Adding a new model for virtio-scsi roughly follows the same scheme
as the previous patch.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

KVM will be able to use a PCI SCSI controller even on POWER.  Let
the user specify the vSCSI controller by other means than a default.

After this patch, the QEMU driver will actually look at the model
and reject anything but auto, lsilogic and ibmvscsi.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NOsier Yang <jyang@redhat.com>

In qemuDomainAttachNetDevice, the guest's tap interface has only been
attached to the bridge if iface_connected is true. It's possible for
an error to occur prior to that happening, and previously we would
attempt to remove the tap interface from the bridge even if it hadn't
been attached.

QMP commands don't need to be escaped since converting them to json
also escapes special characters. When a QMP command fails, however,
libvirt falls back to HMP commands. These fallback functions
(qemuMonitorText*) do their own escaping, and pass the result directly
to qemuMonitorHMPCommandWithFd. If the monitor is in json mode, these
pre-escaped commands will be escaped again when converted to json,
which can result in the wrong arguments being sent.

For example, a filename test\file would be sent in json as
test\\file.

This prevented attaching an image file with a " or \ in its name in
qemu 1.0.50, and also broke rbd attachment (which uses backslashes to
escape some internal arguments.)
Reported-by: NMasuko Tomoya <tomoya.masuko@gmail.com>
Signed-off-by: NJosh Durgin <josh.durgin@dreamhost.com>
Signed-off-by: NEric Blake <eblake@redhat.com>

This patch fixes console corruption, that happens if two concurrent
sessions are opened for a single console on a domain. Result of this
corruption was that each of the console streams recieved just a part
of the data written to the pipe so every console rendered unusable.

New helper function for safe console handling is used to establish the
console stream connection. This function ensures that no other libvirt
client is using the console (with the ability to disconnect consoles of
libvirt clients) and that no UUCP style lockfile is placed on the PTY
device.

* src/qemu/qemu_domain.h
        - add data structure to domain's private data dealing with
          console connections
* src/qemu/qemu_domain.c:
        - allocate/free domain's console data structure
* src/qemu/qemu_driver.c
        - use the new helper function for console handling

This patch adds a set of functions used in creating console streams for
domains using PTYs and ensures mutually exclusive access to the PTYs.

If mutually exclusive access is not used, two clients may open the same
console, which results in corruption on both clients as both of them
race to read data from the PTY.

Two approaches are used to ensure this:
1) Internal data structure holding open PTYs.
        This is used internally and enables the user to forcibly
        terminate another console connection eg. when somebody leaves
        the console open on another host.

2) UUCP style lock files:
        This uses UUCP lock files according to the  FHS
        ( http://www.pathname.com/fhs/pub/fhs-2.3.html#VARLOCKLOCKFILES )
        to check if other programs (like minicom) are not using the pty
        device of the console.

        This feature is disabled by default and may be enabled using
        configure parameter
        --with-console-lock-files=/path/to/lock/file/directory
        or --with-console-lock-files=auto (which tries to infer the
        location from OS used (currently only linux).

        On usual linux systems, normal users may not write to the
        /var/lock directory containing the locks. This poses problems
        while in session mode. If the current user has no access to the
        lockfile directory, check for presence of the file is still
        done, but no lock file is created. This does NOT result in an
        error.

This patch adds another callback to a FDstream object. The original
callback is used by the daemon stream driver to handle events.

This callback is called if and only if the stream is about to be closed.
This might be used to handle cleanup steps after a fdstream exits. This
will be used later on in ensuring mutually exclusive access to consoles.

* src/fdstream.c:
        - emit the callback, when stream is being closed
        - add data structures needed to handle the callback
        - add function to register callback
* src/fdstream.h:
        - define function prototypes for the callback

This patch causes the fdstream driver to call the stream event callback
if virStreamAbort() is called on a stream using this driver.

A remote handler for a stream can only detect changes via stream events,
so this event callback is necessary in order to enable a daemon to abort
a stream in such a way that the client will see the change.

* src/fdstream.c:
        - modify close function to call stream event callback

This patch adds a set of flags to be used with the virDomainOpenConsole
API call to specify if the user wishes to interrupt an existing console
session or just to try open a new one.

VIR_DOMAIN_CONSOLE_SAFE - specifies that the console connection should
                          be opened only if the hypervisor supports
                          mutually exclusive access to console devices

VIR_DOMAIN_CONSOLE_FORCE - specifies that the caller wishes to interrupt
                           existing session and force a creation of a
                           new one.

This patch changes behavior of virPidFileRead to enable passing NULL as
path to the binary the pid file should be checked against to skip this
check. This enables using this function for reading files that have same
semantics as pid files, but belong to unknown processes.

using 'system-wakeup' monitor command. It is supported only in JSON,
as we are enabling it if possible. Moreover, this command is available
in qemu-1.1+ which definitely has JSON.

This API allows a domain which previously called
virDomainPMSuspendForDuration() to be woken up.

Function xmlParseURI does not remove square brackets around IPv6
address when parsing. One of the solutions is making wrappers around
functions working with xmlURI*. This assures that uri->server will be
always properly assigned and it doesn't have to be changed when used
on some new place in the code.
For this purpose, functions virParseURI and virSaveURI were
added. These function are wrappers around xmlParseURI and xmlSaveUri
respectively.
Also there is one new syntax check function to prohibit these functions
anywhere else.

File changes:
 - src/util/viruri.h        -- declaration
 - src/util/viruri.c        -- definition
 - src/libvirt_private.syms -- symbol export
 - src/Makefile.am          -- added source and header files
 - cfg.mk                   -- added sc_prohibit_xmlURI
 - all others               -- ID name and include fixes

The /usr/include/python/pyconfig.h file pollutes the global
namespace with a huge number of HAVE_XXX and WITH_XXX
defines. These change what we detected in our own config.h
In particular if you try to build without DTrace, python's
headers turn it back on with predictable fail.

THe hack to workaround this is to rename WITH_DTRACE to
WITH_DTRACE_PROBES to avoid the namespace clash

It's possible to disable SPICE TLS in qemu.conf. When this happens,
libvirt ignores any SPICE TLS port or x509 directory that may have
been set when it builds the qemu command line to use. However, it's
not ignoring the secure channels that may have been set and adds
tls-channel arguments to qemu command line.
Current qemu versions don't report an error when this happens, and try to use
TLS for the specified channels.

Before this patch

<domain type='kvm'>
  <name>auto-tls-port</name>
  <memory>65536</memory>
  <os>
    <type arch='x86_64' machine='pc'>hvm</type>
  </os>
  <devices>
    <graphics type='spice' port='5900' tlsPort='-1' autoport='yes' listen='0' ke
      <listen type='address' address='0'/>
      <channel name='main' mode='secure'/>
      <channel name='inputs' mode='secure'/>
    </graphics>
  </devices>
</domain>

generates

-spice port=5900,addr=0,disable-ticketing,tls-channel=main,tls-channel=inputs

and starts QEMU.

After this patch, an error is reported if a TLS port is set in the XML
or if secure channels are specified but TLS is disabled in qemu.conf.
This is the behaviour the oVirt people (where I spotted this issue) said
they would expect.

This fixes bug #790436

This patch adds support for vmx files with empty networkName
values (which is the case for vmx generated by Workstation).
It also adds support for vmx containing NATed network interfaces.

Update test suite accordingly

[forwarding this here from RH bug #796732]

When creating a network (virsh net-create) with an erroneous XML
containing an empty <name> element, the error message is misleading:

error: Failed to create network from foo.xml
error: missing domain name information

It took me a bit of time to figure out that it was the *network* name
that was missing (I generate this xml and didn't look at it, first).

I realized that the same message is used for missing name when creating
a domain, network, or device node.

https://bugzilla.redhat.com/show_bug.cgi?id=795656 mentions
that a graceful destroy request can time out, meaning that the
error message is user-visible and should be more appropriate
than just internal error.

* src/qemu/qemu_driver.c (qemuDomainDestroyFlags): Swap error type.

Migrating domains with disks using cache != none is unsafe unless the
disk images are stored on coherent clustered filesystem. Thus we forbid
migrating such domains unless VIR_MIGRATE_UNSAFE flags is used.

This patch adds VIR_MIGRATE_UNSAFE flag for migration APIs and new
VIR_ERR_MIGRATION_UNSAFE error code.  The error code should be returned
whenever migrating a domain is considered unsafe (e.g., it's configured
in a way that does not ensure data integrity once it is migrated).
VIR_MIGRATE_UNSAFE flag may be used to force migration even though it
would normally be considered unsafe and forbidden.

AC_CHECK_PROG checks for program in given path. However, if it doesn't
exists, [variable] is set to [value-if-not-found]. We don't want this
to be the empty string in case of 'modprobe' and 'scrub' as we want to
fallback to runtime detection.

Despite documentation, if we do fork() parent always returns -1
even if file is accessible. Which is wrong obviously.

Clarify the documentation of virConnectListDomains.

Adding "Expect:" to the header list stops libcurl from sending a
Expect header at all.

Before, a dummy Expect header was added that might confuse HTTP
proxies and result in HTTP error code 417 being reported.

Previously we would have:

"os type 'hvm' & arch 'idontexist' combination is not supported"

Now we get

"No guest options available for arch 'idontexist'"

or if options available but guest OS type not applicable:

"No os type 'xen' available for arch 'x86_64'"

* src/util/virfile.h: the virFileWrapperFdFlags being defined as
  a globa variable instead of a type ended up generating a duplicate
  symbol error.
* AUTHORS: added Lincoln Myers

Improve the logging during driver initialization when testing the command
line tools.

* src/qemu/qemu_process.c (qemuFindAgentConfig): avoid crash libvirtd due to
deref a NULL pointer.

* How to reproduce?
1. virsh edit the following xml into guest configuration:
    <channel type='pty'>
      <target type='virtio'/>
    </channel>
2. virsh start <domain>

or
% virt-install -n foo -r 1024 --disk path=/var/lib/libvirt/images/foo.img,size=1 \
--channel pty,target_type=virtio -l <installation tree>
Signed-off-by: NAlex Jia <ajia@redhat.com>

When migrating a qemu domain, we enter the monitor, send some commands,
try to connect to destination qemu, send other commands, end exit the
monitor. However, if we couldn't connect to destination qemu we forgot
to exit the monitor.

Bug introduced by commit d9d518b1.

In case libvirtd cannot detect host CPU model (which may happen if it
runs inside a virtual machine), the daemon is likely to segfault when
starting a new qemu domain. It segfaults when domain XML asks for host
(either model or passthrough) CPU or does not ask for any specific CPU
model at all.

Currently, if scrub (used for wiping algorithms) is not present
at compile time, we don't support any other wiping algorithms than
zeroing, even if it was installed later. Switch to runtime detection
instead.

Bug introduced in commit 35abced2.  On an inactive domain,
$ virsh snapshot-create-as dom snap
$ virsh snapshot-create dom
$ virsh snapshot-create dom
$ virsh snapshot-delete --children dom snap
could crash libvirtd, due to a use-after-free that results
when the callback freed the current element in the iteration.

* src/conf/domain_conf.c (virDomainSnapshotForEachChild)
(virDomainSnapshotActOnDescendant): Allow iteration to delete
current child.

This patch allows libvirt to add interfaces to already
existing Open vSwitch bridges. The following syntax in
domain XML file can be used:

    <interface type='bridge'>
      <mac address='52:54:00:d0:3f:f2'/>
      <source bridge='ovsbr'/>
      <virtualport type='openvswitch'>
        <parameters interfaceid='921a80cd-e6de-5a2e-db9c-ab27f15a6e1d'/>
      </virtualport>
      <address type='pci' domain='0x0000' bus='0x00'
                          slot='0x03' function='0x0'/>
    </interface>

or if libvirt should auto-generate the interfaceid use
following syntax:

    <interface type='bridge'>
      <mac address='52:54:00:d0:3f:f2'/>
      <source bridge='ovsbr'/>
      <virtualport type='openvswitch'>
      </virtualport>
      <address type='pci' domain='0x0000' bus='0x00'
                          slot='0x03' function='0x0'/>
    </interface>

It is also possible to pass an optional profileid. To do that
use following syntax:

   <interface type='bridge'>
     <source bridge='ovsbr'/>
     <mac address='00:55:1a:65:a2:8d'/>
     <virtualport type='openvswitch'>
       <parameters interfaceid='921a80cd-e6de-5a2e-db9c-ab27f15a6e1d'
                   profileid='test-profile'/>
     </virtualport>
   </interface>

To create Open vSwitch bridge install Open vSwitch and
run the following command:

    ovs-vsctl add-br ovsbr

An upcoming patch will add a <virtualport> element to interfaces of
type='bridge', so it makes sense to give this function a more generic
name.

The current default method of terminating the qemu process is to send
a SIGTERM, wait for up to 1.6 seconds for it to cleanly shutdown, then
send a SIGKILL and wait for up to 1.4 seconds more for the process to
terminate. This is problematic because occasionally 1.6 seconds is not
long enough for the qemu process to flush its disk buffers, so the
guest's disk ends up in an inconsistent state.

Since this only occasionally happens when the timeout prior to SIGKILL
is 1.6 seconds, this patch increases that timeout to 10 seconds. At
the very least, this should reduce the occurrence from "occasionally"
to "extremely rarely". (Once SIGKILL is sent, it waits another 5
seconds for the process to die before returning).

Note that in the cases where it takes less than this for qemu to
shutdown cleanly, libvirt will *not* wait for any longer than it would
without this patch - qemuProcessKill polls the process and returns as
soon as it is gone.

This patch is based on an earlier patch by Eric Blake which was never
committed:

https://www.redhat.com/archives/libvir-list/2011-November/msg00243.html

Aside from rebasing, this patch only drops the driver lock once (prior
to the first time the function sleeps), then leaves it dropped until
it returns (Eric's patch would drop and re-acquire the lock around
each call to sleep).

At the time Eric sent his patch, the response (from Dan Berrange) was
that, while it wasn't a good thing to be holding the driver lock while
sleeping, we really need to rethink locking wrt the driver object,
switching to a finer-grained approach that locks individual items
within the driver object separately to allow for greater concurrency.

This is a good plan, and at the time it made sense to not apply the
patch because there was no known bug related to the driver lock being
held in this function.

However, we now know that the length of the wait in qemuProcessKill is
sometimes too short to allow the qemu process to fully flush its disk
cache before SIGKILL is sent, so we need to lengthen the timeout (in
order to improve the situation with management applications until they
can be updated to use the new VIR_DOMAIN_DESTROY_GRACEFUL flag added
in commit 72f8a7f1). But, if we
lengthen the timeout, we also lengthen the amount of time that all
other threads in libvirtd are essentially blocked from doing anything
(since just about everything needs to acquire the driver lock, if only
for long enough to get a pointer to a domain).

The solution is to modify qemuProcessKill to drop the driver lock
while sleeping, as proposed in Eric's patch. Then we can increase the
timeout with a clear conscience, and thus at least lower the chances
that someone running with existing management software will suffer the
consequence's of qemu's disk cache not being flushed.

In the meantime, we still should work on Dan's proposal to make
locking within the driver object more fine grained.

(NB: although I couldn't find any instance where qemuProcessKill() was
called with no jobs active for the domain (or some other guarantee
that the current thread had at least one refcount on the domain
object), this patch still follows Eric's method of temporarily adding
a ref prior to unlocking the domain object, because I couldn't
convince myself 100% that this was the case.)

via user agent. Allow targets mem & hybrid iff system_wakeup
monitor command is available.

In the future (my next patch in fact) we may want to make
decisions depending on qemu having a monitor command or not.
Therefore, we want to set qemuCaps flag instead of querying
on the monitor each time we are about to make that decision.