Since the balloon driver does not guarantee that it returns memory to
the host, using the value in the audit message is not a good idea.
This patch removes auditing from updating the balloon size and reports
the total physical size at startup.
The code which generates paths for UNIX socket blindly used target name
without checking if it was set. Thus for the following device XML
<channel type='unix'>
<source mode='bind'/>
<target type='virtio'/>
</channel>
we would generate "/var/lib/libvirt/qemu/channel/target/NAME.(null)"
path which works but is not really correct. Let's not use the
".target_name" suffix at all if target name is not set.
https://bugzilla.redhat.com/show_bug.cgi?id=1226854
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
While CPU0 was made unpluggable in Linux a while ago it's not desirable
to unplug it since some parts of the kernel (suspend-to-ram) still
depend on it.
This patch fixes the vCPU selection code in libvirt so that it will not
be disabled.
The target type comparison in qemuDomainDetachChrDevice
used the VIR_DOMAIN_CHR_SERIAL_TARGET_TYPE enum, so virtio-serial
addresses were not freed properly for channel devices.
Call qemuDomainReleaseDeviceAddress uncoditionally and decide
based on the address type instead of the target/device types.
Also check the device type when deciding what type the address should
be. Commit 9807c47 (aiming to fix another error in address allocation)
only checked the target type, but its value is different for different
device types. This resulted in an error when trying to attach
a channel with target type 'virtio':
error: Failed to attach device from channel-file.xml
error: internal error: virtio serial device has invalid address type
Make the logic for releasing the address dependent only on
* the address type
* whether it was allocated earlier
to avoid copying the device and target type checks.
https://bugzilla.redhat.com/show_bug.cgi?id=1230039
Signed-off-by: Luyao Huang <lhuang@redhat.com>
Signed-off-by: Ján Tomko <jtomko@redhat.com>
def->vcpus was never updated after successfully changing the live
vcpu count of a domain. Subsequent queries for vcpu info would
return incorrect results. E.g.:
virsh vcpucount test
maximum config 4
maximum live 4
current config 4
current live 4
virsh setvcpus test 2
virsh vcpucount test
maximum config 4
maximum live 4
current config 4
current live 4
After patch, live current config is reported correctly:
virsh vcpucount test
maximum config 4
maximum live 4
current config 4
current live 2
While fixing this, noticed that the live config was not saved
to cfg->stateDir via virDomainSaveStatus. Save the live config
and change error handling of virDomainSave{Config,Status} to
log a message via VIR_WARN, instead of failing the entire
DomainSetVcpusFlags operation.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
The libxl driver always uses virDomainObj->def when formatting
the domain XML description. Use virDomainObj->newDef when
--inactive flag is set.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
libxlDomainCreateXML() would remove a persistent domain if
libxlDomainStart() failed. Check if domain is persistent
before removing.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
When restarting libvirtd and reconnecting to running domains,
libxlReconnectDomain() would unconditionally set the domain state
to VIR_DOMAIN_RUNNING, overwriting the state maintained in
$statedir/<domname>.xml. A domain in a paused state would have
the state changed to running, even though it was actually in a
paused state.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
https://bugzilla.redhat.com/show_bug.cgi?id=1201760
When the domain "<on_crash>coredump-destroy</on_crash>" is set, the
domain wasn't being destroyed, rather it was being rebooted.
Add VIR_DOMAIN_LIFECYCLE_CRASH_COREDUMP_DESTROY to the list of
on_crash types that cause "-no-reboot" to be added to the qemu
command line.
Although defined the same way, fortunately there hadn't been any deviation.
Ensure any assignments to onCrash use VIR_DOMAIN_LIFECYCLE_CRASH_* defs and
not VIR_DOMAIN_LIFECYCLE_* defs
https://bugzilla.redhat.com/show_bug.cgi?id=1232606
Since an mpath pool contains all the Multipath devices on a host, allowing
more than one defined on a host at a time should be disallowed under the
policy of disallowing duplicate source pools for the host.
Adjust to docs to clarify the Multipath target path value usage for both
the storage driver (only 1 pool per host) and formatstorage references
(ignore the target element in favor of the default target mapping of
/dev/mapper).
https://bugzilla.redhat.com/show_bug.cgi?id=1230664
Per the devmapper docs, use "/dev/mapper" or "/dev/dm-n" in order to
determine if a device is under control of DM Multipath.
So add "/dev/mapper" to the virFileExists, leaving the "/dev/mpath"
as a "legacy" option since it appears for a while it was the preferred
mechanism, but is no longer maintained
https://bugzilla.redhat.com/show_bug.cgi?id=1201143
The formatdomain.html description for <disk> device 'lun' indicates that
it must be either a type 'block' or type 'network' with protocol 'iscsi';
however, we did not make that check until domain startup.
This caused issues for virt-manager which had an unexpected failure at
run time rather config time.
This patch adds a check in post part disk device checking for the specific
and supported lun types as well as adjusting the test failure to be for
parse config rather than run time.
Libvirt periodically refreshes all volumes in a storage pool, including
the volumes being cloned.
While cloning a storage volume from parent, we drop pool locks. Subsequent
volume refresh sometimes changes allocation for an ongoing copy, and leads
to corrupt images.
Fix: Introduce a shadow volume that isolates the volume object under refresh
from the base which has a copy ongoing.
Signed-off-by: Prerna Saxena <prerna@linux.vnet.ibm.com>
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Spice events have mostly similar information present in the event JSON
but they differ in the name of the element containing the port.
The JSON event also provides connection ID which might be useful in the
future.
This patch splits up the event parser code into two functions and the
SPICE reimplements the event parsing with correct names and drops the
VNC only stuff.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1236585
https://bugzilla.redhat.com/show_bug.cgi?id=1227664
If the requested format type for the new entry in the file system pool
is a 'dir', then be sure to set the vol->type correctly as would be done
when the pool is refreshed.
Make sure we only assign the default spicevmc channel name to spicevmc
virtio channels. Caused by commits 3269ee65 and 1133ee2b, which moved
the assignment from XML parsing code to QEMU but failed to keep the
logic.
https://bugzilla.redhat.com/show_bug.cgi?id=1179680
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Certain PCI buses don't support hotplug, and when automatically
assigning PCI addresses for devices, libvirt is very conservative in
its assumptions about whether or not a device will need to be
hotplugged/unplugged in the future. But if the user manually assigns
an address, they likely are aware of any hotplug requirements of the
device (or at least they should be).
In short, after this patch, automatically PCI address assignment will
assume that the device must be plugged in to a hot-pluggable slot, but
manually assignment can place the device in any bus that is
compatible, regardless of whether or not it supports hotplug. If the
user makes a mistake and plugs the device into a bus that doesn't
support hotplug, then later tries to do a hot-unplug, qemu will give
an appropriate error.
(in the future we may want to add a "hotpluggable" attribute to all
devices, with default being "yes" for autoassign, and "no" for manual
assign).
When support for the pcie-root and dmi-to-pci-bridge buses on a Q35
machinetype was added, I was concerned that even though qemu at the
time allowed plugging a PCI device into a PCIe port, that it might not
be supported in the future. To prevent painful backtracking in the
possible future where this happened, I disallowed such connections
except in a few specific cases requested by qemu developers (indicated
in the code with the flag VIR_PCI_CONNECT_TYPE_EITHER_IF_CONFIG).
Now that a couple years have passed, there is a clear message from
qemu that there is no danger in allowing PCI devices to be plugged
into PCIe ports. This patch eliminates
VIR_PCI_CONNECT_TYPE_EITHER_IF_CONFIG and changes the code to always
allow PCI->PCIe or PCIe->PCI connection *when the PCI address is
specified in the config. (For newly added devices that haven't yet
been given a PCI address, the auto-placement still prefers using the
correct type of bus).
The PCI case of the switch statement in this function contains another
switch statement with a case for each model. Currently every model
except pci-root and pcie-root has a check for index > 0 (since only
those two can have index==0), and the function should never be called
for those two anyway. If we move the check for !pci[e]-root to the top
of the pci case, then we can move the check for index > 0 out of the
individual model cases. This will save repeating that check for the
three new controller models about to be added.
Instead of using qemuMonitorJSONDevGetBlockExtent (which I plan to
remove later) extract the data in place.
Additionally add a flag that will be set when the wr_highest_offset was
extracted correctly so that callers can act according to that.
The test case addition should help make sure that everything works.
Function prlsdkGetStatsParam was missing a prototype or the static
keyword. I went with static since it built successfully.
Pushed as a build breaker fix.
Implemented counters:
VIR_DOMAIN_MEMORY_STAT_SWAP_IN
VIR_DOMAIN_MEMORY_STAT_SWAP_OUT
VIR_DOMAIN_MEMORY_STAT_MINOR_FAULT
VIR_DOMAIN_MEMORY_STAT_MAJOR_FAULT
VIR_DOMAIN_MEMORY_STAT_AVAILABLE
VIR_DOMAIN_MEMORY_STAT_ACTUAL_BALLOON
VIR_DOMAIN_MEMORY_STAT_UNUSED
Comments.
1. Use vzDomObjFromDomainRef/virDomainObjEndAPI pair to get domain
object as we use prlsdkGetStatsParam. See previous statistics
comments.
2. Balloon statistics is not applicable to containers. Fault
statistics for containers not provided in PCS6 yet.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
Comments.
Replace vzDomObjFromDomain/virObjectUnlock pair
to vzDomObjFromDomainRef/virDomainObjEndAPI as we
use prlsdkGetStatsParam. See previous statistics
comments.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
Make net device lookup by mac return sdk handle
instead of quite ephemeral enumeration index. After
this change there is no need anymore in special
function of removing device by enumeration index.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
Populate counters SDK currenly supports:
rx_bytes
rx_packets
tx_bytes
tx_packets
Comments.
Use vzDomObjFromDomainRef/virDomainObjEndAPI pair to get domain
object as we use prlsdkGetStatsParam that can release domain
object lock and thus we need a reference in case domain
is deleated meanwhile.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
So far the argument has not much meaning and was practically ignored.
This is not good since when doing memory hotplug, the size of desired
hugepage backing is passed in that argument. Taking closer look at the
tests I'm fixing reveals the bug. For instance, while the following is
in the test:
<memory model='dimm'>
<source>
<nodemask>1-3</nodemask>
<pagesize unit='KiB'>4096</pagesize>
</source>
<target>
<size unit='KiB'>524287</size>
<node>0</node>
</target>
<address type='dimm' slot='0' base='0x100000000'/>
</memory>
the generated commandline corresponding to this XML was:
-object memory-backend-ram,id=memdimm0,size=536870912,\
host-nodes=1-3,policy=bind
Have you noticed? Yes, memory-backend-ram! Nothing can be further away
from the right answer. The hugepage backing is requested in the XML
and we happily ignore it. This is just not right. It's
memory-backend-file which should have been used:
-object memory-backend-file,id=memdimm0,prealloc=yes,\
mem-path=/dev/hugepages4M/libvirt/qemu,size=536870912,\
host-nodes=1-3,policy=bind
The problem is, that @pagesize passed to qemuBuildMemoryBackendStr
(where this part of commandline is built) was ignored. The hugepage to
back memory was searched only and only by NUMA nodes pinning. This
works only for regular guest NUMA nodes.
Then, I'm changing the hugepages size in the test XMLs too. This is
simply because in the test suite we create dummy mount points just for
2M and 1G hugepages. And in the test 4M was requested. I'm sticking to
2M, but 1G should just work too.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=1196644
This function constructs the backend (host facing) part of the
memory device. At the beginning, the configured hugepages are
searched to find the best match for given guest NUMA node.
Configured hugepages can have a @nodeset attribute to specify on
which guest NUMA nodes should be the hugepages backing used.
There is, however, one 'corner case'. Users may just tell 'use
hugepages to back all the nodes'. In other words:
<memoryBacking>
<hugepages/>
</memoryBacking>
<cpu>
<numa>
<cell id='0' cpus='0-1' memory='1024000' unit='KiB'/>
</numa>
</cpu>
Our code fails in this case. Well, since there's no @nodeset (nor
any <page/> child element to <hugepages/>) we fail to lookup the
default hugepage size to use.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=1235116
According to our XML definition, zero is as valid as any other value.
Mainly because it should be kernel-agnostic.
Signed-off-by: Luyao Huang <lhuang@redhat.com>
Drop internal data structures and use the proper fields in virDomainDef.
This allows to greatly simplify the code and allows to remove the
private data structure that was holding just redundant data.
This patch also fixes the bogous output where we'd report that a fresh
VM without vCPU pinning would not run on all vcpus.
Only self-locking APIs are used and the pointer is immutable so there's
no need to lock the driver to access the domain list.
This patch removes locking partially for everything that will not be
converted to testDomObjFromDomain in the next patch.
Make testObjectEventQueue tolerant to NULL @event and move it so that it
does not require a prototype. Additionally we are now able to remove
locking when accessing driver->eventState, since it's using self-locking
APIs and the pointer is immutable.
When migration fails in qemuMigrationPrepareAny, we unconditionally call
qemuDomainRemoveInactive, which should only be called for transient
domains. The check for !vm->persistent was accidentally removed by
commit 540c339.
Signed-off-by: Luyao Huang <lhuang@redhat.com>
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
In the e6d180f07f commit the parallels driver was renamed to vz.
However, there was a commit merged later, which was sent to the list
before the rename. The other commit is 6de12b026b. Fix all the
missing renames.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
We already enable the parser option to detect invalid UTF-8, but
didn't test it. Also, JSON states that behavior of an object
with a duplicated key is undefined; we chose to reject it, but
were not testing it.
With the enhanced tests in place, we can simplify yajl2
initialization by relying on parser defaults being sane.
* src/util/virjson.c (virJSONValueFromString): Simplify.
* tests/jsontest.c (mymain): Test more bad usage.
Signed-off-by: Eric Blake <eblake@redhat.com>
Since older yajl ignores trailing garbage, a client can cause
problems by intentionally ending the wrapper array early. Since
we already track nesting, it's not too much harder to reject
invalid nesting pops.
* src/util/virjson. (_virJSONParser): Add field.
(virJSONValueFromString): Set witness.
(virJSONParserHandleEndArray): Use it to catch abuse.
* tests/jsontest.c (mymain): Test it.
Signed-off-by: Eric Blake <eblake@redhat.com>
Yajl 2 has a nice feature that it can be configured whether to
allow multiple JSON objects parsed from a single stream, defaulting
to off. And yajl 1.0.12 at least provided a way to tell if all
input bytes were parsed, or if trailing bytes remained after a
valid JSON object was parsed. But we target RHEL 6 yajl 1.0.7,
which has neither of these. So fake it by always parsing '[...]'
instead, so that trailing garbage either trips up the array parse,
or is easily detected when unwrapping the result.
* src/util/virjson.c (virJSONValueFromString): With older json,
wrap text to avoid trailing garbage.
* tests/jsontest.c (mymain): Add tests for this.
Signed-off-by: Eric Blake <eblake@redhat.com>
We have been allowing javascript style comments in JSON ever
since commit 9428f2c (v0.7.5), but qemu doesn't send them, and
they are not strict JSON. Reject them for now; if we can later
prove that it is worthwhile, we can reinstate it at that point
(or even make it conditional, by adding a bool parameter to
the libvirt entry point).
* src/util/virjson.c (virJSONValueFromString): Don't enable
comment parsing.
* tests/jsontest.c (mymain): Test it.
Signed-off-by: Eric Blake <eblake@redhat.com>
Commit ceb496e5 fails on RHEL 6, with yajl 1.0.7, because that
version of yajl returns yajl_status_insufficient_data when the
parser is waiting for the rest of a token (this enum value was
dropped in yajl 2, so we have to wrap it). It also exposes a
problem where older yajl silently ignores trailing garbage after
a successful parse, so this patch works around that by changing
the testsuite. Another more invasive patch can add tighter
semantics to json parsing, but this is sufficient for a minimal
clean backport.
While touching this, fix up our error message cleanup. Yajl
documents that error messages produced by yajl_get_error()
MUST be cleaned with yajl_free_error(); this is certainly
true if we were to pass non-NULL allocator callbacks during
yajl_alloc(), but probably harmless in our usage of passing
NULL. But better safe than sorry.
* src/util/virjson.c (virJSONValueFromString): Allow different
error code. Use canonical cleanup of error message.
(VIR_YAJL_STATUS_OK): New helper macro.
* tests/jsontest.c (mymain): Wrap text to avoid difference in
trailing garbage handling
Signed-off-by: Eric Blake <eblake@redhat.com>
In this patch we add VIR_DOMAIN_DEVICE_NET handlers implementation
for domainAttachDevice and domainDetachDevice callbacks.
As soon as we don't support this operation for hypervisor type domains,
we implement this functionality for containers only.
In detach procedure we find network device by MAC address.
Because PrlVmDevNet_GetMacAddress() returns MAC as a UTF-8 encoded
null-terminated string, we use memcmp() to compare it.
Also we remove corresponding virtual network by prlsdkDelNetAdapter call.
This patch provides support for the new watchdog model "diag288".
Signed-off-by: Boris Fiuczynski <fiuczy@linux.vnet.ibm.com>
Reviewed-by: Daniel Hansel <daniel.hansel@linux.vnet.ibm.com>
Reviewed-by: Stefan Zimmermann <stzi@linux.vnet.ibm.com>
Reviewed-by: Tony Krowiak <akrowiak@linux.vnet.ibm.com>
This patch provides support for a new watchdog action "inject-nmi" which
allows to define an inject of a non-maskable interrupt into a guest.
Signed-off-by: Boris Fiuczynski <fiuczy@linux.vnet.ibm.com>
Reviewed-by: Daniel Hansel <daniel.hansel@linux.vnet.ibm.com>
Reviewed-by: Stefan Zimmermann <stzi@linux.vnet.ibm.com>
Reviewed-by: Tony Krowiak <akrowiak@linux.vnet.ibm.com>
Related to :
https://bugzilla.redhat.com/show_bug.cgi?id=1171933
Rather than ignore the return status from virStorageBackendSCSIFindLUs,
cause a failure to start the pool if a -1 is returned. Issue was noted
during testing of the bz for iscsi that 'scsi' and 'fc' pools don't fail.
Commit id '1feaccf0' attempted to handle an empty secrettype value; however,
it made a mistake by processing the secretType as if it was the original
secrettype string. The 'secretType' is actually whether 'usage' or 'uuid'
was used.
Thus adjust part of the change to make the same check for def->src->type !=
VIR_STORAGE_TYPE_VOLUME before setting auth_secret_usage from the
secrettype field.
Luckily the aforementioned commits misdeed would be overwritten by the
call to virStorageTranslateDiskSourcePool
Just refactor existing code to use a child buf instead of
check all element before format <blkiotune> and <cputune>.
This will avoid the more and more bigger element check during
we introduce new elements in <blkiotune> and <cputune> in the
future.
Signed-off-by: Luyao Huang <lhuang@redhat.com>
Functions like virDomainOpenConsole() and virDomainOpenChannel() accept
NULL as a dev_name parameter. Try using alias for the error message if
dev_name is not specified.
Before:
error: internal error: character device <null> is not using a PTY
After:
error: internal error: character device serial0 is not using a PTY
Signed-off-by: Luyao Huang <lhuang@redhat.com>
Commit id '832a9256' adjusted the code to recognize when the default
type of "unknown" was provided as the format type and to use "dos" if
found. Since the pool is built with "dos" and it could cause some
confusion when formatting the XML after building by seeing "unknown"
in the output, let's just adjust the pool's setting to "dos" so that
subsequent formats will see the value.
By trying to lead the way of clean includes, I sorted the lines
alphabetically and that is a problem for mingw builds with gnulib.
As 'configmake.h' defines DATADIR and 'datatypes.h' transitively
includes 'winsock.h' that uses 'DATADIR' as a name for a struct,
it's enough to reorder those.
Even though this might be worked around in gnulib later on, this
fixes the build for now.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
The SCSI Architecture Model defines a logical unit address
as 64-bits in length, so change the field accordingly so
that the entire value could be stored.
Signed-off-by: Eric Farman <farman@linux.vnet.ibm.com>
The address elements are all unsigned integers, so we should
use the appropriate print directive when printing it.
Signed-off-by: Eric Farman <farman@linux.vnet.ibm.com>
The SCSI address element attributes bus, target, and unit are expected
to be positive values, so make sure no one provides a negative value since
the value is stored as an unsigned.
Signed-off-by: Eric Farman <farman@linux.vnet.ibm.com>
Rather than grabbing an arbitrary JSON value and then checking
if it has the right type, we might as well request the correct
type to begin with.
* src/qemu/qemu_monitor_json.c (qemuMonitorJSONIOProcessEvent)
(qemuMonitorJSONCommandWithFd, qemuMonitorJSONHandleGraphics)
(qemuMonitorJSONGetStatus, qemuMonitorJSONExtractCPUInfo)
(qemuMonitorJSONGetVirtType, qemuMonitorJSONGetBalloonInfo)
(qemuMonitorJSONGetMemoryStats)
(qemuMonitorJSONDevGetBlockExtent)
(qemuMonitorJSONGetOneBlockStatsInfo)
(qemuMonitorJSONGetAllBlockStatsInfo)
(qemuMonitorJSONBlockStatsUpdateCapacityOne)
(qemuMonitorJSONBlockStatsUpdateCapacity)
(qemuMonitorJSONGetBlockExtent)
(qemuMonitorJSONGetMigrationStatusReply)
(qemuMonitorJSONGetDumpGuestMemoryCapability)
(qemuMonitorJSONAddFd, qemuMonitorJSONQueryRxFilterParse)
(qemuMonitorJSONExtractChardevInfo)
(qemuMonitorJSONDiskNameLookupOne)
(qemuMonitorJSONDiskNameLookup)
(qemuMonitorJSONGetAllBlockJobInfo)
(qemuMonitorJSONBlockIoThrottleInfo, qemuMonitorJSONGetVersion)
(qemuMonitorJSONGetMachines, qemuMonitorJSONGetCPUDefinitions)
(qemuMonitorJSONGetCommands, qemuMonitorJSONGetEvents)
(qemuMonitorJSONGetKVMState, qemuMonitorJSONGetObjectTypes)
(qemuMonitorJSONGetObjectListPaths)
(qemuMonitorJSONGetObjectProps, qemuMonitorJSONGetTargetArch)
(qemuMonitorJSONGetMigrationCapabilities)
(qemuMonitorJSONGetStringArray, qemuMonitorJSONAttachCharDev)
(qemuMonitorJSONGetCPUx86Data, qemuMonitorJSONGetIOThreads)
(qemuMonitorJSONGetMemoryDeviceInfo): Use shorter idioms.
Signed-off-by: Eric Blake <eblake@redhat.com>
While working in qemu_monitor_json, I repeatedly found myself
getting a value then checking if it was an object. Add some
wrappers to make this task easier.
* src/util/virjson.c (virJSONValueObjectGetByType)
(virJSONValueObjectGetObject, virJSONValueObjectGetArray): New
functions.
(virJSONValueObjectGetString, virJSONValueObjectGetNumberInt)
(virJSONValueObjectGetNumberUint)
(virJSONValueObjectGetNumberLong)
(virJSONValueObjectGetNumberUlong)
(virJSONValueObjectGetNumberDouble)
(virJSONValueObjectGetBoolean): Simplify.
(virJSONValueIsNull): Change return type.
* src/util/virjson.h: Reflect changes.
* src/libvirt_private.syms (virjson.h): Export them.
* tests/jsontest.c (testJSONLookup): New test.
Signed-off-by: Eric Blake <eblake@redhat.com>
I was adding a JSON test, and was shocked to find out our parser
treated the input string of "1" as invalid JSON. It turns out
that YAJL specifically documents that it buffers input, and that
if the last input read could be a prefix to a longer token, then
you have to explicitly tell the parser that the buffer has ended
before that token will be processed.
It doesn't help that yajl 2 renamed the function from what it was
in yajl 1.
* src/util/virjson.c (virJSONValueFromString): Complete parse, in
case buffer ends in possible token prefix.
* tests/jsontest.c (mymain): Expose the problem.
Signed-off-by: Eric Blake <eblake@redhat.com>
There's a small formatting problem in the function. One line is
not correctly indented. Fix this.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
This capability specifies that "virt" machine on ARM has PCI controller. Enabled when version is at least 2.3.0.
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
If vm->def->cputune.nvcpupin is 0 in libxlDomainSetVcpuAffinities (as
seems to be the case on arm) then the VIR_FREE after cleanup: would be
operating on an uninitialised pointer in map.map.
Fix this by using libxl_bitmap_init and libxl_bitmap_dispose in the
appropriate places (like VIR_FREE, libxl_bitmap_dispose is also
idempotent, so there is no double free on exit from the loop).
libxl_bitmap_dispose is slightly preferable since it also sets
map.size back to 0, avoiding a potential source of confusion.
This fixes the crashes we've been seeing in the Xen automated tests on
ARM.
I had a glance at the handful of other users of libxl_bitmap and none
of them looked to have a similar issue.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
When a connection to the destination host during a p2p migration drops,
we know we will have to cancel the migration; it doesn't make sense to
waste resources by trying to finish the migration. We already do so
after sending "migrate" command to QEMU and we should do it while
waiting for drive mirrors to become ready too.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Checking status of all part of migration and aborting it when something
failed is a complex thing which makes the waiting loop hard to read.
This patch moves all the checks into a separate function similarly to
what was done for drive mirror loops.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Instead of passing current job name to several functions which already
know what the current job is we can generate the name where we actually
need to use it.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Once we start waiting for migration events instead of polling
query-migrate, priv->job.current will not be regularly updated anymore
because we will get the current status directly from the events. Thus
virDomainGetJob{Info,Stats} will have to query QEMU, but they can't just
blindly update priv->job.current structure. This patch introduces
qemuMigrationFetchJobStatus which just fills in a caller supplied
structure and makes qemuMigrationUpdateJobStatus a tiny wrapper around
it.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Move common parts of qemuDomainGetJobInfo and qemuDomainGetJobStats into
a separate API (qemuDomainGetJobStatsInternal).
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
QEMU_CAPS_SEAMLESS_MIGRATION capability says QEMU supports
SPICE_MIGRATE_COMPLETED event. Thus we can just drop all code which
polls query-spice and replace it with waiting for the event.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
When libvirtd is restarted during migration, we properly cancel the
ongoing migration (unless it managed to almost finished before the
restart). But if we were also migrating storage using NBD, we would
completely forget about the running disk mirrors.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
"query-block-jobs" QMP command returns all running block jobs at once,
while qemuMonitorBlockJobInfo would only report one. This is not very
nice in case we need to check several block jobs. This patch refactors
the monitor code to always parse all block jobs and store them in a
hash.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
So that they can format private data (e.g., disk private data) stored
elsewhere in the domain object.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
This patch reverts commit 76c61cdca2.
VIR_DOMAIN_DISK_MIRROR_STATE_ABORT says we asked for a block job to be
aborted rather than saying it was aborted. Let's just use
VIR_DOMAIN_DISK_MIRROR_STATE_NONE consistently whenever a block job
finishes since no caller depends on VIR_DOMAIN_DISK_MIRROR_STATE_ABORT
(anymore) to check whether a block job failed or it was cancelled.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Abort migration as soon as we detect that some of the disk mirrors
failed. There's no sense in trying to finish memory migration first.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Instead of cancelling disk mirrors sequentially, let's just call
block-job-cancel for all migrating disks and then wait until all
disappear.
In case we cancel disk mirrors at the end of successful migration we
also need to check all block jobs completed successfully. Otherwise we
have to abort the migration.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
By switching block jobs to use domain conditions, we can drop some
pretty complicated code in NBD storage migration.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Because we are polling we may detect some errors after we asked QEMU for
migration status even though they occurred before. If this happens and
QEMU reports migration completed successfully, we would happily report
the migration succeeded even though we should have cancelled it because
of the other error.
In practise it is not a big issue now but it will become a much bigger
issue once the check for storage migration status is moved inside the
loop in qemuMigrationWaitForCompletion.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
The wrapper is useful for calling qemuBlockJobEventProcess with the
event details stored in disk's privateData, which is the most likely
usage of qemuBlockJobEventProcess.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Complex jobs, such as migration, need to monitor several events at once,
which is impossible when each of the event uses its own condition
variable. This patch adds a single condition variable to each domain
object. This variable can be used instead of the other event specific
conditions.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Well, if a server is being destructed, all underlying services and
their sockets should disappear with it. But due to bug in our
implementation this is not the case. Yes, we are closing the sockets,
but that's not enough. We must also:
1) Unregister them from the event loop
2) Unref the service for each socket
The last step is needed, because each socket callback holds a
reference to the service object. Since in the first step we are
unregistering the callbacks, they no longer need the reference.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Although highly unlikely, nobody says that virEventAddHandle()
can't return 0 as a handle to socket callback. It can't happen
with our default implementation since all watches will have value
1 or greater, but users can register their own callback functions
(which can re-use unused watch IDs for instance). If this is the
case, weird things may happen.
Also, there's a little bug I'm fixing too, upon
virNetSocketRemoveIOCallback(), the variable holding callback ID
was not reset. Therefore calling AddIOCallback() once again would
fail. Not that we are doing it right now, but we might.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
When going through the code I've notice that
virNetSocketAddIOCallback() increases the reference counter of
@socket. However, its counter part RemoveIOCallback does not. It took
me a while to realize this disproportion. The AddIOCallback registers
our own callback which eventually calls the desired callback and then
unref the @sock. Yeah, a bit complicated but it works. So, lets note
this hard learned fact in a comment in RemoveIOCallback().
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=1224233
Currently it's not possible to determine the difference between a
fatal memory allocation or failure to open/read the directory error
with a perhaps less fatal, I didn't find the "block" device in the
directory (which may be a disk entry without a block device).
In the case of the latter, we shouldn't cause failure to continue
searching in the caller (virStorageBackendSCSIFindLUs), rather we
should allow trying reading the next directory entry.
Signed-off-by: John Ferlan <jferlan@redhat.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=1203032
Implement a `migrate_disks' parameters for the QEMU driver. This multi-
value parameter can be used to explicitly specify what block devices
are to be migrated using the NBD server. Tunnelled migration using NBD
is to be done.
Signed-off-by: Pavel Boldin <pboldin@mirantis.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
The `virTypedParamsAddStringList' function provides interface to add a
NULL-terminated array of string values as a multi-value to the params.
Signed-off-by: Pavel Boldin <pboldin@mirantis.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Add multikey API:
* virTypedParamsFilter that filters all the parameters with specified name.
* virTypedParamsGetStringList that returns a list with all the values for
specified name and string type.
Signed-off-by: Pavel Boldin <pboldin@mirantis.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Allow multi-value parameters to be build using virTypedParamsAdd*
functions by removing check for duplicates.
Signed-off-by: Pavel Boldin <pboldin@mirantis.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
The `virTypedParamsValidate' function now can be instructed to allow
multiple entries for some of the keys. For this flag the type with
the `VIR_TYPED_PARAM_MULTIPLE' flag.
Add unit tests for this new behaviour.
Signed-off-by: Pavel Boldin <pboldin@mirantis.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
When playing with disk migration lately, I've noticed this warning in
domain logs:
WARNING: Image format was not specified for 'nbd://masina:49153/drive-virtio-disk0' and probing guessed raw.
Automatically detecting the format is dangerous for raw images, write operations on block 0 will be restricted.
Specify the 'raw' format explicitly to remove the restrictions.
So I started digging into qemu source code to see what has triggered
the warning. I'd expect qemu to know formats of guest's disks since we
tell them on command line. This lead me to qmp_drive_mirror() where
the following can be found:
if (!has_format) {
format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
}
So, format is automatically initialized from the disk iff mode !=
"existing". Unfortunately, in migration we are tied to use this mode
(NBD doesn't support creating new images). Therefore the only way to
avoid this warning is to pass format. The discussion on the mail-list [1]
resulted in the code that always forces NBD export as "raw" format.
[1] https://www.redhat.com/archives/libvir-list/2015-June/msg00153.html
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Pavel Boldin <pboldin@mirantis.com>
This function is returning a string (domain XML). Since d3ce7363
when it was first introduced, it was indented incorrectly:
static char
*qemuMigrationBeginPhase(..)
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
If virDomainObjGetDefs used in qemuDomainPinIOThread would fail the code
would jump to the 'cleanup' label after acquiring the job, thus the VM
would be locked forever.
Introduced in commit cac6d639.
The privileged flag will not change while the configuration might
change. Make the 'privileged' flag member of the driver again and mark
it immutable. Should that ever change add an accessor that will group
reads of the state.
virDomainObjGetOneDef will help to retrieve the correct definition
pointer from @vm in cases where VIR_DOMAIN_AFFECT_LIVE and
VIR_DOMAIN_AFFECT_CONFIG are mutually exclusive. The function simply
returns the correct pointer. This similarly to virDomainObjGetDefs will
greatly simplify the code.
If @flags contains only VIR_DOMAIN_AFFECT_CONFIG and @vm is active, the
function would return the active config rather than the persistent one
that it should return. This happened due to the fact that
virDomainObjGetDefs was checking the updated flags which may not contain
VIR_DOMAIN_AFFECT_LIVE if it is not requested even if @vm is active.
Additionally the function would not take the flags into account when
setting the pointers which was later used to determine whether the code
needs to update the given configuration.
The mistake was caught by the virt-test suite.
When hotplugging a memory device, there wasn't a check to determine
if there is a conflict with the address space being used by the to
be added memory device and any existing device which is disallowed by qemu.
This patch adds a check to ensure the new device address doesn't
conflict with any existing device.
Signed-off-by: Luyao Huang <lhuang@redhat.com>
VIR_APPEND_ELEMENT would clear @srv to NULL after it successfully
inserted it thus the reference count could not be increased afterwards.
Switch to VIR_APPEND_ELEMENT_COPY. This fixes crash after terminating
the daemon.
There was many errors in libvirt.log caused by prlsdkDelNet function because
job variable was always initialized as PRL_INVALID_HANDLE
In this patch job variable gets return value of PrlSrv_DeleteVirtualNetwork function()
https://bugzilla.redhat.com/show_bug.cgi?id=1220527
This type of information defines attributes of a system
baseboard. With one exception: board type is yet not implemented
in qemu so it's not introduced here either.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Some machine types are only reported as canonical names for other
machine types, which make it a bit harder to find what machine types are
supported by a specific QEMU binary. Ideally, one would just use
/capabilities/guest/arch[@name='...']/machine/text() XPath to get a list
of all supported machine types, but it doesn't work right now.
For example, we report
<machine canonical='pc-i440fx-2.3' maxCpus='255'>pc</machine>
in guest capabilities, but the corresponding
<machine maxCpus='255'>pc-i440fx-2.3</machine>
is missing.
This is a result of QMP probing. With "-machine ?" parsing QEMU sends
us two lines:
pc Standard PC (i440FX + PIIX, 1996) (alias of pc-i440fx-2.3)
pc-i440fx-2.3 Standard PC (i440FX + PIIX, 1996) (default)
while query-machines QMP command reports both in the same entry:
{"name": "pc-i440fx-2.3", "is-default": true, "cpu-max": 255, "alias": "pc"}
Let's make sure we always report separate <machine/> for both the
canonical name and its alias and using the canonical name as the default
machine type (i.e., inserting it before its alias) in case is-default is
true.
https://bugzilla.redhat.com/show_bug.cgi?id=1229666
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
As with several other attributes of devices (link status, sriov VF
list, IOMMU group list), the detdev feature bits aren't automatically
updated in the nodedev driver's cache when they change. In order to
get a properly up-to-date list when getting the XML of a device, we
must reget them in update-caps prior to each dumpxml.
This resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1232880
