When reverting a live internal snapshot with a live guest the ABI
compatiblity check was comparing a "migratable" definition with a normal
one. This resulted in the check failing with:
revert requires force: Target device address type none does not match source pci
This patch generates a "migratable" definition from the actual one to
check against the definition from the snapshot to avoid this problem.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1006886
Osier Yang pointed out that ever since commit 31cb030, the
signature of qemuDomainObjEndJob was changed to return a bool.
While comparison against 0 or > 0 still gives the right results,
it looks fishy; we also had one place that was comparing < 0
which is effectively dead code.
* src/qemu/qemu_migration.c (qemuMigrationPrepareAny): Fix dead
code bug.
(qemuMigrationBegin): Use more canonical form of bool check.
* src/qemu/qemu_driver.c (qemuAutostartDomain)
(qemuDomainCreateXML, qemuDomainSuspend, qemuDomainResume)
(qemuDomainShutdownFlags, qemuDomainReboot, qemuDomainReset)
(qemuDomainDestroyFlags, qemuDomainSetMemoryFlags)
(qemuDomainSetMemoryStatsPeriod, qemuDomainInjectNMI)
(qemuDomainSendKey, qemuDomainGetInfo, qemuDomainScreenshot)
(qemuDomainSetVcpusFlags, qemuDomainGetVcpusFlags)
(qemuDomainRestoreFlags, qemuDomainGetXMLDesc)
(qemuDomainCreateWithFlags, qemuDomainAttachDeviceFlags)
(qemuDomainUpdateDeviceFlags, qemuDomainDetachDeviceFlags)
(qemuDomainBlockResize, qemuDomainBlockStats)
(qemuDomainBlockStatsFlags, qemuDomainMemoryStats)
(qemuDomainMemoryPeek, qemuDomainGetBlockInfo)
(qemuDomainAbortJob, qemuDomainMigrateSetMaxDowntime)
(qemuDomainMigrateGetCompressionCache)
(qemuDomainMigrateSetCompressionCache)
(qemuDomainMigrateSetMaxSpeed)
(qemuDomainSnapshotCreateActiveInternal)
(qemuDomainRevertToSnapshot, qemuDomainSnapshotDelete)
(qemuDomainQemuMonitorCommand, qemuDomainQemuAttach)
(qemuDomainBlockJobImpl, qemuDomainBlockCopy)
(qemuDomainBlockCommit, qemuDomainOpenGraphics)
(qemuDomainGetBlockIoTune, qemuDomainGetDiskErrors)
(qemuDomainPMSuspendForDuration, qemuDomainPMWakeup)
(qemuDomainQemuAgentCommand, qemuDomainFSTrim): Likewise.
Signed-off-by: Eric Blake <eblake@redhat.com>
Failure to attach to a domain during 'virsh qemu-attach' left
the list of domains in an odd state:
$ virsh qemu-attach 4176
error: An error occurred, but the cause is unknown
$ virsh list --all
Id Name State
----------------------------------------------------
2 foo shut off
$ virsh qemu-attach 4176
error: Requested operation is not valid: domain is already active as 'foo'
$ virsh undefine foo
error: Failed to undefine domain foo
error: Requested operation is not valid: cannot undefine transient domain
$ virsh shutdown foo
error: Failed to shutdown domain foo
error: invalid argument: monitor must not be NULL
It all stems from leaving the list of domains unmodified on
the initial failure; we should follow the lead of createXML
which removes vm on failure (the actual initial failure still
needs to be fixed in a later patch, but at least this patch
gets us to the point where we aren't getting stuck with an
unremovable "shut off" transient domain).
While investigating, I also found a leak in qemuDomainCreateXML;
the two functions should behave similarly. Note that there are
still two unusual paths: if dom is not allocated, the user will
see an OOM error even though the vm remains registered (but oom
errors already indicate tricky cleanup); and if the vm starts
and then quits again all before the job ends, it is possible
to return a non-NULL dom even though the dom will no longer be
useful for anything (but this at least lets the user know their
short-lived vm ran).
* src/qemu/qemu_driver.c (qemuDomainCreateXML): Don't leak vm on
failure to obtain job.
(qemuDomainQemuAttach): Match cleanup of qemuDomainCreateXML.
Signed-off-by: Eric Blake <eblake@redhat.com>
Currently, only X86 provides users CPU features with CPUID instruction.
If users specify the features for non-x86, it should tell users to
remove them.
This patch is to report one error if features are specified by
users for non-x86 platform.
Signed-off-by: Li Zhang <zhlcindy@linux.vnet.ibm.com>
While debugging a failure of 'virsh qemu-attach', I noticed that
we were leaking the count of active domains on failure. This
means that a libvirtd session that is supposed to quit after
active domains disappear will hang around forever.
* src/qemu/qemu_process.c (qemuProcessAttach): Undo count of
active domains on failure.
Signed-off-by: Eric Blake <eblake@redhat.com>
In Fedora 19, 'qemu-kvm' is a simple wrapper that calls
'qemu-system-x86_64 -machine accel=kvm'. Attempting
to use 'virsh qemu-attach $pid' to a machine started as:
qemu-kvm -cdrom /var/lib/libvirt/images/foo.img \
-monitor unix:/tmp/demo,server,nowait -name foo \
--uuid cece4f9f-dff0-575d-0e8e-01fe380f12ea
was failing with:
error: XML error: No PCI buses available
because we did not see 'kvm' in the executable name read from
/proc/$pid/cmdline, and tried to assign os.machine as
"accel=kvm" instead of "pc"; this in turn led to refusal to
recognize the pci bus.
Noticed while investigating https://bugzilla.redhat.com/995312
although there are still other issues to fix before that bug
will be completely solved.
I've concluded that the existing parser code for native-to-xml
is a horrendous hodge-podge of ad-hoc approaches; I basically
rewrote the -machine section to be a bit saner.
* src/qemu/qemu_command.c (qemuParseCommandLine): Don't assume
-machine argument is always appropriate for os.machine; set
virtType if accel is present.
Signed-off-by: Eric Blake <eblake@redhat.com>
'virsh domxml-from-native' and 'virsh qemu-attach' could misbehave
for an emulator installed in (a somewhat unlikely) location
such as /usr/local/qemu-1.6/qemu-system-x86_64 or (an even less
likely) /opt/notxen/qemu-system-x86_64. Limit the strstr seach
to just the basename of the file where we are assuming details
about the binary based on its name.
While testing, I accidentally triggered a core dump during strcmp
when I forgot to set os.type on one of my code paths; this patch
changes such a coding error to raise a nicer internal error instead.
* src/qemu/qemu_command.c (qemuParseCommandLine): Compute basename
earlier.
* src/conf/domain_conf.c (virDomainDefPostParseInternal): Avoid
NULL deref.
Signed-off-by: Eric Blake <eblake@redhat.com>
CPU features are not supported on non-x86 and hasFeatures will be NULL.
This patch is to remove CPU features functions calling to avoid errors.
Signed-off-by: Li Zhang <zhlcindy@linux.vnet.ibm.com>
The VIR_FREE() macro will cast away any const-ness. This masked a
number of places where we passed a 'const char *' string to
VIR_FREE. Fortunately in all of these cases, the variable was not
in fact const data, but a heap allocated string. Fix all the
variable declarations to reflect this.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
No need to open code now that we have a nice function.
Interestingly, our virStringFreeList function is typed correctly
(a malloc'd list of malloc'd strings is NOT const, whether at the
point where it is created, or at the point where it is cleand up),
so using it with a 'const char **' argument would require a cast
to keep the compiler. I chose instead to remove const from code
even where we don't modify the argument, just to avoid the need
to cast.
* src/qemu/qemu_command.h (qemuParseCommandLine): Drop declaration.
* src/qemu/qemu_command.c (qemuParseProcFileStrings)
(qemuStringToArgvEnv): Don't force malloc'd result to be const.
(qemuParseCommandLinePid, qemuParseCommandLineString): Simplify
cleanup.
(qemuParseCommandLine, qemuFindEnv): Drop const-correctness to
avoid the need to cast in callers.
Signed-off-by: Eric Blake <eblake@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=999352
Since commit v1.0.5-56-g449e6b1 (Pull parsing of migration xml up into
QEMU driver APIs) any attempt to rename a domain during migration fails
with the following error message:
internal error Incoming cookie data had unexpected name DOM vs DOM2
This is because migration cookies always use the original domain name
and the mentioned commit failed to propagate the name back to
qemuMigrationPrepareAny.
Currently, kernel supports up to 8 queues for a multiqueue tap device.
However, if user tries to enter a huge number (e.g. one million) the tap
allocation fails, as expected. But what is not expected is the log full
of warnings:
warning : virFileClose:83 : Tried to close invalid fd 0
The problem is, upon error we iterate over an array of FDs (handlers to
queues) and VIR_FORCE_CLOSE() over each item. However, the array is
pre-filled with zeros. Hence, we repeatedly close stdin. Ouch.
But there's more. The queues allocation is done in virNetDevTapCreate()
which cleans up the FDs in case of error. Then, its caller, the
virNetDevTapCreateInBridgePort() iterates over the FD array and tries to
close them too. And so does qemuNetworkIfaceConnect() and
qemuBuildInterfaceCommandLine().
Starting with qemu 1.6, the qemu-system-arm vexpress-a9 model has a
hardcoded virtio-mmio transport which enables attaching all virtio
devices.
On the command line, we have to use virtio-XXX-device rather than
virtio-XXX-pci, thankfully s390 already set the precedent here so
it's fairly straight forward.
At the XML level, this adds a new device address type virtio-mmio.
The controller and addressing don't have any subelements at the
moment because we they aren't needed for this usecase, but could
be added later if needed.
Add a test case for an ARM guest with one of every virtio device
enabled.
Similar to the chardev bit, ARM boards depend on the old style '-net nic'
for actually instantiating net devices. But we can't block out
-netdev altogether since it's needed for upcoming virtio support.
And add tests for working ARM XML with console, disk, and networking.
This corresponds to '-sd' and '-drive if=sd' on the qemu command line.
Needed for many ARM boards which don't provide any other way to
pass in storage.
QEMU ARM boards don't give us any way to explicitly wire in
a -chardev, so use the old style -serial options.
Unfortunately this isn't as simple as just turning off the CHARDEV flag
for qemu-system-arm, as upcoming virtio support _will_ use device/chardev.
On my machine, a guest fails to boot if it has a sound card, but not
graphical device/display is configured, because pulseaudio fails to
initialize since it can't access $HOME.
A workaround is removing the audio device, however on ARM boards there
isn't any option to do that, so -nographic always fails.
Set QEMU_AUDIO_DRV=none if no <graphics> are configured. Unfortunately
this has massive test suite fallout.
Add a qemu.conf parameter nographics_allow_host_audio, that if enabled
will pass through QEMU_AUDIO_DRV from sysconfig (similar to
vnc_allow_host_audio)
Add an attribute named 'removable' to the 'target' element of disks,
which controls the removable flag. For instance, on a Linux guest it
controls the value of /sys/block/$dev/removable. This option is only
valid for USB disks (i.e. bus='usb'), and its default value is 'off',
which is the same behaviour as before.
To achieve this, 'removable=on' (or 'off') is appended to the '-device
usb-storage' parameter sent to qemu when adding a USB disk via
'-disk'. A capability flag QEMU_CAPS_USB_STORAGE_REMOVABLE was added
to keep track if this option is supported by the qemu version used.
Bug: https://bugzilla.redhat.com/show_bug.cgi?id=922495
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Allow use of the usb-storage device only if the new capability flag
QEMU_CAPS_DEVICE_USB_STORAGE is set, which it is for qemu(-kvm)
versions >= 0.12.1.2-rhel62-beta.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
vhost only works in KVM mode at the moment, and is infact compiled
out if the emulator is built for non-native architecture. While it
may work at some point in the future for plain qemu, for now it's
just noise on the command line (and which contributes to arm cli
breakage).
When using a <interface type="network"> that points to a network with
hostdev forwarding mode a hostdev alias is created for the network. This
allias is inserted into the hostdev list, but is backed with a part of
the network object that it is connected to.
When a VM is being stopped qemuProcessStop() calls
networkReleaseActualDevice() which eventually frees the memory for the
hostdev object. Afterwards when the domain definition is being freed by
virDomainDefFree() an invalid pointer is accessed by
virDomainHostdevDefFree() and may cause a crash of the daemon.
This patch removes the entry in the hostdev list before freeing the
depending memory to avoid this issue.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1000973
QEMU commit 3984890 introduced the "pci-hole64-size" property,
to i440FX-pcihost and q35-pcihost with a default setting of 2 GB.
Translate <pcihole64>x<pcihole64/> to:
-global q35-pcihost.pci-hole64-size=x for q35 machines and
-global i440FX-pcihost.pci-hole64-size=x for i440FX-based machines.
Error out on other machine types or if the size was specified
but the pcihost device lacks 'pci-hole64-size' property.
https://bugzilla.redhat.com/show_bug.cgi?id=990418
The ftp protocol is already recognized by qemu/KVM so add this support to
libvirt as well.
The xml should be as following:
<disk type='network' device='cdrom'>
<source protocol='ftp' name='/url/path'>
<host name='host.name' port='21'/>
</source>
</disk>
Signed-off-by: Aline Manera <alinefm@br.ibm.com>
QEMU/KVM already allows a HTTP URL for the cdrom ISO image so add this support
to libvirt as well.
The xml should be as following:
<disk type='network' device='cdrom'>
<source protocol='http' name='/url/path'>
<host name='host.name' port='80'/>
</source>
</disk>
Signed-off-by: Aline Manera <alinefm@br.ibm.com>
If there's no hard_limit set and domain uses VFIO we still must lock
the guest memory (prerequisite from qemu). Hence, we should compute
the amount to be locked from max_balloon.
When cpu hotplug fails without reporting an error, we would fail the
command but update the count of vCPUs anyways.
Commit 761fc48136 fixed the case when CPU
hot-unplug failed silently, but forgot to fix up the value in this case.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1000357
The virDomainOpenGraphics method accepts a UNIX socket FD from
the client app. It must set the label on this FD otherwise QEMU
will be prevented from receiving it with recvmsg.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
If user requested multiqueue networking, beside multiple /dev/tap and
/dev/vhost-net openings, we forgot to pass mq=on onto the -device
virtio-net-pci command line. This is advised at:
http://www.linux-kvm.org/page/Multiqueue#Enable_MQ_feature
Re-arrange the code so that the returned bitmap is always initialized to
NULL even on early failures and return an error message as some callers
are already expecting it. Fix up the rest not to shadow the error.
https://bugzilla.redhat.com/show_bug.cgi?id=822052
When doing a live migration, if the destination fails for any
reason after the point in which files should be labeled, then
the cleanup of the destination would restore the labels to their
defaults, even though the source is still trying to continue
running with the image open. Bug 822052 mentioned one source
of live migration failure - a mismatch in SELinux virt_use_nfs
settings (on for source, off for destination); but I found other
situations that would also trigger it (for example, having a
graphics device tied to port 5999 on the source, and a different
domain on the destination already using that port, so that the
destination cannot reuse the port).
In short, just as cleanup of the source on a successful migration
must not relabel files (because the destination would be crippled
by the relabel), cleanup of the destination on a failed migration
must not relabel files (because the source would be crippled).
* src/qemu/qemu_process.c (qemuProcessStart): Set flag to avoid
label restoration when cleaning up on failed migration.
Signed-off-by: Eric Blake <eblake@redhat.com>
Each of the modules handled reporting error messages from the secret fetching
slightly differently with respect to the error. Provide a similar message
for each error case and provide as much data as possible.
Following XML would fail :
<disk type='network' device='lun'>
<driver name='qemu' type='raw'/>
<source protocol='iscsi' name='iqn.2013-07.com.example:iscsi/1'>
<host name='example.com' port='3260'/>
</source>
<target dev='sda' bus='scsi'/>
</disk>
With the message:
error: Failed to start domain iscsilun
error: Unable to get device ID 'iqn.2013-07.com.example:iscsi/1': No such fi
Cause was commit id '1f49b05a' which added 'virDomainDiskSourceIsBlockType'
If there's no hard_limit set and domain uses VFIO we still must lock the
guest memory (prerequisite from qemu). Hence, we should compute the
amount to be locked from max_balloon.
Since 16bcb3 we have a regression. The hard_limit is set
unconditionally. By default the limit is zero. Hence, if user hasn't
configured any, we set the zero in cgroup subsystem making the kernel
kill the corresponding qemu process immediately. The proper fix is to
set hard_limit iff user has configured any.
This function is to guess the correct limit for maximal memory
usage by qemu for given domain. This can never be guessed
correctly, not to mention all the pains and sleepless nights this
code has caused. Once somebody discovers algorithm to solve the
Halting Problem, we can compute the limit algorithmically. But
till then, this code should never see the light of the release
again.
Currently the virConnectBaselineCPU API does not expose the CPU features
that are part of the CPU's model. This patch adds a new flag,
VIR_CONNECT_BASELINE_CPU_EXPAND_FEATURES, that causes the API to explicitly
list all features that are part of that model.
Signed-off-by: Don Dugger <donald.d.dugger@intel.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
Hotplugging a single SCSI device works, but adding additional ones
result in an error from QEMU:
[root@gpok197 ~]# virsh attach-device guest01 blah.xml
Device attached successfully
[root@gpok197 ~]# virsh attach-device guest01 blah2.xml
error: Failed to attach device from blah2.xml
error: internal error unable to execute QEMU command 'device_add': Duplicate ID 'hostdev0' for device
The hostdev ID that is created is always set to zero, regardless
of the contents of the XML. Changing the index in the hotplug case
to a negative one so the next available index is used.
Signed-off-by: Eric Farman <farman@linux.vnet.ibm.com>
Reviewed-by: Viktor Mihajlovski <mihajlov@linux.vnet.ibm.com>
Go through disks of guest, if one disk doesn't exist or its backing
chain is broken, with 'optional' startupPolicy, for CDROM and Floppy
we only discard its source path definition in xml, for disks we drop
it from disk list and free it.
This patch addresses two concerns with the error reporting when an
incompatible PCI address is specified for a device:
1) It wasn't always apparent which device had the problem. With this
patch applied, any error about an incompatible address will always
contain the full address as given in the config, so it will be easier
to determine which device's config aused the problem.
2) In some cases when the problem came from bad config, the error
message was erroneously classified as VIR_ERR_INTERNAL_ERROR. With
this patch applied, the same error message will be changed to indicate
either "internal" or "xml" error depending on whether the address came
from the config, or was automatically generated by libvirt.
Note that in the case of "internal" (due to bad auto-generation)
errors, the PCI address won't be of much use in finding the location
in config to change (because it was automatically generated). Of
course that makes perfect sense, but still the address could provide a
clue about a bug in libvirt attempting to use a type of pci bus that
doesn't have its flags set correctly (or something similar). In other
words, it's not perfect, but it is definitely better.
q35 machines have an implicit ahci (sata) controller at 00:1F.2 which
has no "id" associated with it. For this reason, we can't refer to it
as "ahci0". Instead, we don't give an id on the commandline, which
qemu interprets as "use the first ahci controller". We then need to
specify the unit with "unit=%d" rather than adding it onto the bus
arg.
https://bugzilla.redhat.com/show_bug.cgi?id=979477
Since 1.0.3 we are using the new way to copy non shared storage during
migration (the NBD way). However, whether the new or old way is used is
not controllable by user but unconditionally turned on if both sides of
migration support it. Moreover, the implementation is not complete: the
combination for VIR_MIGRATE_TUNNELLED flag is missing (as we need to
open new port on the destination) in which case we just error out. This
is a deadly combination: not letting users choose their destiny and
erroring out. We should not do that but VIR_WARN and turn the NBD off
instead.
We had been setting the device alias in the devinceinfo for pci
controllers to "pci%u", but then hardcoding "pci.%u" when creating the
device address for other devices using that pci bus. This all worked
just fine until we encountered the built-in "pcie.0" bus (the PCIe
root complex) in Q35 machines.
In order to create the correct commandline for this one case, this
patch:
1) sets the alias for PCI controllers correctly, to "pci.%u" (or
"pcie.%u" for the pcie-root controller)
2) eliminates the hardcoded "pci.%u" for pci controllers when
generatuing device address strings, and instead uses the controller's
alias.
3) plumbs a pointer to the virDomainDef all the way down to
qemuBuildDeviceAddressStr. This was necessary in order to make the
aliase of the controller *used by a device* available (previously
qemuBuildDeviceAddressStr only had the deviceinfo of the device
itself, *not* of the controller it was connecting to). This made for a
larger than desired diff, but at least in the future we won't have to
do it again, since all the information we could possibly ever need for
future enhancements is in the virDomainDef. (right?)
This should be done for *all* controllers, but for now we just do it
in the case of PCI controllers, to reduce the likelyhood of
regression.
This patch adds in special handling for a few devices that need to be
treated differently for q35 domains:
usb - there is no implicit/default usb controller for the q35
machinetype. This is done because normally the default usb controller
is added to a domain by just adding "-usb" to the qemu commandline,
and it's assumed that this will add a single piix3 usb1 controller at
slot 1 function 2. That's not what happens when the machinetype is
q35, though. Instead, adding -usb to the commandline adds 3 usb
(version 2) controllers to the domain at slot 0x1D.{1,2,7}. Rather
than having
<controller type='usb' index='0'/>
translate into 3 separate devices on the PCI bus, it's cleaner to not
automatically add a default usb device; one can always be added
explicitly if desired. Or we may decide that on q35 machines, 3 usb
controllers will be automatically added when none is given. But for
this initial commit, at least we aren't locking ourselves into
something we later won't want.
video - qemu always initializes the primary video device immediately
after any integrated devices for the machinetype. Unless instructed
otherwise (by using "-device vga..." instead of "-vga" which libvirt
uses in many cases to work around deficiencies and bugs in various
qemu versions) qemu will always pick the first unused slot. In the
case of the "pc" machinetype and its derivatives, this is always slot
2, but on q35 machinetypes, the first free slot is slot 1 (since the
q35's integrated peripheral devices are placed in other slots,
e.g. slot 0x1f). In order to make the PCI address of the video device
predictable, that slot (1 or 2, depending on machinetype) is reserved
even when no video device has been specified.
sata - a q35 machine always has a sata controller implicitly added at
slot 0x1F, function 2. There is no way to avoid this controller, so we
always add it. Note that the xml2xml tests for the pcie-root and q35
cases were changed to use DO_TEST_DIFFERENT() so that we can check for
the sata controller being automatically added. This is especially
important because we can't check for it in the xml2argv output (it has
no effect on that output since it's an implicit device).
ide - q35 has no ide controllers.
isa and smbus controllers - these two are always present in a q35 (at
slot 0x1F functions 0 and 3) but we have no way of modelling them in
our config. We do need to reserve those functions so that the user
doesn't attempt to put anything else there though. (note that the "pc"
machine type also has an ISA controller, which we also ignore).
This PCI controller, named "dmi-to-pci-bridge" in the libvirt config,
and implemented with qemu's "i82801b11-bridge" device, connects to a
PCI Express slot (e.g. one of the slots provided by the pcie-root
controller, aka "pcie.0" on the qemu commandline), and provides 31
*non-hot-pluggable* PCI (*not* PCIe) slots, numbered 1-31.
Any time a machine is defined which has a pcie-root controller
(i.e. any q35-based machinetype), libvirt will automatically add a
dmi-to-pci-bridge controller if one doesn't exist, and also add a
pci-bridge controller. The reasoning here is that any useful domain
will have either an immediate (startup time) or eventual (subsequent
hot-plug) need for a standard PCI slot; since the pcie-root controller
only provides PCIe slots, we need to connect a dmi-to-pci-bridge
controller to it in order to get a non-hot-plug PCI slot that we can
then use to connect a pci-bridge - the slots provided by the
pci-bridge will be both standard PCI and hot-pluggable.
Since pci-bridge devices themselves can not be hot-plugged into a
running system (although you can hot-plug other devices into a
pci-bridge's slots), any new pci-bridge controller that is added can
(and will) be plugged into the dmi-to-pci-bridge as long as it has
empty slots available.
This patch is also changing the qemuxml2xml-pcie test from a "DO_TEST"
to a "DO_DIFFERENT_TEST". This is so that the "before" xml can omit
the automatically added dmi-to-pci-bridge and pci-bridge devices, and
the "after" xml can include it - this way we are testing if libvirt is
properly adding these devices.
This controller is implicit on q35 machinetypes. It provides 31 PCIe
(*not* PCI) slots as controller 0.
Currently there are no devices that can connect to pcie-root, and no
implicit pci controller on a q35 machine, so q35 is still
unusable. For a usable q35 system, we need to add a
"dmi-to-pci-bridge" pci controller, which can connect to pcie-root,
and provides standard pci slots that can be used to connect other
devices.
Previous refactoring of the guest PCI address reservation/allocation
code allowed for slot types other than basic PCI (e.g. PCI express,
non-hotpluggable slots, etc) but would not auto-allocate a slot for a
device that required any type other than a basic hot-pluggable
PCI slot.
This patch refactors the code to be aware of different slot types
during auto-allocation of addresses as well - as long as there is an
empty slot of the required type, it will be found and used.
The piece that *wasn't* added is that we don't auto-create a new PCI
bus when needed for anything except basic PCI devices. This is because
there are multiple different types of controllers that can provide,
for example, a PCI express slot (in addition to the pcie-root
controller, these can also be found on a "root-port" or on a
"downstream-switch-port"). Since we currently don't support any PCIe
devices (except pending support for dmi-to-pci-bridge), we can defer
any decision on what to do about this.
* The functions qemuDomainPCIAddressReserveAddr and
qemuDomainPCIAddressReserveSlot were very similar (and should have
been more similar) and were about to get more code added to them which
would create even more duplicated code, so this patch gives
qemuDomainPCIAddressReserveAddr a "reserveEntireSlot" arg, then
replaces the body of qemuDomainPCIAddressReserveSlot with a call to
qemuDomainPCIAddressReserveAddr.
You will notice that addrs->lastaddr was previously set in
qemuDomainPCIAddressReserveAddr (but *not* set in
qemuDomainPCIAddressReserveSlot). For consistency and cleanliness of
code, that bit was removed and put into the one caller of
qemuDomainPCIAddressReserveAddr (there is a similar place where the
caller of qemuDomainPCIAddressReserveSlot sets lastaddr). This does
guarantee identical functionality to pre-patch code, but in practice
isn't really critical, because lastaddr is just keeping track of where
to start when looking for a free slot - if it isn't updated, we will
just start looking on a slot that's already occupied, then skip up to
one that isn't.
* qemuCollectPCIAddress was essentially doing the same thing as
qemuDomainPCIAddressReserveAddr, but with some extra special case
checking at the beginning. The duplicate code has been replaced with
a call to qemuDomainPCIAddressReserveAddr. This required adding a
"fromConfig" boolean, which is only used to change the log error
code from VIR_ERR_INTERNAL_ERROR (when the address was
auto-generated by libvirt) to VIR_ERR_XML_ERROR (when the address is
coming from the config); without this differentiation, it would be
difficult to tell if an error was caused by something wrong in
libvirt's auto-allocate code or just bad config.
* the bit of code in qemuDomainPCIAddressValidate that checks the
connect type flags is going to be used in a couple more places where
we don't need to also check the slot limits (because we're generating
the slot number ourselves), so that has been pulled out into a
separate qemuDomainPCIAddressFlagsCompatible function.
* qemuDomainPCIAddressSetNextAddr
The name of this function was confusing because 1) other functions in
the file that end in "Addr" are only operating on a single function of
one PCI slot, not the entire slot, while functions that do something
with the entire slot end in "Slot", and 2) it didn't contain a verb
describing what it is doing (the "Set" refers to the set that contains
all PCI buses in the system, used to keep track of which slots in
which buses are already reserved for use).
It is now renamed to qemuDomainPCIAddressReserveNextSlot, which more
clearly describes what it is doing. Arguably, it could have been
changed to qemuDomainPCIAddressSetReserveNextSlot, but 1) the word
"set" is confusing in this context because it could be intended as a
verb or as a noun, and 2) most other functions that operate on a
single slot or address within this set are also named
qemuDomainPCIAddress... rather than qemuDomainPCIAddressSet... Only
the Create, Free, and Grow functions for an address set (which modify the
entire set, not just one element) use "Set" in their name.
* qemuPCIAddressAsString, qemuPCIAddressValidate
All the other functions in this set are named
qemuDomainPCIAddressxxxxx, so I renamed these to be consistent.
The parser shouldn't be doing arch-specific things like adding in
implicit controllers to the config. This should instead be done in the
hypervisor's post-parse callback.
This patch removes the auto-add of a usb controller from the domain
parser, and puts it into the qemu driver's post-parse callback (just
as is already done with the auto-add of the pci-root controller). In
the future, any machine/arch that shouldn't have a default usb
controller added should just set addDefaultUSB = false in this
function.
We've recently seen that q35 and ARMV7L domains shouldn't get a default USB
controller, so I've set addDefaultUSB to false for both of those.
If upgrading from a libvirt that is older than 1.0.5, we can
not assume that vm->def->resource is non-NULL. This bogus
assumption caused libvirtd to crash
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
*src/util/virstoragefile.c: Add a helper function to get
the first name of missing backing files, if the name is NULL,
it means the diskchain is not broken.
*src/qemu/qemu_domain.c: qemuDiskChainCheckBroken(disk) to
check if its chain is broken
Refactor this function to make it focus on disk presence checking,
including diskchain checking, and not only for CDROM and Floppy.
This change is good for the following patches.
Make the virCgroupNewMachine method try to use systemd-machined
first. If that fails, then fallback to using the traditional
cgroup setup code path.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Although this isn't apparently needed for the guest agent itself, the
test I will be adding later depends on the newline as a separator of
messages to process.
The VIR_DOMAIN_PAUSED_GUEST_PANICKED constant is badly named,
leaking the QEMU event name. Elsewhere in the API we use
'CRASHED' rather than 'PANICKED', and the addition of 'GUEST'
is redundant since all events are guest related.
Thus rename it to VIR_DOMAIN_PAUSED_CRASHED, which matches
with VIR_DOMAIN_RUNNING_CRASHED and VIR_DOMAIN_EVENT_CRASHED.
It was added in commit 14e7e0ae8d
which post-dates v1.1.0, so is safe to rename before 1.1.1
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=981094
The commit 0ad9025ef introduce qemu flag QEMU_CAPS_DEVICE_VIDEO_PRIMARY
for using -device VGA, -device cirrus-vga, -device vmware-svga and
-device qxl-vga. In use, for -device qxl-vga, mouse doesn't display
in guest window like the desciption in above bug.
This patch try to use -device for primary video when qemu >=1.6 which
contains the bug fix patch
Instead of requiring drivers to use a combination of calls
to virCgroupNewDetect and virCgroupIsValidMachine, combine
the two into virCgroupNewDetectMachine
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Both virStoragePoolFree and virStorageVolFree reset the last error,
which might lead to the cryptic message:
An error occurred, but the cause is unknown
When the volume wasn't found, virStorageVolFree was called with NULL,
leading to an error:
invalid storage volume pointer in virStorageVolFree
This patch changes it to:
Storage volume not found: no storage vol with matching name 'tomato'
Function qemuOpenFile() haven't had any idea about seclabels applied
to VMs only, so in case the seclabel differed from the "user:group"
from configuration, there might have been issues with opening files.
Make qemuOpenFile() VM-aware, but only optionally, passing NULL
argument means skipping VM seclabel info completely.
However, all current qemuOpenFile() calls look like they should use VM
seclabel info in case there is any, so convert these calls as well.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=869053
Since PCI bridges, PCIe bridges, PCIe switches, and PCIe root ports
all share the same namespace, they are all defined as controllers of
type='pci' in libvirt (but with a differing model attribute). Each of
these controllers has a certain connection type upstream, allows
certain connection types downstream, and each can either allow a
single downstream connection at slot 0, or connections from slot 1 -
31.
Right now, we only support the pci-root and pci-bridge devices, both
of which only allow PCI devices to connect, and both which have usable
slots 1 - 31. In preparation for adding other types of controllers
that have different capabilities, this patch 1) adds info to the
qemuDomainPCIAddressBus object to indicate the capabilities, 2) sets
those capabilities appropriately for pci-root and pci-bridge devices,
and 3) validates that the controller being connected to is the proper
type when allocating slots or validating that a user-selected slot is
appropriate for a device..
Having this infrastructure in place will make it much easier to add
support for the other PCI controller types.
While it would be possible to do all the necessary checking by just
storing the controller model in the qemyuDomainPCIAddressBus, it
greatly simplifies all the validation code to also keep a "flags",
"minSlot" and "maxSlot" for each - that way we can just check those
attributes rather than requiring a nearly identical switch statement
everywhere we need to validate compatibility.
You may notice many places where the flags are seemingly hard-coded to
QEMU_PCI_CONNECT_HOTPLUGGABLE | QEMU_PCI_CONNECT_TYPE_PCI
This is currently the correct value for all PCI devices, and in the
future will be the default, with small bits of code added to change to
the flags for the few devices which are the exceptions to this rule.
Finally, there are a few places with "FIXME" comments. Note that these
aren't indicating places that are broken according to the currently
supported devices, they are places that will need fixing when support
for new PCI controller models is added.
To assure that there was no regression in the auto-allocation of PCI
addresses or auto-creation of integrated pci-root, ide, and usb
controllers, a new test case (pci-bridge-many-disks) has been added to
both the qemuxml2argv and qemuxml2xml tests. This new test defines a
domain with several dozen virtio disks but no pci-root or
pci-bridges. The .args file of the new test case was created using
libvirt sources from before this patch, and the test still passes
after this patch has been applied.
Although these two enums are named ..._LAST, they really had the value
of ..._SIZE. This patch changes their values so that, e.g.,
QEMU_PCI_ADDRESS_SLOT_LAST really is the slot number of the last slot
on a PCI bus.
The implicit IDE, USB, and video controllers provided by the PIIX3
chipset in the pc-* machinetypes are not present on other
machinetypes, so we shouldn't be doing the special checking for
them. This patch places those validation checks into a separate
function that is only called for machine types that have a PIIX3 chip
(which happens to be the i440fx-based pc-* machine types).
One qemuxml2argv test data file had to be changed - the
pseries-usb-multi test had included a piix3-usb-uhci device, which was
being placed at a specific address, and also had slot 2 auto reserved
for a video device, but the pseries virtual machine doesn't actually
have a PIIX3 chip, so even if there was a piix3-usb-uhci driver for
it, the device wouldn't need to reside at slot 1 function 2. I just
changed the .argv file to have the generic slot info for the two
devices that results when the special PIIX3 code isn't executed.
qemuDomainPCIAddressBus was an array of QEMU_PCI_ADDRESS_SLOT_LAST
uint8_t's, which worked fine as long as every PCI bus was
identical. In the future, some PCI busses will allow connecting PCI
devices, and some will allow PCIe devices; also some will only allow
connection of a single device, while others will allow connecting 31
devices.
In order to keep track of that information for each bus, we need to
turn qemuDomainPCIAddressBus into a struct, for now with just one
member:
uint8_t slots[QEMU_PCI_ADDRESS_SLOT_LAST];
Additional members will come in later patches.
The item in qemuDomainPCIAddresSet that contains the array of
qemuDomainPCIAddressBus is now called "buses" to be more consistent
with the already existing "nbuses" (and with the new "slots" array).
Currently the QEMU driver creates the VM's cgroup prior to
forking, and then uses a virCommand hook to move the child
into the cgroup. This won't work with systemd whose APIs
do the creation of cgroups + attachment of processes atomically.
Fortunately we have a handshake taking place between the
QEMU driver and the child process prior to QEMU being exec()d,
which was introduced to allow setup of disk locking. By good
fortune this synchronization point can be used to enable the
QEMU driver to do atomic setup of cgroups removing the use
of the hook script.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Use the new virCgroupNewDetect function to determine cgroup
placement of existing running VMs. This will allow the legacy
cgroups creation APIs to be removed entirely
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
During qemuTranslateDiskSourcePool() execution, if the srcpool has been
defined with authentication information, then for iSCSI pools copy the
authentication and host information to virDomainDiskDef.
Due to a goto statement missed when refactoring in 2771f8b74c
when acquiring of a domain job failed the error path was not taken. This
resulted into a crash afterwards as an extra reference was removed from a
domain object leading to it being freed. An attempt to list the domains
leaded to a crash of the daemon afterwards.
https://bugzilla.redhat.com/show_bug.cgi?id=928672
The translation must be done before both of cgroup and security
setting, otherwise since the disk source is not translated yet,
it might be skipped on cgroup and security setting.
The difference with already supported pool types (dir, fs, block)
is: there are two modes for iscsi pool (or network pools in future),
one can specify it either to use the volume target path (the path
showed up on host) with mode='host', or to use the remote URI qemu
supports (e.g. file=iscsi://example.org:6000/iqn.1992-01.com.example/1)
with mode='direct'.
For 'host' mode, it copies the volume target path into disk->src. For
'direct' mode, the corresponding info in the *one* pool source host def
is copied to disk->hosts[0].
Convert the remaining methods in vircgroup.c to report errors
instead of returning errno values.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The alias for hostdevs of type SCSI can be too long for QEMU if
larger LUNs are encountered. Here's a real life example:
<hostdev mode='subsystem' type='scsi' managed='no'>
<source>
<adapter name='scsi_host0'/>
<address bus='0' target='19' unit='1088634913'/>
</source>
<address type='drive' controller='0' bus='0' target='0' unit='0'/>
</hostdev>
this results in a too long drive id, resulting in QEMU yelling
Property 'scsi-generic.drive' can't find value 'drive-hostdev-scsi_host0-0-19-1088634913'
This commit changes the alias back to the default hostdev$(index)
scheme.
Signed-off-by: Viktor Mihajlovski <mihajlov@linux.vnet.ibm.com>
In case libvirtd is asked to unplug a device but the device is actually
unplugged later when libvirtd is not running, we need to detect that and
remove such device when libvirtd starts again and reconnects to running
domains.
A future patch wants the DAC security manager to be able to safely
get the supplemental group list for a given uid, but at the time
of a fork rather than during initialization so as to pick up on
live changes to the system's group database. This patch adds the
framework, including the possibility of a pre-fork callback
failing.
For now, any driver that implements a prefork callback must be
robust against the possibility of being part of a security stack
where a later element in the chain fails prefork. This means
that drivers cannot do any action that requires a call to postfork
for proper cleanup (no grabbing a mutex, for example). If this
is too prohibitive in the future, we would have to switch to a
transactioning sequence, where each driver has (up to) 3 callbacks:
PreForkPrepare, PreForkCommit, and PreForkAbort, to either clean
up or commit changes made during prepare.
* src/security/security_driver.h (virSecurityDriverPreFork): New
callback.
* src/security/security_manager.h (virSecurityManagerPreFork):
Change signature.
* src/security/security_manager.c (virSecurityManagerPreFork):
Optionally call into driver, and allow returning failure.
* src/security/security_stack.c (virSecurityDriverStack):
Wrap the handler for the stack driver.
* src/qemu/qemu_process.c (qemuProcessStart): Adjust caller.
Signed-off-by: Eric Blake <eblake@redhat.com>
When user does not specify any model for scsi controller, or worse, no
controller at all, but libvirt automatically adds scsi controller with
no model, we are not searching for virtio-scsi and thus this can fail
for example on qemu which doesn't support lsi logic adapter.
This means that when qemu on x86 doesn't support lsi53c895a and the
user adds the following to an XML without any scsi controller:
<disk ...>
...
<target dev='sda'>
</disk>
libvirt fails like this:
# virsh define asdf.xml
error: Failed to define domain from asdf.xml
error: internal error Unable to determine model for scsi controller
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=974943
When virAsprintf was changed from a function to a macro
reporting OOM error in dc6f2da, it was documented as returning
0 on success. This is incorrect, it returns the number of bytes
written as asprintf does.
Some of the functions were converted to use virAsprintf's return
value directly, changing the return value on success from 0 to >= 0.
For most of these, this is not a problem, but the change in
virPCIDriverDir breaks PCI passthrough.
The return value check in virhashtest pre-dates virAsprintf OOM
conversion.
vmwareMakePath seems to be unused.
Merge the virCommandPreserveFD / virCommandTransferFD methods
into a single virCommandPasFD method, and use a new
VIR_COMMAND_PASS_FD_CLOSE_PARENT to indicate their difference
in behaviour
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Introduced in commit 24b08219; compilation on RHEL 6.4 complained:
qemu/qemu_hotplug.c: In function 'qemuDomainAttachChrDevice':
qemu/qemu_hotplug.c:1257: error: declaration of 'remove' shadows a global declaration [-Wshadow]
/usr/include/stdio.h:177: error: shadowed declaration is here [-Wshadow]
* src/qemu/qemu_hotplug.c (qemuDomainAttachChrDevice): Avoid the
name 'remove'.
Signed-off-by: Eric Blake <eblake@redhat.com>
A part of the returned monitor response was freed twice and caused
crashes of the daemon when using guest agent cpu count retrieval.
# virsh vcpucount dom --guest
Introduced in v1.0.6-48-gc6afcb0
Implement the new API that will handle setting the balloon driver statistics
collection period in order to enable or disable the collection dynamically.
This patch will add the qemuMonitorJSONGetMemoryStats() to execute a
"guest-stats" on the balloonpath using "get-qom" replacing the former
mechanism which looked through the "query-ballon" returned data for
the fields. The "query-balloon" code only returns 'actual' memory.
Rather than duplicating the existing code, have the JSON API use the
GetBalloonInfo API.
A check in the qemuMonitorGetMemoryStats() will be made to ensure the
balloon driver path has been set. Since the underlying JSON code can
return data not associated with the balloon driver, we don't fail on
a failure to get the balloonpath. Of course since we've made the check,
we can then set the ballooninit flag. Getting the path here is primarily
due to the process reconnect path which doesn't attempt to set the
collection period.
At vm startup and attach attempt to set the balloon driver statistics
collection period based on the value found in the domain xml file. This
is not done at reconnect since it's possible that a collection period
was set on the live guest and making the set period call would reset to
whatever value is stored in the config file.
Setting the stats collection period has a side effect of searching through
the qom-list output for the virtio balloon driver and making sure that it
has the right properties in order to allow setting of a collection period
and eventually fetching of statistics.
The walk through the qom-list is expensive and thus the balloonpath will
be saved in the monitor private structure as well as a flag indicating
that the initialization has already been attempted (in the event that a
path is not found, no sense to keep checking).
This processing model conforms to the qom object model model which
requires setting object properties after device startup. That is, it's
not possible to pass the period along via the startup code as it won't
be recognized.
If users haven't configured guest agent then qemuAgentCommand() will
dereference a NULL 'mon' pointer, which causes crash of libvirtd when
using agent based cpu (un)plug.
With the patch, when the qemu-ga service isn't running in the guest,
a expected error "error: Guest agent is not responding: Guest agent
not available for now" will be raised, and the error "error: argument
unsupported: QEMU guest agent is not configured" is raised when the
guest hasn't configured guest agent.
GDB backtrace:
(gdb) bt
#0 virNetServerFatalSignal (sig=11, siginfo=<value optimized out>, context=<value optimized out>) at rpc/virnetserver.c:326
#1 <signal handler called>
#2 qemuAgentCommand (mon=0x0, cmd=0x7f39300017b0, reply=0x7f394b090910, seconds=-2) at qemu/qemu_agent.c:975
#3 0x00007f39429507f6 in qemuAgentGetVCPUs (mon=0x0, info=0x7f394b0909b8) at qemu/qemu_agent.c:1475
#4 0x00007f39429d9857 in qemuDomainGetVcpusFlags (dom=<value optimized out>, flags=9) at qemu/qemu_driver.c:4849
#5 0x00007f3957dffd8d in virDomainGetVcpusFlags (domain=0x7f39300009c0, flags=8) at libvirt.c:9843
How to reproduce?
# To start a guest without guest agent configuration
# then run the following cmdline
# virsh vcpucount foobar --guest
error: End of file while reading data: Input/output error
error: One or more references were leaked after disconnect from the hypervisor
error: Failed to reconnect to the hypervisor
RHBZ: https://bugzilla.redhat.com/show_bug.cgi?id=984821
Signed-off-by: Alex Jia <ajia@redhat.com>
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
There are two levels on which a device may be hotplugged: config
and live. The config level requires just an insert or remove from
internal domain definition structure, which is exactly what this
patch does. There is currently no implementation for a chardev
update action, as there's not much to be updated. But more
importantly, the only thing that can be updated is path or socket
address by which chardevs are distinguished. So the update action
is currently not supported.
Add a new qemuMonitorJSONSetObjectProperty() method to support invocation
of the 'qom-set' JSON monitor command with a provided path, property, and
expected data type to set.
NOTE: The set API was added only for the purpose of the qemumonitorjsontest
The test code uses the same "/machine/i440fx" property as the get test and
attempts to set the "realized" property to "true" (which it should be set
at anyway).
Add a new qemuMonitorJSONGetObjectProperty() method to support invocation
of the 'qom-get' JSON monitor command with a provided path, property, and
expected data type return. The qemuMonitorJSONObjectProperty is similar to
virTypedParameter; however, a future patch will extend it a bit to include
a void pointer to balloon driver statistic data.
NOTE: The ObjectProperty structures and API are added only for the
purpose of the qemumonitorjsontest
The provided test will execute a qom-get on "/machine/i440fx" which will
return a property "realized".
Add a new qemuMonitorJSONGetObjectListPaths() method to support invocation
of the 'qom-list' JSON monitor command with a provided path.
NOTE: The ListPath structures and API's are added only for the
purpose of the qemumonitorjsontest
The returned list of paired data fields of "name" and "type" that can
be used to peruse QOM configuration data and eventually utilize for the
balloon statistics.
The test does a "{"execute":"qom-list", "arguments": { "path": "/"}}" which
returns "{"return": [{"name": "machine", "type": "child<container>"},
{"name": "type", "type": "string"}]}" resulting in a return of an array
of 2 elements with [0].name="machine", [0].type="child<container>". The [1]
entry appears to be a header that could be used some day via a command such
as "virsh qemuobject --list" to format output.
If an error occurs during qemuDomainAttachNetDevice after the macvtap
was created in qemuPhysIfaceConnect, the macvtap device gets left behind.
This patch adds code to the cleanup routine to delete the macvtap.
Signed-off-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Reviewed-by: Viktor Mihajlovski <mihajlov@linux.vnet.ibm.com>
I recently patches the callers to virPCIDeviceReset() to not call it
if the current driver for a device was vfio-pci (since that driver
will always reset the device itself when appropriate. At the time, Dan
Berrange suggested that I could instead modify virPCIDeviceReset
to check the currently bound driver for the device, and decide
for itself whether or not to go ahead with the reset.
This patch removes the previously added checks, and replaces them with
a check down in virPCIDeviceReset(), as suggested.
The functional difference here is that previously we were deciding
based on either the hostdev configuration or the value of
stubDriverName in the virPCIDevice object, but now we are actually
comparing to the "driver" link in the device's sysfs entry
directly. In practice, both should be the same.
The function being introduced is responsible for creating command
line argument for '-device' for given character device. Based on
the chardev type, it calls appropriate qemuBuild.*ChrDeviceStr(),
e.g. qemuBuildSerialChrDeviceStr() for serial chardev and so on.
The chardev alias assignment is going to be needed in a separate
places, so it should be moved into a separate function rather
than copying code randomly around.
The function being introduced is responsible for preparing and
executing 'chardev-add' qemu monitor command. Moreover, in case
of PTY chardev, the corresponding pty path is updated.
Currently, we are building InetSocketAddress qemu json type
within the qemuMonitorJSONNBDServerStart function. However, other
future functions may profit from the code as well. So it should
be moved into a static function.
Recent changes uncovered a possibility that 'last_processed_hostdev_vf'
was set to -1 in 'qemuPrepareHostdevPCIDevices' and would cause problems
in for loop end condition in the 'resetvfnetconfig' label if the
variable was never set to 'i' due to 'qemuDomainHostdevNetConfigReplace'
failure.
With current code, error reporting for unsupported devices for hot plug,
unplug and update is total mess. The VIR_ERR_CONFIG_UNSUPPORTED error
code is reported instead of VIR_ERR_OPERATION_UNSUPPORTED. Moreover, the
error messages are not helping to find the root cause (lack of
implementation).
For low-memory domains (roughly under 400MB) our automatic memory limit
computation comes up with a limit that's too low. This is because the
0.5 multiplication does not add enough for such small values. Let's
increase the constant part of the computation to fix this.
Convert the type of loop iterators named 'i', 'j', k',
'ii', 'jj', 'kk', to be 'size_t' instead of 'int' or
'unsigned int', also santizing 'ii', 'jj', 'kk' to use
the normal 'i', 'j', 'k' naming
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Whenever virPortAllocatorRelease is called with port == 0, it complains
that the port is not in an allowed range, which is expectable as the
port was never allocated. Let's make virPortAllocatorRelease ignore 0
ports in a similar way free() ignores NULL pointers.
https://bugzilla.redhat.com/show_bug.cgi?id=981139
If a domain is paused before migration starts, we need to tell that to
the destination libvirtd to prevent it from resuming the domain at the
end of migration. This regression was introduced by commit 5379bb0.
Since commit 23e8b5d8, the code is refactored in a way that supports
domains with multiple graphics elements and commit 37b415200 allows
starting such domains. However none of those commits take migration
into account. Even though qemu doesn't support relocation for
anything else than SPICE and for no more than one graphics, there is no
reason to hardcode one graphics into this part of the code as well.
Add monitor callback API domainGuestPanic, that implements
'destroy', 'restart' and 'preserve' events of the 'on_crash'
in the XML when domain crashed.
After abf75aea24 the compiler screams:
qemu/qemu_driver.c: In function 'qemuNodeDeviceDetachFlags':
qemu/qemu_driver.c:10693:9: error: 'domain' may be used uninitialized in this function [-Werror=maybe-uninitialized]
pci = virPCIDeviceNew(domain, bus, slot, function);
^
qemu/qemu_driver.c:10693:9: error: 'bus' may be used uninitialized in this function [-Werror=maybe-uninitialized]
qemu/qemu_driver.c:10693:9: error: 'slot' may be used uninitialized in this function [-Werror=maybe-uninitialized]
qemu/qemu_driver.c:10693:9: error: 'function' may be used uninitialized in this function [-Werror=maybe-uninitialized]
Since the other functions qemuNodeDeviceReAttach and qemuNodeDeviceReset
looks exactly the same, I've initialized the variables there as well.
However, I am still wondering why those functions don't matter to gcc
while the first one does.
Implement check whether (maximum) vCPUs doesn't exceed machine
type's cpu-max settings.
On older versions of QEMU the check is disabled.
Signed-off-by: Michal Novotny <minovotn@redhat.com>
A loop in qemuPrepareHostdevPCIDevices() intended to cycle through all
the objects on the list pcidevs was doing "while (listcount > 0)", but
nothing in the body of the loop was reducing the size of the list - it
was instead removing items from a *different* list. It has now been
safely changed to a for() loop.
(This isn't as bad as it sounds - it's only a problem in case of an
OOM error.)
qemuGetActivePciHostDeviceList() had been creating a list that
contained pointers to objects that were also on the activePciHostdevs
list. In case of an OOM error, this newly created list would be
virObjectUnref'ed, which would cause everything on the list to be
freed. But all of those objects would still be on the
activePciHostdevs list, which could have very bad consequences if that
list was ever again accessed.
The solution used here is to populate the new list with *copies* of
the objects from the original list. It turns out that on return from
qemuGetActivePciHostDeviceList(), the caller would almost immediately
go through all the device objects and "steal" them (i.e. remove the
pointer from the list but not delete it) all from either one list or
the other; we now instead just *delete* (remove from the list and
free) each device from one list or the other, so in the end we have
the same state.
I realized after the fact that it's probably better in the long run to
give this function a name that matches the name of the link used in
sysfs to hold the group (iommu_group).
I'm changing it now because I'm about to add several more functions
that deal with iommu groups.
The driver arg to virPCIDeviceDetach is no longer used (the name of the stub driver is now set in the virPCIDevice object, and virPCIDeviceDetach retrieves it from there). Remove it.
I just learned that VFIO resets PCI devices when they are assigned to
guests / returned to the host, so it is redundant for libvirt to reset
the devices. This patch inhibits calling virPCIDeviceReset to devices
that will be/were assigned using VFIO.
virPCIDeviceDetach would previously sometimes consume the input device
object (to put it on the inactive list) and sometimes not. Avoiding
memory leaks required checking beforehand to see if the device was
already on the list, and freeing the device object in the caller only
if there wasn't already an identical object on the inactive list.
This patch makes it consistent - virPCIDeviceDetach will *never*
consume the input virPCIDevice object; if it needs to put one on the
inactive list, it will create a copy and put *that* on the list. This
way the caller knows that it is always their responsibility to free
the device object they created.
Previously stubDriver was always set from a string literal, so it was
okay to use a const char * that wasn't freed when the virPCIDevice was
freed. This will not be the case in the near future, so it is now a
char* that is allocated in virPCIDeviceSetStubDriver() and freed
during virPCIDeviceFree().
Add new CPU features for HyperV:
vapic for virtual APIC support
spinlocks for setting spinlock support
<features>
<hyperv>
<vapic state='on'/>
<spinlocks state='on' retries='4096'/>
</hyperv>
</features>
https://bugzilla.redhat.com/show_bug.cgi?id=784836
Commit 752596b5 broke the build with -Werror
qemu/qemu_hotplug.c: In function 'qemuDomainChangeGraphics':
qemu/qemu_hotplug.c:1980:39: error: declaration of 'listen' shadows a
global declaration [-Werror=shadow]
Fix with s/listen/newlisten/
Currently, we have a bug when updating a graphics device. A graphics device can
have a listen address set. This address is either defined by user (in which case
it's type is VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_ADDRESS) or it can be inherited
from a network (in which case it's type is
VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_NETWORK). However, in both cases we have a
listen address to process (e.g. during migration, as I've tried to fix in
7f15ebc7).
Later, when a user tries to update the graphics device (e.g. set a password),
we check if listen addresses match the original as qemu doesn't know how to
change listen address yet. Hence, users are required to not change the listen
address. The implementation then just dumps listen addresses and compare them.
Previously, while dumping the listen addresses, NULL was returned for NETWORK.
After my patch, this is no longer true, and we get a listen address for olddev
even if it is a type of NETWORK. So we have a real string on one side, the NULL
from user's XML on the other side and hence we think user wants to change the
listen address and we refuse it.
Therefore, we must take the type of listen address into account as well.
As a consequence of the cgroup layout changes from commit '632f78ca', the
qemuDomainGetSchedulerParameters[Flags]()' and qemuGetSchedulerType() APIs
failed to return data for a non running domain. This can be seen through
a 'virsh schedinfo <domain>' command which returns:
Scheduler : Unknown
error: Requested operation is not valid: cgroup CPU controller is not mounted
Prior to that change a non running domain would return:
Scheduler : posix
cpu_shares : 0
vcpu_period : 0
vcpu_quota : 0
emulator_period: 0
emulator_quota : 0
This patch will restore the capability to return configuration only data
for a non running domain regardless of whether cgroups are available.
This flag is meant for errors happening on the source of the migration
and isn't used on the destination. To allow better migration
compatibility, don't propagate it to the destination.
Paolo Bonzini pointed out that it's actually possible to migrate a qemu
instance that was paused due to I/O error and it will be able to work on
the destination if the storage is accessible.
This patch introduces flag VIR_MIGRATE_ABORT_ON_ERROR that cancels the
migration in case an I/O error happens while it's being performed and
allows migration without this flag. This flag can be possibly used for
other error reasons that may be introduced in the future.
Currently, we wait for SPICE to migrate in the very same loop where we
wait for qemu to migrate. This has a disadvantage of slowing seamless
migration down. One one hand, we should not kill the domain until all
SPICE data has been migrated. On the other hand, there is no need to
wait in the very same loop and hence slowing down 'cont' on the
destination. For instance, if users are watching a movie, they can
experience the movie to be stopped for a couple of seconds, as
processors are not running nor on src nor on dst as libvirt waits for
SPICE to migrate. We should move the waiting phase to migration CONFIRM
phase.
When qemu >= 1.20, it is safe to use -device for primary video
device as described in 4c993d8ab.
So, we are missing the cap flag in QMP capabilities detection, this
flag can be initialized safely in virQEMUCapsInitQMPBasic.
Convert input XML to migratable before using it in
qemuDomainSaveImageOpen.
XML in the save image is migratable, i.e. doesn't contain implicit
controllers. If these controllers were in a non-default order in the
input XML, the ABI check would fail. Removing and re-adding these
controllers fixes it.
https://bugzilla.redhat.com/show_bug.cgi?id=834196
During a live migration the guest may receive a disk access I/O error.
In this state the guest is unable to continue running on a remote host
after migration as some state may be present in the kernel and not
migrated.
With this patch, the migration is canceled in such case so it can either
continue on the source if the I/O issues are recovered or has to be
destroyed anyways.
https://bugzilla.redhat.com/show_bug.cgi?id=971485
As of d7f9d82753 we copy the listen
address from the qemu.conf config file in case none has been provided
via XML. But later, when migrating, we should not include such listen
address in the migratable XML as it is something autogenerated, not
requested by user. Moreover, the binding to the listen address will
likely fail, unless the address is '0.0.0.0' or its IPv6 equivalent.
This patch introduces a new boolean attribute to virDomainGraphicsListenDef
to distinguish autofilled listen addresses. However, we must keep the
attribute over libvirtd restarts, so it must be kept within status XML.
This patch fixes changes done in commit 29c1e913e4
that was pushed without implementing review feedback.
The flag introduced by the patch is changed to VIR_DOMAIN_VCPU_GUEST and
documentation makes the difference between regular hotplug and this new
functionality more explicit.
The virsh options that enable the use of the new flag are changed to
"--guest" and the documentation is fixed too.
Currently, there's a path to use the ncpuinfo variable uninitialized,
which leads to a compiler warning:
qemu/qemu_driver.c: In function 'qemuDomainGetVcpusFlags':
qemu/qemu_driver.c:4573:9: error: 'ncpuinfo' may be used
uninitialized in this function [-Werror=maybe-uninitialized]
for (i = 0; i < ncpuinfo; i++) {
^
This patch implements support for the "cpu-add" QMP command that plugs
CPUs into a live guest. The "cpu-add" command was introduced in QEMU
1.5. For the hotplug to work machine type "pc-i440fx-1.5" is required.
The qemu guest agent allows to online and offline CPUs from the
perspective of the guest. This patch adds helpers that call
'guest-get-vcpus' and 'guest-set-vcpus' guest agent functions and
convert the data for internal libvirt usage.
https://bugzilla.redhat.com/show_bug.cgi?id=964177
Though both libvirt and QEMU's document say RTC_CHANGE returns
the offset from the host UTC, qemu actually returns the offset
from the specified date instead when specific date is provided
(-rtc base=$date).
It's not safe for qemu to fix it in code, it worked like that
for 3 years, changing it now may break other QEMU use cases.
What qemu tries to do is to fix the document:
http://lists.gnu.org/archive/html/qemu-devel/2013-05/msg04782.html
And in libvirt side, instead of replying on the value from qemu,
this converts the offset returned from qemu to the offset from
host UTC, by:
/*
* a: the offset from qemu RTC_CHANGE event
* b: The specified date (-rtc base=$date)
* c: the host date when libvirt gets the RTC_CHANGE event
* offset: What libvirt will report
*/
offset = a + (b - c);
The specified date (-rtc base=$date) is recorded in clock's def as
an internal only member (may be useful to exposed outside?).
Internal only XML tag "basetime" is introduced to not lose the
guest's basetime after libvirt restarting/reloading:
<clock offset='variable' adjustment='304' basis='utc' basetime='1370423588'/>
Currently, a listen address for a SPICE server can be specified. Later,
when the domain is migrated, we need to relocate the graphics which
involves telling new destination to the SPICE server. However, we can't
just assume the listen address is the new location, because the listen
address can be ANYCAST (0.0.0.0 for IPv4, :: for IPv6). In which case,
we want to pass the remote hostname. But there are some troubles with
ANYCAST. In both IPv4 and IPv6 it has many ways for specifying such
address. For instance, in IPv4: 0, 0.0, 0.0.0, 0.0.0.0. The number of
variations gets bigger in IPv6 world. Hence, in order to check for
ANYCAST address sanely, we should take the provided listen address,
parse it and format back in it's full form. Which is exactly what this
patch does.
The work was done at the time of snapshot xmlstring parsing
if (offline && def->memory &&
def->memory != VIR_DOMAIN_SNAPSHOT_LOCATION_NONE) {
virReportError(...);
}
This should resolve:
https://bugzilla.redhat.com/show_bug.cgi?id=959191
The problem was that qemuUpdateActivePciHostdevs was returning 0
(success) when no hostdevs were present, but would otherwise return -1
(failure) even when it completed successfully. It is only called from
qemuProcessReconnect(), and when qemuProcessReconnect got back an
error, it would not only stop reconnecting, but would terminate the
guest qemu process "to remove danger of it ending up running twice if
user tries to start it again later".
(This bug was introduced in commit 011cf7ad, which was pushed between
v1.0.2 and v1.0.3, so all maintenance branches from v1.0.3 up to 1.0.5
will need this one line patch applied.)
A literal IPv6 must be escaped, otherwise migration fails with:
unable to execute QEMU command 'drive-mirror': address resolution failed
for f0::0d:5901: Servname not supported for ai_socktype
since QEMU treats everything after the first ':' as the port.
If snapshot creation failed for example due to invalid use of the
"REUSE_EXTERNAL" flag, libvirt killed access to the original image file
instead of the new image file. On machines with selinux this kills the
whole VM as the selinux context is enforced immediately.
* qemu_driver.c:qemuDomainSnapshotUndoSingleDiskActive():
- Kill access to the new image file instead of the old one.
Partially resolves: https://bugzilla.redhat.com/show_bug.cgi?id=906639
A bug in Cygwin [1] and poor error messages from gcc [2] lead
to this confusing compilation error:
qemu/qemu_monitor.c:418:9: error: passing argument 2 of 'sendmsg' from incmpatible pointer type
/usr/include/sys/socket.h:42:11: note: expected 'const struct msghdr *' but argument is of type 'struct msghdr *'
[1] http://cygwin.com/ml/cygwin/2013-05/msg00451.html
[2] http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57475
* src/qemu/qemu_monitor.c (includes): Include <sys/socket.h>
before <sys/un.h>.
Signed-off-by: Eric Blake <eblake@redhat.com>
This is a recurring problem for cygwin :)
For example, see commit 23a4df88.
qemu/qemu_driver.c: In function 'qemuStateInitialize':
qemu/qemu_driver.c:691:13: error: format '%d' expects type 'int', but argument 8 has type 'uid_t' [-Wformat]
* src/qemu/qemu_driver.c (qemuStateInitialize): Add casts.
* daemon/remote.c (remoteDispatchAuthList): Likewise.
Signed-off-by: Eric Blake <eblake@redhat.com>
A cygwin build of the qemu driver fails with:
qemu/qemu_process.c: In function 'qemuPrepareCpumap':
qemu/qemu_process.c:1803:31: error: 'CPU_SETSIZE' undeclared (first use in this function)
CPU_SETSIZE is a Linux extension in <sched.h>; a bit more portable
is using sysconf if _SC_NPROCESSORS_CONF is defined (several platforms
have it, including Cygwin). Ultimately, I would have preferred to
use gnulib's 'nproc' module, but it is currently under an incompatible
license.
* src/qemu/qemu_conf.h (QEMUD_CPUMASK_LEN): Provide definition on
cygwin.
Signed-off-by: Eric Blake <eblake@redhat.com>
Currently, if there's an error opening /dev/vhost-net (e.g. because
it doesn't exist) but it's not required we proceed with vhostfd array
filled with -1 and vhostfdSize unchanged. Later, when constructing
the qemu command line only non-negative items within vhostfd array
are taken into account. This means, vhostfdSize may be greater than
the actual count of non-negative items in vhostfd array. This results
in improper command line arguments being generated, e.g.:
-netdev tap,fd=21,id=hostnet0,vhost=on,vhostfd=(null)
If we are just ejecting media, ret == -1 even after the retry loop
determines that the tray is open, as requested. This means media
disconnect always report's error.
Fix it, and fix some other mini issues:
- Don't overwrite the 'eject' error message if the retry loop fails
- Move the retries decrement inside the loop, otherwise the final loop
might succeed, yet retries == 0 and we will raise error
- Setting ret = -1 in the disk->src check is unneeded
- Fix comment typos
cc: mprivozn@redhat.com
Currently qemuDomainReboot() does reboot in two phases:
qemuMonitorSystemPowerdown() and qemuProcessFakeReboot().
qemuMonitorSystemPowerdown() shutdowns the domain and saves domain
state/reason as VIR_DOMAIN_SHUTDOWN_UNKNOWN.
qemuProcessFakeReboot() sets domain state/reason to
VIR_DOMAIN_RESUMED_UNPAUSED but does not save domain state changes.
Subsequent restart of libvirtd leads to restoring domain state/reason to
saved that is VIR_DOMAIN_SHUTDOWN_UNKNOWN and to automatic shutdown of
the domain. This commit adds virDomainSaveStatus() into
qemuProcessFakeReboot() to avoid unexpected shutdowns.
With previous patch, we accept negative value as length of string to
duplicate. So there is no need to pass strlen(src) in case we want to do
duplicate the whole string.
Function qemuDomainSetBlockIoTune() was checking QEMU capabilities
even when !(flags & VIR_DOMAIN_AFFECT_LIVE) and the domain was
shutoff, resulting in the following problem:
virsh # domstate asdf; blkdeviotune asdf vda --write-bytes-sec 100
shut off
error: Unable to change block I/O throttle
error: unsupported configuration: block I/O throttling not supported with this QEMU binary
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=965016
Since f03dcc5 we use [::] as the listening address both on qemu
command line in -incoming and in nbd-server-start QMP command.
However the latter requires just :: without the braces.
In order to learn libvirt multiqueue several things must be done:
1) The '/dev/net/tun' device needs to be opened multiple times with
IFF_MULTI_QUEUE flag passed to ioctl(fd, TUNSETIFF, &ifr);
2) Similarly, '/dev/vhost-net' must be opened as many times as in 1)
in order to keep 1:1 ratio recommended by qemu and kernel folks.
3) The command line construction code needs to switch from 'fd=X' to
'fds=X:Y:...:Z' and from 'vhostfd=X' to 'vhostfds=X:Y:...:Z'.
4) The monitor handling code needs to learn to pass multiple FDs.
In 84c59ffa I've tried to fix changing ejectable media process. The
process should go like this:
1) we need to call 'eject' on the monitor
2) we should wait for 'DEVICE_TRAY_MOVED' event
3) now we can issue 'change' command
However, while waiting in step 2) the domain monitor was locked. So
even if qemu reported the desired event, the proper callback was not
called immediately. The monitor handling code needs to lock the
monitor in order to read the event. So that's the first lock we must
not hold while waiting. The second one is the domain lock. When
monitor handling code reads an event, the appropriate callback is
called then. The first thing that each callback does is locking the
corresponding domain as a domain or its device is about to change
state. So we need to unlock both monitor and VM lock. Well, holding
any lock while sleep()-ing is not the best thing to do anyway.
Since 0d70656afd, it starts to access the sysfs files to build
the qemu command line (by virSCSIDeviceGetSgName, which is to find
out the scsi generic device name by adpater🚌target:unit), there
is no way to work around, qemu wants to see the scsi generic device
like "/dev/sg6" anyway.
And there might be other places which need to access sysfs files
when building qemu command line in future.
Instead of increasing the arguments of qemuBuildCommandLine, this
introduces a new callback for qemuBuildCommandLine, and thus tests
can register their own callbacks for sysfs test input files accessing.
* src/qemu/qemu_command.h: (New callback struct
qemuBuildCommandLineCallbacks;
extern buildCommandLineCallbacks)
* src/qemu/qemu_command.c: (wire up the callback struct)
* src/qemu/qemu_driver.c: (Use the new syntax of qemuBuildCommandLine)
* src/qemu/qemu_hotplug.c: Likewise
* src/qemu/qemu_process.c: Likewise
* tests/testutilsqemu.[ch]: (Helper testSCSIDeviceGetSgName;
callback struct testCallbacks;)
* tests/qemuxml2argvtest.c: (Use testCallbacks)
* src/tests/qemuxmlnstest.c: (Like above)
Resolves:https://bugzilla.redhat.com/show_bug.cgi?id=927620
#kill -STOP `pidof qemu-kvm`
#virsh destroy $guest --graceful
error: Failed to destroy domain testVM
error: An error occurred, but the cause is unknown
With --graceful, SIGTERM always is emitted to kill driver
process, but it won't success till burning out waiting time
in case of process being stopped.
But domain destroy without --graceful can work, SIGKILL will
be emitted to the stopped process after 10 secs which always
kills a process even one that is currently stopped.
So report an error after burning out waiting time in this case.
