qemu: Remove explicit values from virQEMUCapsFlags
This gets rid of the partially enforced alignment and makes it less
likely for a bogus value to be introduced in the enumeration.
Capabilities are divided in five-element groups for better readability.
Use #define for QEMU_CAPS_NET_NAME and QEMU_CAPS_HOST_NET_ADD, both
of which are aliases for QEMU_CAPS_0_10.
Peter Krempa [Tue, 6 Oct 2015 15:02:09 +0000 (17:02 +0200)]
qemu: migration: Skip few checks while doing offline migration
qemuMigrationIsAllowed would disallow offline migration if the VM
contained host devices or memory modules. Since during offline migration
we don't transfer any state we can safely migrate VMs with such
configuration.
Peter Krempa [Tue, 6 Oct 2015 14:24:48 +0000 (16:24 +0200)]
qemu: migration: Use migration flags in qemuMigrationIsAllowed
Use the migration @flags for checking various migration aspects rather
than picking them out as booleans. Document the new semantics in the
function header.
Peter Krempa [Tue, 6 Oct 2015 13:39:48 +0000 (15:39 +0200)]
qemu: migration: Split source and destination migration checks
Extract the hostdev check from qemuMigrationIsAllowed into a separate
function since that is the only part that needs to be done in the v2
migration protocol prepare phase on the destination. All other checks
were added when the v3 protocol existed so they don't need to be
extracted.
This change will allow to drop the @def argument for
qemuMigrationIsAllowed and further simplify the function.
We git-ify the libvirt directory as part of applying patches in the spec
file, but 'git clean' will ignore subfolders that appear to be standalone
git repos.
Let's just delete the .git directory after we're done with it.
In fact, it was never used as far as vz has no features supporting it.
That is why there will be no harm to anyone if we just remove this code to
prevent further misunderstanding and efforts to support dead code.
At the time this code was added we had intentions to support libvirt interface
to manage vz networks. In fact, it was never implemented completely to work
correctly that makes me think that there will be no harm to anyone if we just
rip it off. Moreover, in vz7 we started to use libvirt bridge network driver to
manage networks.
- qemuxml2argv-aarch64-mmio-default-pci: Verify that we still default
to virtio-mmio even if qemu is new enough to support PCI
- qemuxml2argv-aarch64-virtio-pci: Check generated arm virtio PCI args
Jiri Denemark [Fri, 2 Oct 2015 10:08:26 +0000 (12:08 +0200)]
qemu: Wait until destination QEMU consumes all migration data
Even though QEMU on the source host reports completed migration and thus
we move to the Finish phase, QEMU on the destination host may still be
processing migration data. Thus before we can start guest CPUs on the
destination, we have to wait for a completed migration event.
Jiri Denemark [Mon, 5 Oct 2015 12:32:58 +0000 (14:32 +0200)]
qemu: Make updating stats in qemuMigrationCheckJobStatus optional
With new QEMU which supports migration events,
qemuMigrationCheckJobStatus needs to explicitly query QEMU for migration
statistics once migration is completed to make sure the caller sees
up-to-date statistics with both old and new QEMU. However, some callers
are not interested in the statistics at all and once we start waiting
for a completed migration on the destination host too, checking the
statistics would even fail. Let's push the decision whether to update
the statistics or not to the caller.
Jiri Denemark [Thu, 1 Oct 2015 19:54:38 +0000 (21:54 +0200)]
qemu: Copy completed migration stats only on success
The destination host gets detailed statistics about the current
migration form the source host via migration cookie and copies them to
the domain object so that they can be queried using
virDomainGetJobStats. However, we should only copy statistics to the
domain object when migration finished successfully.
Jiri Denemark [Thu, 1 Oct 2015 19:39:35 +0000 (21:39 +0200)]
qemu: Always update migration times on destination
Even if we are migrating a domain with VIR_MIGRATE_PAUSED flag set, we
should still update the total time of the migration. Updating downtime
doesn't hurt either, even though we don't actually start guest CPUs.
We are distributing virkeymaps.h and all the tools needed to rebuild
that file. On top of that, we are generating that file into the
$(srcdir) and that sometimes fails when trying to do make dist in VPATH
on rawhide fedora. And we don't clean the file when maintainer-clean
make target is requested. So let's not distribute the file and rather
let everyone rebuild it when needed and clean it when appropriate.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
apparmor: differentiate between error and unconfined profiles
profile_status function was not making any difference between error
cases and unconfined profiles. The problem with this approach is that
dominfo was throwing an error on unconfined domains.
Peter Krempa [Mon, 5 Oct 2015 07:58:30 +0000 (09:58 +0200)]
conf: Remove <metadata> elements with no namespace
Our docs state that subelements of <metadata> shall have a namespace
and the medatata APIs expect that too. To avoid inaccessible
<metadata> sub-elements, just remove those that don't conform to the
documentation.
Apart from adding the new condition this patch renames the function and
refactors the code flow to allow the changes.
- It poorly duplicates virtTestRun pass/fail reporting logic
- It doesn't have virtTestRun's alloc testing support
- It only reports the test name _after_ the test has run.
- It doesn't follow the standard virtTestRun pattern that most other
tests use.
There's no users left, so drop it. If any other async tests like eventtest
spring up that don't cleanly fit the virtTestRun pattern, I suggest they
just open code the support for it around virtTestRun
These event tests aren't run synchronously, so there isn't an obvious
function to pass to virtTestRun. Instead, open code roughly what
virtTestResult did before: printing an error message if a test failed.
Calculation of the extended and logical partition values for the disk
pool is complex. As the bz points out an extended partition should have
it's allocation initialized to 0 (zero) and keep the capacity as the size
dictated by the extents read. Then for each logical partition found,
adjust the allocation of the extended partition.
Finally, previous logic tried to avoid recalculating things if a logical
partition was deleted; however, since we now have special logic to handle
the allocation of the extended partition, just make life easier by reading
the partition table again - rather than doing the reverse adjustment.
When 'starting' up a disk pool, we need to make sure the label on the
device is valid; otherwise, the followup refreshPool will assume the
disk has been properly formatted for use. If we don't find the valid
label, then refuse the start and give a proper reason.
John Ferlan [Thu, 1 Oct 2015 12:24:14 +0000 (08:24 -0400)]
storage: Add additional errors/checks for disk label
Let's check to ensure we can find the Partition Table in the label
and that libvirt actually recognizes that type; otherwise, when we
go to read the partitions during a refresh operation we may not be
reading what we expect.
This will expand upon the types of errors or reason that a build
would fail, so we can create more direct error messages.
John Ferlan [Thu, 1 Oct 2015 13:30:28 +0000 (09:30 -0400)]
storage: Add param to check whether we can write a disk label
Modify virStorageBackendDiskValidLabel to add a 'writelabel' parameter.
While initially for the purpose of determining whether the label should
be written during DiskBuild, a future use during DiskStart could determine
whether the pool should be started using the label found. Augment the
error messages also to give a hint as to what someone may need to do
or why the command failed.
John Ferlan [Thu, 1 Oct 2015 12:07:35 +0000 (08:07 -0400)]
storage: Refactor disk label checking
Create a new function virStorageBackendDiskValidLabel to handle checking
whether there is a label on the device and whether it's valid or not.
While initially for the purpose of determining whether the label can be
overwritten during DiskBuild, a future use during DiskStart could determine
whether the pool should be started using the label found.
Although perhaps bordering on a don't do that type scenario, if
someone creates a volume in a pool outside of libvirt, then uses that
same name to create a volume in the pool via libvirt, then the creation
will fail and in some cases cause the same name volume to be deleted.
This patch will refresh the pool just prior to checking whether the
named volume exists prior to creating the volume in the pool. While
it's still possible to have a timing window to create a file after the
check - at least we tried. At that point, someone is being malicious.
John Ferlan [Wed, 30 Sep 2015 21:37:27 +0000 (17:37 -0400)]
virfile: Fix error path for forked virFileRemove
As it turns out the caller in this case expects a return < 0 for failure
and to get/use "errno" rather than using the negative of returned status.
Again different than the create path.
If someone "deleted" a file from the pool without using virsh vol-delete,
then the unlink/rmdir would return an error (-1) and set errno to ENOENT.
The caller checks errno for ENOENT when determining whether to throw an
error message indicating the failure. Without the change, the error
message is:
John Ferlan [Wed, 30 Sep 2015 21:25:34 +0000 (17:25 -0400)]
virfile: Add extra check for direct delete in virFileRemove
Unlike create options, if the file to be removed is already in the
pool, then the uid/gid will come from the pool. If it's the same as the
currently running process, then just do the unlink/rmdir directly
rather than going through the fork processing unnecessarily
Andrea Bolognani [Wed, 30 Sep 2015 15:58:58 +0000 (11:58 -0400)]
qemu: Add conditions for qemu-kvm use on ppc64
qemu-kvm can be used to run ppc64 guests on ppc64le hosts and vice
versa, since the hardware is actually the same and the endianness
is chosen by the guest kernel.
Up until now, however, libvirt didn't allow the use of qemu-kvm
to run guests if their endianness didn't match the host's.
Peter Krempa [Fri, 2 Oct 2015 13:49:01 +0000 (15:49 +0200)]
rpc: libssh2: Fix regression in ssh host key verification
Commit 792f81a40e caused a regression in the libssh2 host key
verification code by changing the variable type of 'i' to unsigned.
Since one of the loops used -1 as a special value if the asking
callback was found the conversion made a subsequent test always fail.
The bug was stealth enough to pass review, compilers and coverity.
Peter Krempa [Tue, 29 Sep 2015 16:13:06 +0000 (18:13 +0200)]
qemu: Perform the disk WWN check only on fresh starts
Since we'd disallow migration of a guest that would have possibly
invalid config but still be able to work, relax the WWN check to be
performed only on new starts of the VM.
Laine Stump [Fri, 18 Sep 2015 17:51:33 +0000 (13:51 -0400)]
interface: let netcf pre-filter for active vs. inactive
If a system has a large number of active or active interfaces, it can
be a big waste of time to retrieve and qualify all interfaces if the
caller only wanted one subset. Since netcf has a simple flag for this,
translate the libvirt flag into a netcf flag and let netcf pre-filter.
Laine Stump [Fri, 18 Sep 2015 17:19:27 +0000 (13:19 -0400)]
interface: re-use name and mac address rather than re-retrieving
Getting the MAC address of an interface is actually fairly expensive,
and we've already gotten it and stored it into def, so just keep def
around a bit longer and retrieve it from there.
This reduces the time for "virsh iface-list --all" from 28 to 23
seconds when there are 400 interfaces.
Laine Stump [Fri, 18 Sep 2015 17:24:47 +0000 (13:24 -0400)]
interface: report correct interface count when not returning list
The spec for virConnectListAllInterfaces says that if the pointer that
is supposed to hold the list of interfaces is NULL, the function
should just return the count of interfaces that matched the filter,
but the code never increments the count if the list pointer is NULL.
We are using memory-backing-file even when it's not needed, for example
if user requests hugepages for memory backing, but does not specify any
pagesize or memory node pinning. This causes migrations to fail when
migrating from older libvirt that did not do this. So similarly to
commit 7832fac84741d65e851dbdbfaf474785cbfdcf3c which does it for
memory-backend-ram, this commit makes is more generic and
backend-agnostic, so the backend is not used if there is no specific
pagesize of hugepages requested, no nodeset the memory node should be
bound to, no memory access change required, and so on.
So since the introduction of the memory-backend-file object until now we
only added '-mem-path' for non-NUMA guests and we used the parameters of
the memory-backend-file object to specify the path to the hugetlbfs
mount. But hugepages can be also used without memory-backend-file
object, as it used to be before its introduction. Let's just get this
part of the code back and properly append the '-mem-path' for NUMA
guests as well, but only when the memory backend is not needed.
This parameter is already being applied when no numa is requested and
because we still use memory-object-file unconditionally for
hugepage-backed NUMA guests, this should not fire until later.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
qemu: Extract -mem-path building into its own function
That function is called qemuBuildMemPathStr() and will be used in
other places in the future. The change in the test suite is proper due
to the fact that -mem-prealloc makes only sense with -mem-path (from
qemu documentation -- html/qemu-doc.html).
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Pavel Fedin [Wed, 30 Sep 2015 11:04:10 +0000 (14:04 +0300)]
qemu: Add support for gic-version machine option
Support for GICv3 has been recently introduced in qemu using gic-version
option for the 'virt' machine. The option can actually take values of
'2', '3' and 'host', however, since in libvirt this is a numeric
parameter, we limit it only to 2 and 3. Value of 2 is not added to the
command line in order to keep backward compatibility with older qemu
versions.
Unfortunately qemu currently doesn't offer introspection for machine types,
so we have to rely on version number, similar to QEMU_CAPS_MACHINE_USB_OPT.
Even after my rework of startupPolicy handling, one command
slipped my attention. The change-media command has a very unique
approach to constructing disk XML. However, it will not preserve
startupPolicy attribute.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Ján Tomko [Thu, 24 Sep 2015 15:01:40 +0000 (17:01 +0200)]
Update pool allocation with new values on volume creation
Since commit e0139e3, we update the pool allocation with
the user-provided allocation values.
For qcow2, the allocation is ignored for volume building,
but we still subtracted it from pool's allocation.
This can result in interesting values if the user-provided
allocation is large enough:
John Ferlan [Fri, 25 Sep 2015 21:33:05 +0000 (17:33 -0400)]
conf: Fix virtType check
Commit id '7383b8cc' changed virDomainDef 'virtType' to an enum, that
caused a build failure on some archs due to comparing an unsigned value
to < 0. Adjust the fetch of 'type' to be into temporary 'int virtType'
and then assign that virtType to the def->virtType
Introduce VIR_DOMAIN_VIRT_NONE to give domaintype the default value of zero.
This is specially helpful in constructing better error messages
when we don't want to look up the default emulator by virtType.
The test data in vircapstest.c is also modified to reflect this change.
Michal Privoznik [Fri, 25 Sep 2015 11:20:38 +0000 (13:20 +0200)]
tools: Link libvirt.la and readline to libvirt_shell.a
So, our mingw build is broken. It's because while libvirt_shell
library is using some of our internal APIs, e.g. virStrndup, and
readline API but it's not being linked with nor libvirt.la nor
libreadline. Only subsequent users of the library, like virsh,
do link to the needed libraries. In fact, I'm surprised Linux
linker doesn't care, because how can it make a static library
with missing symbols is mystery to me.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
As of commit 6992994, we set graphics/@listen attribute according to the
first listen child element even if that element is of type='network'.
This was done for backward compatibility with applications which only
support the original listen attribute. However, by doing so we broke
migration to older libvirt which tried to check that the listen
attribute matches one of the listen child elements but which did not
take type='network' elements into account.
We are not concerned about compatibility with old applications when
formatting domain XML for migration for two reasons. The XML is consumed
only by libvirtd and the IP address associated with type='network'
listen address on the source host is just useless on the destination
host. Thus, we can safely avoid propagating the type='network' IP
address to graphics/@listen attribute when creating migratable XML.
John Ferlan [Tue, 8 Sep 2015 11:05:55 +0000 (07:05 -0400)]
qemu: Resolve Coverity RESOURCE_LEAK
This seemed to be more of a false positive as for some reason Coverity
was missing the "ret < 0" goto error condition and somehow believing that
event could be overwritten. At first I thought it was just the ret != 0
condition difference, but it wasn't.
In any case, make use of the recent change to qemuDomainEventQueue to
check event == NULL and just pass it as a parameter directly in the
error path. That avoids the error.
John Ferlan [Wed, 23 Sep 2015 22:13:57 +0000 (18:13 -0400)]
qemu: Resolve Coverity CHECKED_RETURN
Coverity complains that return from virHookCall is not checked in
one place in qemuProcessStop. Since the comment notes that we cannot
stop the operation even it if fails, just added the ignore_value.
Michal Privoznik [Wed, 23 Sep 2015 12:13:53 +0000 (14:13 +0200)]
virDomainRestore: Don't keep transient domains around
So while working on my previous patches, I've noticed that
virDomainRestore implementation in qemu and test drivers has the
same problem as I am fixing.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
So, you want to create a domain from XML. The domain already
exists in libvirt's database of domains. It's okay, because name
and UUID matches. However, on domain startup, internal
representation of the domain is overwritten with your XML even
though we claim that the XML you've provided is a transient one.
The bug is to be found across nearly all the drivers.
Le sigh.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Okay, so we allow users to 'virsh create' an already existing
domain, providing completely different XML than the one stored in
Libvirt. Well, as long as name and UUID matches. However, in some
drivers the code that handles errors unconditionally removes the
domain that failed to start even though the domain might have
been persistent. Fortunately, the domain is removed just from the
internal list of domains and the config file is kept around.
Steps to reproduce:
1) virsh dumpxml $dom > /tmp/dom.xml
2) change XML so that it is still parse-able but won't boot, e.g.
change guest agent path to /foo/bar
3) virsh create /tmp/dom.xml
4) virsh dumpxml $dom
5) Observe "No such domain" error
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Michal Privoznik [Wed, 23 Sep 2015 08:03:10 +0000 (10:03 +0200)]
qemuDomainDeviceDefPostParse: Drop useless checks
Now that virQEMUDriverCreateXMLConf is never called with NULL
(after 086f37e97aab) we can safely drop useless check in
qemuDomainDeviceDefPostParse as we are guaranteed to be always
called with the driver initialized. Therefore checking if driver
is NULL makes no sense. Moreover, if we mix it with direct driver
dereference. And after that, we are sure that nor @cfg will be
NULL, therefore we can drop checks for that too.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Peter Krempa [Fri, 11 Sep 2015 19:34:24 +0000 (21:34 +0200)]
virsh: Fix job status indicator for 0 length block jobs
Although 0 length block jobs aren't entirely useful, the output of virsh
blockjob is empty due to the condition that suppresses the output for
migration jobs that did not start. Since the only place that actually
uses the condition that suppresses the output is in migration, let's
move the check there and thus add support for 0 of 0 equaling to 100%.
Peter Krempa [Wed, 23 Sep 2015 12:19:06 +0000 (14:19 +0200)]
qemu: Refresh memory size only on fresh starts
Qemu unfortunately doesn't update internal state right after migration
and so the actual balloon size as returned by 'query-balloon' are
invalid for a while after the CPUs are started after migration. If we'd
refresh our internal state at this point we would report invalid current
memory size until the next balloon event would arrive.
Peter Krempa [Wed, 23 Sep 2015 11:07:03 +0000 (13:07 +0200)]
qemu: Align memory module sizes to 2MiB
My original implementation was based on a qemu version that still did
not have all the checks in place. Using sizes that would align to odd
megabyte increments will produce the following error:
qemu-kvm: -device pc-dimm,node=0,memdev=memdimm0,id=dimm0: backend memory size must be multiple of 0x200000
qemu-kvm: -device pc-dimm,node=0,memdev=memdimm0,id=dimm0: Device 'pc-dimm' could not be initialized
Introduce an alignment retrieval function for memory devices and use it
to align the devices separately and modify a test case to verify it.
After my "client rpc: Report proper error for keepalive disconnections"
patch, virsh would no long print a warning when it closes a connection
to a daemon after a keepalive timeout. Although the warning
virsh # 2015-09-15 10:59:26.729+0000: 642080: info :
libvirt version: 1.2.19
2015-09-15 10:59:26.729+0000: 642080: warning :
virKeepAliveTimerInternal:143 : No response from client
0x7efdc0a46730 after 1 keepalive messages in 2 seconds
was pretty ugly, it was still useful. This patch brings the useful part
back while making it much nicer:
virsh # error: Disconnected from qemu:///system due to keepalive timeout
Even though we hit an error in client's IO loop, we still want to
process any pending data. So instead of reporting the error right away,
we can finish the current iteration and report the error once we're done
with it. Note that the error is stored in client->error by
virNetClientMarkClose so we don't need to worry about it being reset or
rewritten by any API we call in the meantime.
CC qemu/libvirt_driver_qemu_impl_la-qemu_command.lo
qemu/qemu_capabilities.c:46:27: fatal error: qemu_capspriv.h: No such file or directory
#include "qemu_capspriv.h"
Michal Privoznik [Tue, 22 Sep 2015 14:27:57 +0000 (16:27 +0200)]
tests: Avoid use of virQEMUDriverCreateXMLConf(NULL)
We use the function to create a virDomainXMLOption object that is
required for some functions. However, we don't pass the driver
pointer to the object anywhere - rather than pass NULL. This
causes trouble later when parsing a domain XML and calling post
parse callbacks:
Program received signal SIGSEGV, Segmentation fault.
0x000000000043fa3e in qemuDomainDefPostParse (def=0x7d36c0, caps=0x7caf10, opaque=0x0) at qemu/qemu_domain.c:1043
1043 qemuCaps = virQEMUCapsCacheLookup(driver->qemuCapsCache, def->emulator);
(gdb) bt
#0 0x000000000043fa3e in qemuDomainDefPostParse (def=0x7d36c0, caps=0x7caf10, opaque=0x0) at qemu/qemu_domain.c:1043
#1 0x00007ffff2928bf9 in virDomainDefPostParse (def=0x7d36c0, caps=0x7caf10, xmlopt=0x7c82c0) at conf/domain_conf.c:4269
#2 0x00007ffff294de04 in virDomainDefParseXML (xml=0x7da8c0, root=0x7dab80, ctxt=0x7da980, caps=0x7caf10, xmlopt=0x7c82c0, flags=0) at conf/domain_conf.c:16400
#3 0x00007ffff294e5b5 in virDomainDefParseNode (xml=0x7da8c0, root=0x7dab80, caps=0x7caf10, xmlopt=0x7c82c0, flags=0) at conf/domain_conf.c:16582
#4 0x00007ffff294e424 in virDomainDefParse (xmlStr=0x0, filename=0x7c7ef0 "/home/zippy/work/libvirt/libvirt.git/tests/securityselinuxlabeldata/disks.xml", caps=0x7caf10, xmlopt=0x7c82c0, flags=0) at conf/domain_conf.c:16529
#5 0x00007ffff294e4b2 in virDomainDefParseFile (filename=0x7c7ef0 "/home/zippy/work/libvirt/libvirt.git/tests/securityselinuxlabeldata/disks.xml", caps=0x7caf10, xmlopt=0x7c82c0, flags=0) at conf/domain_conf.c:16553
#6 0x00000000004303ca in testSELinuxLoadDef (testname=0x53c929 "disks") at securityselinuxlabeltest.c:192
#7 0x00000000004309e8 in testSELinuxLabeling (opaque=0x53c929) at securityselinuxlabeltest.c:313
#8 0x0000000000431207 in virtTestRun (title=0x53c92f "Labelling \"disks\"", body=0x430964 <testSELinuxLabeling>, data=0x53c929) at testutils.c:211
#9 0x0000000000430c5d in mymain () at securityselinuxlabeltest.c:373
#10 0x00000000004325c2 in virtTestMain (argc=1, argv=0x7fffffffd7e8, func=0x430b4a <mymain>) at testutils.c:863
#11 0x0000000000430deb in main (argc=1, argv=0x7fffffffd7e8) at securityselinuxlabeltest.c:381
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Michal Privoznik [Tue, 22 Sep 2015 14:12:39 +0000 (16:12 +0200)]
qemuTestDriverInit: init the driver lock too
Even though usage of the lock is limited to a very few cases,
it's still needed. Therefore we should initialize it too.
Otherwise we may get some random test failures:
==1204== Conditional jump or move depends on uninitialised value(s)
==1204== at 0xEF7F7CF: pthread_mutex_lock (in /lib64/libpthread-2.20.so)
==1204== by 0x9CA89A5: virMutexLock (virthread.c:89)
==1204== by 0x450B2A: qemuDriverLock (qemu_conf.c:83)
==1204== by 0x45549C: virQEMUDriverGetConfig (qemu_conf.c:869)
==1204== by 0x448E29: qemuDomainDeviceDefPostParse (qemu_domain.c:1240)
==1204== by 0x9CC9B13: virDomainDeviceDefPostParse (domain_conf.c:4224)
==1204== by 0x9CC9B91: virDomainDefPostParseDeviceIterator (domain_conf.c:4251)
==1204== by 0x9CC7843: virDomainDeviceInfoIterateInternal (domain_conf.c:3440)
==1204== by 0x9CC9C25: virDomainDefPostParse (domain_conf.c:4276)
==1204== by 0x9CEEE03: virDomainDefParseXML (domain_conf.c:16400)
==1204== by 0x9CEF5B4: virDomainDefParseNode (domain_conf.c:16582)
==1204== by 0x9CEF423: virDomainDefParse (domain_conf.c:16529)
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Peter Krempa [Mon, 21 Sep 2015 16:10:55 +0000 (18:10 +0200)]
qemu: ppc64: Align memory sizes to 256MiB blocks
For some machine types ppc64 machines now require that memory sizes are
aligned to 256MiB increments (due to the dynamically reconfigurable
memory). As now we treat existing configs reasonably in regards to
migration, we can round all the sizes unconditionally. The only drawback
will be that the memory size of a VM can potentially increase by
(256MiB - 1byte) * number_of_NUMA_nodes.
Peter Krempa [Thu, 17 Sep 2015 06:14:05 +0000 (08:14 +0200)]
qemu: command: Align memory sizes only on fresh starts
When we are starting a qemu process for an incomming migration or
snapshot reloading we should not modify the memory sizes in the domain
since we could potentially change the guest ABI that was tediously
checked before. Additionally the function now updates the initial memory
size according to the NUMA node size, which should not happen if we are
restoring state.
Peter Krempa [Fri, 18 Sep 2015 15:24:32 +0000 (17:24 +0200)]
conf: Don't always recalculate initial memory size from NUMA size totals
When implementing memory hotplug I've opted to recalculate the initial
memory size (contents of the <memory> element) as a sum of the sizes of
NUMA nodes when NUMA was enabled. This was based on an assumption that
qemu did not allow starting when the NUMA node size total didn't equal
to the initial memory size. Unfortunately the check was introduced to
qemu just lately.
This patch uses the new XML parser flag to decide whether it's safe to
update the memory size total from the NUMA cell sizes or not.
As an additional improvement we now report an error in case when the
size of hotplug memory would exceed the total memory size.
The rest of the changes assures that the function is called with correct
flags.
Peter Krempa [Thu, 13 Aug 2015 14:39:28 +0000 (16:39 +0200)]
conf: Pre-calculate initial memory size instead of always calculating it
Add 'initial_memory' member to struct virDomainMemtune so that the
memory size can be pre-calculated once instead of inferring it always
again and again.
Separating of the fields will also allow finer granularity of decisions
in later patches where it will allow to keep the old initial memory
value in cases where we are handling incomming migration from older
versions that did not always update the size from NUMA as the code did
previously.
The change also requires modification of the qemu memory alignment
function since at the point where we are modifying the size of NUMA
nodes the total size needs to be recalculated too.
The refactoring done in this patch also fixes a crash in the hyperv
driver that did not properly initialize def->numa and thus
virDomainNumaGetMemorySize(def->numa) crashed.
In summary this patch should have no functional impact at this point.
Peter Krempa [Tue, 15 Sep 2015 14:59:23 +0000 (16:59 +0200)]
conf: Split memory related post parse stuff into separate function
The post parse func is growing rather large. Since later patches will
introduce more logic in the memory post parse code, split it into a
separate handler.
Peter Krempa [Tue, 15 Sep 2015 15:04:55 +0000 (17:04 +0200)]
conf: Add XML parser flag that will allow us to do incompatible updates
Add a new parser flag that will mark code paths that parse XML files
wich will not be used with existing VM state so that post parse
callbacks can possibly do ABI incompatible changes if needed.
projects / people / liuw / libxenctrl-split / libvirt.git / log