Michal Privoznik [Wed, 23 Mar 2016 08:57:06 +0000 (09:57 +0100)]
qemuxml2argv: Mock virSCSIDeviceGetSgName
When constructing SCSI hostdev command line for qemu, the
/sys/bus/scsi/devices/... dir is scanned. Unfortunately, even in
the tests. This is needed to determine the name of SCSI device to
passthrough to qemu, because in the domain XML we were given its
address instead. Anyway, we should not be touching live system
data in our test suite as it produced unpredictable results. The
test is regressing from 1e9a083742efe on.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Vasiliy Tolstov [Wed, 23 Mar 2016 11:37:59 +0000 (11:37 +0000)]
autocreate tap device for ethernet network type
If a user specify network type ethernet, then create it via libvirt and run
script if it provided. After this commit user does not need to
run external script to create tap device or add root permissions to qemu
process.
Andrea Bolognani [Fri, 18 Mar 2016 17:03:53 +0000 (18:03 +0100)]
tests: hostdev: Add more tests
Ensure the code behaves properly even for situations that were not
being considered before, such as simply detaching devices from the
host without attaching them to a guest and attaching devices as
managed even though they had already been manually detached from
the host.
Andrea Bolognani [Fri, 18 Mar 2016 17:03:52 +0000 (18:03 +0100)]
hostdev: Use actual device when reattaching
Instead of forcing the values for the unbind_from_stub, remove_slot
and reprobe properties, look up the actual device and use that when
calling virPCIDeviceReattach().
This ensures the device is restored to its original state after
reattach: for example, if it was not bound to any driver before
detach, it will not be bound forcefully during reattach.
Andrea Bolognani [Fri, 18 Mar 2016 17:03:51 +0000 (18:03 +0100)]
hostdev: Save netdev configuration of actual device
We would be just fine looking up the information in pcidevs most
of the time; however, some corner cases would not be handled
properly, so look up the actual device instead.
Andrea Bolognani [Fri, 18 Mar 2016 17:03:50 +0000 (18:03 +0100)]
hostdev: Streamline device ownership tracking
After this patch, ownership of virPCIDevice instances is very easy
to keep track of: for each host PCI device, the only instance that
actually matters is the one inside one of the bookkeeping list.
Whenever some operation needs to be performed on a PCI device, the
actual device is looked up first; when this is not the case, a
comment explains the reason.
hostdev: Stop early if unmanaged devices have not been detached
Unmanaged devices, as the name suggests, are not detached
automatically from the host by libvirt before being attached to a
guest: it's the user's responsability to detach them manually
beforehand. If that preliminary step has not been performed, the
attach operation can't complete successfully.
Instead of relying on the lower layers to error out with cryptic
messages such as
error: Failed to attach device from /tmp/hostdev.xml
error: Path '/dev/vfio/12' is not accessible: No such file or directory
prevent the situation altogether and provide the user with a more
useful error message.
Andrea Bolognani [Thu, 17 Mar 2016 16:41:31 +0000 (17:41 +0100)]
hostdev: Detect untracked inactive devices
Unmanaged devices are attached to guests in two steps: first,
the device is detached from the host and marked as inactive;
subsequently, it is marked as active and attached to the guest.
If the daemon is restarted between these two operations, we lose
track of the inactive device.
Steps 5 and 6 of virHostdevPreparePCIDevices() already subtly
take care of this situation, but some planned changes will make
it so that's no longer the case. Plus, explicit is always better
than implicit.
Pavel Hrdina [Tue, 15 Mar 2016 12:54:24 +0000 (13:54 +0100)]
qemuxml2argvtest: use driver.config and priv for qemuDomainSetPrivatePaths
Update testutilsqemu to overwrite libDir and channelTargetDir and set
private paths using domain's privateData. This changes is required for
following patch.
This will skip few steps from qemuProcessStart in order to create only
qemu CMD. Use a VIR_QEMU_PROCESS_START_PRETEND for all the qemuProcess*
functions called by this one to not modify or check host.
This new function will be used later on for XMLToNative API and also for
qemuxml2argvtest to make sure that both API and test uses the same code
as qemuProcessStart.
We need also update qemuProcessInit to wrap few lines of code with check
that VIR_QEMU_PROCESS_START_PRETEND that makes sense only for
qemuProcessStart.
Pavel Hrdina [Tue, 15 Mar 2016 12:00:59 +0000 (13:00 +0100)]
qemu_process: introduce qemuProcessPrepareDomain
Move all code that modifies only live XML to this function. The new
VIR_QEMU_PROCESS_START_PRETEND flag will be used by qemuXMLToNative and
qemuxml2argvtest later in order to reuse the same code as
qemuProcessStart uses.
Every aligning requires at least one cast and it's hard to read. Let's
make a function that makes sure the pointer is moved according to the
alignment and use that to move throughout the data buffer.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Michal Privoznik [Mon, 21 Mar 2016 13:52:49 +0000 (14:52 +0100)]
tests: Produce predictable results in nsstest
Problem is that in the test any status file matching
tests/nssdata/*.status is loaded as it contains IP addresses that
are parsed. However, there's no order specified in which the
files are loaded. Therefore on different systems the order may be
different. This is then producing an unexpected results.
Instead of defining an order in which the files are loaded, make
the code that checks for missing IP addresses (or redundant ones)
cope with unordered list of addresses. The reasoning behind is
that the code doing the parsing is used in real NSS module where
we don't care for ordering.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Jim Fehlig [Tue, 23 Feb 2016 03:06:57 +0000 (20:06 -0700)]
libxl: support enabling and disabling <hap> feature
Until now, the libxl driver ignored any <hap> setting in domain XML
and deferred to libxl, which enables hap if not specified. While
this is a good default, it prevents disabling hap if desired.
This change allows disabling hap with <hap state='off'/>. hap is
explicitly enabled with <hap/> or <hap state='on/>. Absense of <hap>
retains current behavior of deferring default state to libxl.
Jim Fehlig [Tue, 23 Feb 2016 01:50:19 +0000 (18:50 -0700)]
xenconfig: change 'hap' setting to align with Xen behavior
hap is enabled by default in xm and xl config and usually only
specified when it is desirable to disable hap (hap = 0). Change
the xm,xl <-> xml converter to behave similarly. I.e. only
produce 'hap = 0' when <hap state='off'/> and vice versa.
Jim Fehlig [Tue, 23 Feb 2016 00:03:21 +0000 (17:03 -0700)]
conf: add 'state' attribute to <hap> feature
Most hypervisors use Hardware Assisted Paging by default and don't
require specifying the feature in domain conf. But some hypervisors
support disabling HAP on a per-domain basis. To enable HAP by default
yet provide a knob to disable it, extend the <hap> feature with a
'state=on|off' attribute, similar to <pvspinlock> and <vmport> features.
In the absence of <hap>, the hypervisor default (on) is used. <hap>
without the state attribute would be the same as <hap state='on'/> for
backwards compatibility. And of course <hap state='off'/> disables hap.
Jiri Denemark [Thu, 17 Dec 2015 12:46:18 +0000 (13:46 +0100)]
qemu: Add flags to qemuMigrationWaitForCompletion
The function already takes two bool arguments, switching to flags makes
it a lot easier to read. Especially in case we need to add another
boolean in the future.
Jiri Denemark [Thu, 14 Jan 2016 23:10:37 +0000 (00:10 +0100)]
qemu: Refuse to abort migration in post-copy mode
In post-copy mode none of the hosts has a complete guest state and
rolling back migration is impossible. Thus aborting it would be
equivalent to destroying the domain.
Jiri Denemark [Wed, 13 Jan 2016 15:29:58 +0000 (16:29 +0100)]
qemu: Handle post-copy migration failures
When migration fails in the post-copy mode, it's impossible to just kill
the destination domain and resume the source since the source no longer
contains current guest state. Let's mark domains on both sides as
VIR_DOMAIN_PAUSED_POSTCOPY_FAILED to let the upper layer decide what to
do with them.
Jiri Denemark [Thu, 26 Nov 2015 14:37:23 +0000 (15:37 +0100)]
qemu: Handle postcopy-active migration state
Migration enters "postcopy-active" state after QEMU switches to
post-copy and pauses guest CPUs. From libvirt's point of view this state
is similar to "completed" because we need to transfer guest execution to
the destination host.
Cristian Klein [Mon, 1 Dec 2014 15:59:53 +0000 (16:59 +0100)]
Add public APIs for post-copy migration
To use post-copy one has to start the migration with
VIR_MIGRATE_POSTCOPY flag and, while migration is in progress, call
virDomainMigrateStartPostCopy() to switch from pre-copy to post-copy.
Signed-off-by: Cristian Klein <cristiklein@gmail.com> Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Jiri Denemark [Tue, 5 Jan 2016 11:08:37 +0000 (12:08 +0100)]
Add event and state details for post-copy
VIR_DOMAIN_EVENT_SUSPENDED_POSTCOPY and VIR_DOMAIN_PAUSED_POSTCOPY are
used on the source host once migration enters post-copy mode (which
means the domain gets paused on the source. After the destination host
takes over the execution of the domain, its virtual CPUs are resumed and
the domain enters VIR_DOMAIN_RUNNING_POSTCOPY state and
VIR_DOMAIN_EVENT_RESUMED_POSTCOPY event is emitted.
In case migration fails during post-copy mode and none of the hosts have
complete state of the domain, both domains will remain paused with
VIR_DOMAIN_PAUSED_POSTCOPY_FAILED reason and an upper layer may decide
what to do.
util: Add virSocketAddrSetIPv[46]AddrNetOrder and use it
This allows setting the address in host and/or network order and makes
the naming consistent. Now you don't need to call [hn]to[nh]l()
functions as that is taken care of by these functions. Also, now
the *NetOrder take the address in network order, the other functions in
host order so the naming and usage is consistent. Some places were
having the address in network order and calling ntohl() just so the
original function can call htonl() again. This makes it nicer to read.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
qemu: Don't add -spice port=0 when no port is specified
If a <graphics type='spice'> has no port nor tlsPort set, the generated
QEMU command line will contain -spice port=0.
This is later going to be ignored by spice-server, but it's better not
to add it at all in this situation.
As an empty -spice is not allowed, we still need to append port=0 if we
did not add any other argument.
The end goal is to avoid adding -spice port=0,addr=127.0.0.1 to QEMU command
line when no SPICE port is specified in libvirt XML.
Currently, the code relies on port=xx to always be present, so subsequent
args can be unconditionally appended with a leading ','. Since port=0
will no longer be added in a subsequent commit, we append a ',' to every
arg instead of prepending, and remove the last one before adding it to
the arg list.
Cole Robinson [Fri, 8 Jan 2016 03:49:58 +0000 (22:49 -0500)]
domain: Add virDomainDefAddImplicitDevices
It's just a combination of AddImplicitControllers, and AddConsoleCompat.
Every caller that wants ImplicitControllers also wants the ConsoleCompat
AFAICT, so lump them together. We also need it for future patches.
Even if nss is disabled, the build system tries to build some
targets like libnss_libvirt_impl.la and nsstest. Hide those
under the "if WITH_NSS" block like the rest of NSS plugin bits.
If we expose this information, which is one byte in every PCI config
file, we let all mgmt apps know whether the device itself is an endpoint
or not so it's easier for them to decide whether such device can be
passed through into a VM (endpoint) or not (*-bridge).
The only purpose of this test is to catch possible linking
problems with libnss_libvirt.so.2.
One of the problems I faced was that the NSS plugin was unloaded
immediately after it got loaded and the name resolution process
continued with next configured option. Without any error. It was
very hard to debug why until I created this simple test and found
out immediately that there were some symbols missing. The reason
why problem was not caught in nsstest is that in the test we want
to use all the fancy stuff and therefore link it with libvirt.la.
So even if there's a symbol missing in the NSS plugin it will be
found in the libvirt.la.
But even after I resolved the issue we still need this test
because files the NSS plugin is built from are still live (mostly
those under utils/ dir). So as they change new symbol might be
required which would render the NSS plugin unusable.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Michal Privoznik [Sun, 14 Feb 2016 10:38:37 +0000 (11:38 +0100)]
nss: Implement _nss_libvirt_gethostbyname3_r
The implementation is pretty straightforward. Moreover, because
of the nature of things, gethostbyname_r and gethostbyname2_r can
be implemented at the same time too.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Michal Privoznik [Sat, 13 Feb 2016 08:27:50 +0000 (09:27 +0100)]
Initial support for NSS plugin skeleton
Name Service Switch is a glibc feature responsible for many
things. Translating domain names into IP addresses and vice versa
is just one of them. However, currently it's the only
functionality that this commit is tickling. Well, in this commit
the plugin skeleton is introduced. Implementation to come in next
patches.
Because of the future testing, where the implementation is to be
linked with a test, this needs to go into static library. Linking
a program with an .so statically is not portable. Therefore a
dummy libnss_libvirt_impl library is being introduced too.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Michal Privoznik [Tue, 15 Mar 2016 15:49:37 +0000 (16:49 +0100)]
virLeaseReadCustomLeaseFile: Allow server_duid to be NULL
This function is going to be used later in such context where the
argument makes no sense. Teach this function to cope with that
instead of the caller having to deal with passing some dummy
argument.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Jim Fehlig [Mon, 14 Mar 2016 22:35:02 +0000 (16:35 -0600)]
tests: add schema test for default cache mode
None of the existing domXML configs under tests/* specify a
default cache mode since default generally means "use the
hypervisor default" and is left unset by the various hypervisor
drivers. Add a config to tests/domainschemadata that specifies
cache='default'.
John Ferlan [Thu, 17 Mar 2016 11:29:23 +0000 (07:29 -0400)]
conf: Format disk pool part_separator attribute for running pool
Commit id '4f846170' added printing of a new field 'part_separator';
however, neglected to do so when there was an "freeExtent" defined
for the device (as there would be when the disk pool was started).
This patch adjusts the logic to appropriately format the device path and
if there the part_separator attribute.
John Ferlan [Tue, 15 Mar 2016 11:15:30 +0000 (07:15 -0400)]
qemu: Move last error save/restore to qemuBuildNetCommandLine
Commit 'ef2ab8fd' moved just the virDomainConfNWFilterTeardown and left
the logic to save/restore the current error essentially doing nothing
in the error path for qemuBuildCommandLine. So move it to where it
was meant to be.
Although the original code would reset the filter on command creation
errors after building the network command portion and commit 'ef2ab8fd'
altered that logic, the teardown is called during qemuProcessStop from
virDomainConfVMNWFilterTeardown and that code has the save/restore
last error logic, so just allow that code to handle the teardown rather
than running it twice. The qemuProcessStop would be called in the failure
path of qemuBuildCommandLine.
Michal Privoznik [Thu, 17 Mar 2016 12:30:42 +0000 (13:30 +0100)]
tests: Set PATH in each test
Currently we spawn couple of binaries in our test suite.
Moreover, we provide some spoofed versions of system binaries
hoping that those will be executed instead of the system ones.
For instance, for testing SSH socket we have written our own ssh
binary for producing predictable results. We certainly don't want
to execute the system ssh binary.
However, in order to prefer our binaries over system ones, we
need to set PATH environment variable. But this is done only at
the Makefile level. So if anybody runs a test by hand that
expects our spoofed binary, the test ends up executing real
system binaries. This is not good. In fact, it's terribly wrong.
The fix lies in a small trick - putting our build directory at
the beginning of the PATH environment variable in each test.
Hopefully, since every test has this VIRT_TEST_MAIN* wrapper, we
can fix this at a single place.
Moreover, while this removes setting PATH for our tests written
in bash, it's safe as we are not calling anything ours that would
require PATH change there.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Michal Privoznik [Thu, 17 Mar 2016 10:44:01 +0000 (11:44 +0100)]
Drop paths.h include
We include the file in plenty of places. This is mostly due to
historical reasons. The only place that needs something from the
header file is storage_backend_fs which opens _PATH_MOUNTED. But
it gets the file included indirectly via mntent.h. At no other
place in our code we need _PATH_.*. Drop the include and
configure check then.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Missing modules is a common expected scenario for most libvirt usage on
RPM distributions like Fedora, so it doesn't really warrant logging at
WARN level. Use INFO instead
It does not have a suffix ByName because there are no other means of
looking up the server and since the name is known, this should be the
preferred one.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
qemu: implement setting target disks migration port
Mostly it is just passing new parameter here and there. In case
of zero value we fallback to auto selecting port and thus keep
backward compatibility.
Also we need to fix places of auto selected port managment.
We should bother only when auto selected was done that is
when externally specified port is not 0.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
Some hypervisors (namely qemu) can have a separate connecton for
non-shared disks migration of active domains. Currently we have
no means to control the port of such a connection. At the same
time we have options to control port of memory migration traffic
(thru migration uri) as well as interfaces that target server
is bound to for incoming migration (thru VIR_MIGRATE_PARAM_LISTEN_ADDRESS).
Let's add the option for setting disks port too.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
We only wait 0.5 seconds for the session daemon to start up and present
its socket, which isn't sufficient for many users. Bump up the sleep
interval and retry amount so we wait for a total of 5.0 seconds.
Erik Skultety [Wed, 16 Mar 2016 15:22:30 +0000 (16:22 +0100)]
virlog: Fix build breaker with "comparison between signed and unsigned"
Refactor series 0b231195 worked with virLogDestination type which, depending
on the compiler, might be (and probably will be) an unsigned data type.
However, virEnumFromString may return -1 in case of an error. So, when enum
happens to be unsigned, some compilers will naturally complain about foo:
'if (foo < 0)'
Each version of virtuozzo supports only one type of SCSI controller
So if we add disk on SCSI bus, we should set SCSI controller model.
We can take it from vzCapabilities structure.
Because Vz6 supports SCSI(BUSLOGIC), IDE and SATA controllers only and
Vz7 supports SCSI(VIRTIO_SCSI) and IDE only we add list of supported
controllers and scsi models to vzCapabilities structure.
When a new connection opens, we select proper capabilities values according
to Virtuozzo version and check them in XMLPostParse.
vz: report correct disk format in domainGetXMLDesc
We should report correct disk format depending on vz version and domain type.
Since we support only one disk format for each domain type, we can take it
from vzCapabilities structure.
Maxim Nestratov [Wed, 16 Mar 2016 11:55:27 +0000 (14:55 +0300)]
vz: move prlsdkCheckDiskUnsupportedParams to vz_utils.c
As long as we have another function checking disk parameters correctness,
let's have them in one place. Here we change prefix of the moved function and
start to call it from vzCheckUnsupportedDisks rather than add disk.
As far as Virtuozzo6 and Virtuozzo7 support different disk types for virtual
machines (ploop and qcow2 respectively) and different buses (vz6: IDE, SCSI,
SATA; vz7: IDE SCSI) we add vzCapabilities structure to help undestand which
disk formats and buses are supported in the context of a current connection.
When a new connection opens, we select proper capabilities in accordance to
current Virtuozzo version.
Erik Skultety [Tue, 15 Mar 2016 21:15:02 +0000 (22:15 +0100)]
virlog: Refactor virLogParseOutputs
The problem with the original virLogParseOutputs method was that the way it
parsed the input, walking the string char by char and using absolute jumps
depending on the virLogDestination type, was rather complicated to read.
This patch utilizes virStringSplit method twice, first time to filter out any
spaces and split the input to individual log outputs and then for each
individual output to tokenize it by to the parts according to our
PRIORITY:DESTINATION?(:DATA) format. Also, to STREQLEN for matching destination
was replaced with virDestinationTypeFromString call.
Erik Skultety [Tue, 15 Mar 2016 20:35:17 +0000 (21:35 +0100)]
virlog: Introduce Type{To,From}String for virLogDestination
In order to refactor the ugly virLogParseOutputs method, this is a neat way of
finding out whether the destination type (in the form of a string) user
provided is a valid one. As a bonus, if it turns out it is valid, we get the
actual enum which will later be passed to any of virLogAddOutput methods right
away.
Erik Skultety [Wed, 16 Mar 2016 09:55:38 +0000 (10:55 +0100)]
tests: Slightly tweak virlogtest
Patch adds a generic DO_TEST_FULL macro, some PASS/FAIL macros to better
visually distinguish tests that should fail and tests that should pass. Also,
some cosmetic changes like renames and direct call to fprintf is replaced with
our VIR_TEST_DEBUG macro, as using testutils should be our preferred way of
reporting errors in tests.
In qemuConnectDomainXMLToNative() we set up the monitor, but we never
memset() it to zeros. Thanks to the introduction of the logfile
parameter of chardevs (and the logfile member of the struct), we started
checking whether that's non-NULL and that exposed this old error.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
projects / libvirt.git / log