Don't ship probes.h;
rather, make it a dependency.

Fix a logic error, the initial value of ret = -1, if just set --config,
it will goto endjob directly without doing its really job here.
Signed-off-by: NLei Li <lilei@linux.vnet.ibm.com>

The glibc time.h header has an undocumented __isleap macro
that we are using. Since it is undocumented & does not appear
on any other OS, stop using it and just define the macro in
libvirt code instead.

* src/util/virtime.c: Remove __isleap usage

Only one IPv4 DHCP definition is supported. Originally the code checked
for a multiple definition and returned an error, but the new domain
definition was already added to networks. This patch moves the check
before the newly defined network is added to active networks.

 *src/network/bridge_driver.c: networkDefine(): - move multiple IPv4
                                                  addresses check before
                                                  definition is used.

Detected by Coverity. Leak introduced in commit 90074ecf.
Signed-off-by: NAlex Jia <ajia@redhat.com>
Signed-off-by: NEric Blake <eblake@redhat.com>

Detected by Coverity. Leak introduced in commit c1df2c14.

Two bugs here:
1. memory leak on successful parse
2. failure to parse still returned success
Signed-off-by: NAlex Jia <ajia@redhat.com>
Signed-off-by: NEric Blake <eblake@redhat.com>

Detected by Coverity.  Leak introduced in commit 8866eed0.

Two bugs here:
1. logfd wasn't closed on all return paths
2. if we failed to mark a domain autodestroy, then the domain
was not made transient but we still returned success
Signed-off-by: NAlex Jia <ajia@redhat.com>
Signed-off-by: NEric Blake <eblake@redhat.com>

Detected by Coverity. Leak introduced in commit 673adba5.

Two separate bugs here:
1. call was not freed on all error paths
2. virCondDestroy was called even if virCondInit failed
Signed-off-by: NAlex Jia <ajia@redhat.com>
Signed-off-by: NEric Blake <eblake@redhat.com>

Detected by Coverity. Leak introduced in commit 0873b688.
Signed-off-by: NAlex Jia <ajia@redhat.com>

Detected by Coverity. Leak introduced in commit 109efd79.
Signed-off-by: NAlex Jia <ajia@redhat.com>

To add support for running libvirt on PowerPC, a CPU driver for the
PowerPC platform must be added.
Most generic cpu driver routines such as CPU compare, decode, etc
are based on CPUID comparison and are not relevant for non-x86
platforms.
Here, we introduce stubs for relevant PowerPC routines invoked by libvirt.
Signed-off-by: NPrerna Saxena <prerna@linux.vnet.ibm.com>
Signed-off-by: NAnton Blanchard <anton@au.ibm.com>

filter 0-device-weight when:

  - getting blkio parameters with --config
  - starting up a domain

When testing with blkio, I found these issues:

  (dom is down)
  virsh blkiotune dom --device-weights /dev/sda,300,/dev/sdb,500
  virsh blkiotune dom --device-weights /dev/sda,300,/dev/sdb,0
  virsh blkiotune dom
  weight         : 800
  device_weight  : /dev/sda,200,/dev/sdb,0

  # issue 1: shows 0 device weight of /dev/sdb that may confuse user

  (continued)
  virsh start dom

  # issue 2: If /dev/sdb doesn't exist, libvirt refuses to bring the
  # dom up because it wants to set the device weight to 0 of a
  # non-existing device. Since 0 means no weight-limit, we really don't
  # have to set it.

Prior to this patch, for a running dom, the commands:

$ virsh blkiotune dom --device-weights /dev/sda,502,/dev/sdb,498
$ virsh blkiotune dom --device-weights /dev/sda,503
$ virsh blkiotune dom
weight         : 500
device_weight  : /dev/sda,503

claim that /dev/sdb no longer has a non-default weight, but
directly querying cgroups says otherwise:

$ cat /cgroup/blkio/libvirt/qemu/dom/blkio.weight_device
8:0     503
8:16    498

After this patch, an explicit 0 is required to remove a device path
from the XML, and omitting a device path that was previously
specified leaves that device path untouched in the XML, to match
cgroups behavior.

* src/qemu/qemu_driver.c (parseBlkioWeightDeviceStr): Rename...
(qemuDomainParseDeviceWeightStr): ...and use correct type.
(qemuDomainSetBlkioParameters): After parsing string, modify
rather than replacing existing table.
* tools/virsh.pod (blkiotune): Tweak wording.

The next patch will make it possible to have virDomainSetBlkioParameters
leave device weights unchanged if they are not mentioned in the incoming
string, but this only works if the list of block weights does not allow
duplicate paths.  Technically, a user can still confuse libvirt by
passing alternate spellings that resolve to the same device, but it
is not worth worrying about working around that kind of abuse.

* src/conf/domain_conf.c (virDomainDefParseXML): Require unique
paths.

Implement the block I/O throttle setting and getting support to qemu
driver.
Signed-off-by: NLei Li <lilei@linux.vnet.ibm.com>
Signed-off-by: NZhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Signed-off-by: NEric Blake <eblake@redhat.com>

Enable block I/O throttle for per-disk in XML, as the first
per-disk IO tuning parameter.
Signed-off-by: NLei Li <lilei@linux.vnet.ibm.com>
Signed-off-by: NZhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Signed-off-by: NEric Blake <eblake@redhat.com>

Support Block I/O Throttle setting and query to remote driver.
Signed-off-by: NLei Li <lilei@linux.vnet.ibm.com>
Signed-off-by: NZhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Signed-off-by: NEric Blake <eblake@redhat.com>

* src/libvirt.c: Free user directory path

The virTimestamp and virTimeMs functions in src/util/util.h
duplicate functionality from virtime.h, in a non-async signal
safe manner. Remove them, and convert all code over to the new
APIs.

* src/util/util.c, src/util/util.h: Delete virTimeMs and virTimestamp
* src/lxc/lxc_driver.c, src/qemu/qemu_domain.c,
  src/qemu/qemu_driver.c, src/qemu/qemu_migration.c,
  src/qemu/qemu_process.c, src/util/event_poll.c: Convert to use
  virtime APIs

Use the new virTimeStringNowRaw() API for generating log timestamps
in an async signal safe manner

* src/util/logging.c: Use virTimeStringNowRaw

The logging APIs need to be able to generate formatted timestamps
using only async signal safe functions. This rules out using
gmtime/localtime/malloc/gettimeday(!) and much more.

Introduce a new internal API which is async signal safe.

  virTimeMillisNowRaw replacement for gettimeofday. Uses clock_gettime
                      where available, otherwise falls back to the unsafe
                      gettimeofday

  virTimeFieldsNowRaw  replacements for gmtime(), convert a timestamp
  virTimeFieldsThenRaw into a broken out set of fields. No localtime()
                       replacement is provided, because converting to
                       local time is not practical with only async signal
                       safe APIs.

  virTimeStringNowRaw  replacements for strftime() which print a timestamp
  virTimeStringThenRaw into a string, using a pre-determined format, with
                       a fixed size buffer (VIR_TIME_STRING_BUFLEN)

For each of these there is also a version without the Raw postfix
which raises a full libvirt error. These versions are not async
signal safe

* src/Makefile.am, src/util/virtime.c, src/util/virtime.h: New files
* src/libvirt_private.syms: New APis
* configure.ac: Check for clock_gettime in -lrt
* tests/virtimetest.c, tests/Makefile.am: Test new APIs

Fix the build on Mingw32 by removing the now obsolete
virGetPMCapabilities symbol from the private exports file

* src/libvirt_private.syms: Remove virGetPMCapabilities

If suspend failed for some reason (e.g. too short duration) then
subsequent attempts to trigger suspend were rejected because we
had already marked a suspend as being in progress

* src/util/virnodesuspend.c: Don't mark suspend as active
  until we've successfully triggered it

* src/lxc/lxc_conf.c, src/uml/uml_conf.c,
  src/xen/xen_hypervisor.c: Initialize suspend capabilities
* tests/xencapsdata/*xml: Add empty powermgmt capabilities

The command name for the suspend action does not need to be
strdup'd. The constant string can be used directly. This
also means the code can be trivially rearranged to make the
switch clearer

* src/util/virnodesuspend.c: Remove strdup of cmdString

To avoid probing the host power management features on any
call to virInitialize, only initialize the mutex in
virNodeSuspendInit. Do lazy load of the supported PM target
mask when it is actually needed

* src/util/virnodesuspend.c: Lazy init of supported features

If we ensure that virNodeSuspendGetTargetMask always resets
*bitmask to zero upon failure, there is no need for the
powerMgmt_valid field.

* src/util/virnodesuspend.c: Ensure *bitmask is zero upon
  failure
* src/conf/capabilities.c, src/conf/capabilities.h: Remove
  powerMgmt_valid field
* src/qemu/qemu_capabilities.c: Remove powerMgmt_valid

* src/libvirt_private.syms: Export virNodeSuspendSupportsTarget
  and virNodeSuspendGetTargetMask

The node suspend capabilities APIs should not have been put into
util.[ch]. Instead move them into virnodesuspend.[ch]

* src/util/util.c, src/util/util.h: Remove suspend capabilities APIs
* src/util/virnodesuspend.c, src/util/virnodesuspend.h: Add
  suspend capabilities APIs
* src/qemu/qemu_capabilities.c: Include virnodesuspend.h

Rename virGetPMCapabilities to virNodeSuspendGetTargetMask and
virDiscoverHostPMFeature to virNodeSuspendSupportsTarget.

* src/util/util.c, src/util/util.h: Rename APIs
* src/qemu/qemu_capabilities.c, src/util/virnodesuspend.c: Adjust
  for new names