Some code paths already assume that it is allocated since it was always
allocated by virDomainPerfDefParseXML. Make it member of virDomainDef
directly so that we don't have to allocate it all the time.
This fixes crash when attempting to connect to an existing process via
virDomainQemuAttach since we would not allocate it in that code path.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1350688
Ensure that the given controller and all controllers with a smaller
index exist; there must not be any missing index in between.
Reviewed-by: Boris Fiuczynski <fiuczy@linux.vnet.ibm.com>
Reviewed-by: Bjoern Walk <bwalk@linux.vnet.ibm.com>
Signed-off-by: Marc Hartmayer <mhartmay@linux.vnet.ibm.com>
The commit "qemu: hot-plug: Assume support for -device in
qemuDomainAttachSCSIDisk" dropped the code for the automatic SCSI
controller creation used in SCSI disk hot-plugging. If we are
hot-plugging a SCSI disk to a domain and there is no proper SCSI
controller defined, it results in an "error: internal error: Could not
find scsi controller with index X required for device" error.
For that reason reverting a hunk of the commit
d4d32005d6.
This patch also adds an extra comment to the code to clarify the
loop.
Reviewed-by: Boris Fiuczynski <fiuczy@linux.vnet.ibm.com>
Reviewed-by: Bjoern Walk <bwalk@linux.vnet.ibm.com>
Signed-off-by: Marc Hartmayer <mhartmay@linux.vnet.ibm.com>
CVE-2016-5008
Setting an empty graphics password is documented as a way to disable
VNC/SPICE access, but QEMU does not always behaves like that. VNC would
happily accept the empty password. Let's enforce the behavior by setting
password expiration to "now".
https://bugzilla.redhat.com/show_bug.cgi?id=1180092
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
libvirt's qemu driver doesn't have direct access to the config on the
guest side of a network interface, and currently doesn't have any
method in place to even inform the guest of the desired config. In the
future, an unenforceable attempt to set the guest-side IP info could
be made by adding a static host entry to the appropriate dnsmasq
configuration (or changing the default dhcp client address on the qemu
commandline for type='user' interfaces), or enhancing the guest agent
to allow setting an IP address, but for now it can't have any effect,
and we don't want to give the illusion that it does.
To prevent the "disappearance" of any existing configs with ip
address/route info (due to parser failure), this check is added in the
newly implemented qemuDomainDeviceDefValidate(), which is only called
when a domain is defined or started, *not* when it is reread from disk
at libvirtd startup.
When support for <interface type='ethernet'> was added in commit
9a4b705f back in 2010, it erroneously looked at <source dev='blah'/>
for a user-specified guest-side interface name. This was never
documented though. (that attribute already existed at the time in the
data.ethernet union member of virDomainNetDef, but apparently had no
practical use - it was only used as a storage place for a NetDef's
bridge name during qemuDomainXMLToNative(), but even then that was
never used for anything).
When support for similar guest-side device naming was added to the lxc
driver several years later, it was put in a new subelement <guest
dev='blah'/>.
In the intervening years, since there was no validation that
ethernet.dev was NULL in the other drivers that didn't actually use
it, innocent souls who were adding other features assuming they needed
to account for non-NULL ethernet.dev when really they didn't, so
little bits of the usual pointless cargo-cult code showed up.
This patch not only switches the openvz driver to use the documented
<guest dev='blah'/> notation for naming the guest-side device (just in
case anyone is still using the openvz driver), and logs an error if
anyone tries to set <source dev='blah'/> for a type='ethernet'
interface, it also removes the cargo-cult uses of ethernet.dev and
<source dev='blah'/>, and eliminates if from the RNG and from
virDomainNetDef.
NB: I decided on this course of action after mentioning the
inconsistency here:
https://www.redhat.com/archives/libvir-list/2016-May/msg02038.html
and getting encouragement do eliminate it in a later IRC discussion
with danpb.
in qemuConnectDomainXMLToNative. This function was only accounting for
about 1/10 of all the allocated items in the NetDef prior to memseting
it to all 0's. On top of that, it was going to great pains to learn
the name of the bridge device, but then never doing anything useful
with it (just putting it into data.ethernet.dev, which is *never* used
when building a qemu commandline). (I think this again all started off
as code with good intentions, but it was never completed, and instead
was just Frankensteinically cargo-culted into the odd mish mash we
have today).
The resulting code is much simpler, produces exactly the same output,
and doesn't leak memory.
This patch removes the expanded and duplicated code that all sprung
out of two well-intentioned-but-useless settings of
net->data.(bridge|ethernet).ipaddr.
qemu has never supported even a single IP address in the interface
config, much less a list of them. All of the instances of "clearing
out the IP addresses" that are now in this function originated with
commit d8dbd6 "Basic domain XML conversions for Xen/QEMU drivers" in
May 2009, but even then the single "ipaddr" in the struct for
type='ethernet' and type='bridge' wasn't used in the qemu driver (only
in xen and openvz). Since then anyone who added a new interface type
also tacked on another unnecessary clearing of ipaddr, and when it was
made into a list of IPs (so far supported only by the LXC driver) this
simple setting was turned into a loop (well, multiple loops) to clear
them all.
I'm tired of mistyping this all the time, so let's do it the same all
the time (similar to how we changed all "Pci" to "PCI" awhile back).
(NB: I've left alone some things in the esx and vbox drivers because
I'm unable to compile them and they weren't obviously *not* a part of
some API. I also didn't change a couple of variables named,
e.g. "somethingIptables", because they were derived from the name of
the "iptables" command)
Rather than pass authdef, pass the 'authdef->username' and the
'&authdef->secdef'
Note that a username may be NULL.
Signed-off-by: John Ferlan <jferlan@redhat.com>
Rather than assume/pass the protocol to the qemuDomainSecretPlainSetup
and qemuDomainSecretAESSetup, set and pass the secretUsageType based
on the src->protocol type. This will eventually be used by the
virSecretGetSecretString call
Signed-off-by: John Ferlan <jferlan@redhat.com>
This kvmGetMaxVCPUs() needs to be used at two different places
so move it to utils with appropriate name and mark it as private
global now.
Signed-off-by: Shivaprasad G Bhat <sbhat@linux.vnet.ibm.com>
On PPC the legacy passthrough is not supported and only
VFIO is supported. So, the checks at places to confirm if the
host is passthrough capable checks only legacy, fix it. This
is seen at only one place now.
Signed-off-by: Shivaprasad G Bhat <sbhat@linux.vnet.ibm.com>
Unfortunately, we can't just call qemuDomainMachineIsPSeries()
here, because we don't have a virDomainDef instance; that said,
the open-coded check should match said function as closely as
possible.
The directories we iterate over are unlikely to contain any entries
starting with a dot, other than '.' and '..' which is already skipped
by virDirRead.
Move the enum into a new src/util/virsecret.h, rename it to be
virSecretLookupType. Add a src/util/virsecret.h in order to perform
a couple of simple operations on the secret XML and virSecretLookupTypeDef
for clearing and copying.
This includes quite a bit of collateral damage, but the goal is to remove
the "virStorage*" and replace with the virSecretLookupType so that it's
easier to to add new lookups that aren't necessarily storage pool related.
Signed-off-by: John Ferlan <jferlan@redhat.com>
One can not issue monitor commands manually during async calls thru
designated API while this could be useful for testing/debugging purposes.
qemuDomainQemuMonitorCommand uses job of type QEMU_JOB_MODIFY and any async
call disable parallel execution of this type of job. The only state that is
changed is taint variable. AFAIU the only place we can mess is resetting
taint flag in qemuProcessStop routine under some async job. But this can not
happen thanx to both virDomainObjIsActive check in qemuDomainQemuMonitorCommand
and resetting active status in qemuProcessStop before taint flag.
Change job type to QEMU_JOB_QUERY and thus make the API call available for
most of async jobs.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
Several places in the code update qemuMonitorMigrationParams structure
and qemuMigrationSetParams is then used to set them all at once.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
We should not require any parameters to be present. After all we have
the *_set bools to express that some parameters were not set.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
qemuMonitorMigrationParams is a better name for a structure which
contains various migration parameters. While doing that, we should use
full names for individual parameters.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Checking whether the function has anything to do is better done in the
function rather then requiring callers to do that.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Since virQEMUCapsNewForBinaryInternal was introduced,
virQEMUCapsNewForBinary is no longer used outside qemu_capabilities.c.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Allow gathering available vcpu ids, their state and offlinability via
the qemu guest agent. The maximum id was chosen arbitrarily and ought
to be enough for everybody.
Documentation for the "guest-set-vcpus" command describes a proper
algorithm how to set vcpus. This patch makes the following changes:
- state of cpus that has not changed is not updated
- if the command was partially successful the command is re-tried with
the rest of the arguments to get a proper error message
- code is more robust against malicious guest agent
- fix testsuite to the new semantics
So far this is only useful for recalculating NUMA memory size,
which this function cannot parse.
This will let us generate USB addresses based on this flag.
The '-usb' option doesn't have any effect for aarch64 mach-virt
guests, so the fact that it's currently enabled by default is not
really causing any issue.
However, that might change in the future (although unlikely), and
having it as part of the QEMU command line can cause confusion to
someone looking through the process list.
Avoid it completely, like it's already happening for q35.
There has been some progress lately in enabling virtio-pci on
aarch64 guests; however, guest OS support is still spotty at best,
so most guests are going to be using virtio-mmio instead.
Currently, mach-virt guests are closely modeled after q35 guests,
and that includes always adding a dmi-to-pci-bridge that's just
impossible to get rid of. While that's acceptable (if suboptimal)
for q35, where you will always need some kind of PCI device anyway,
mach-virt guests should be allowed to avoid it.
While we need to know the difference between the total memory stored in
<memory> and the actual size not included in the possible memory modules
we can't pre-calculate it reliably. This is due to the fact that
libvirt's XML is copied via formatting and parsing the XML and the
initial memory size can be reliably calculated only when certain
conditions are met due to backwards compatibility.
This patch removes the storage of 'initial_memory' and fixes the helpers
to recalculate the initial memory size all the time from the total
memory size. This conversion is possible when we also make sure that
memory hotplug accounts properly for the update of the total memory size
and thus the helpers for inserting and removing memory devices need to
be tweaked too.
This fixes a bug where a cold-plug and cold-remove of a memory device
would increase the size reported in <memory> in the XML by the size of
the memory device. This would happen as the persistent definition is
copied before attaching the device and this would lead to the loss of
data in 'initial_memory'.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1344892
Until now, a Q35 domain (or arm/virt, or any other domain that has a
pcie-root bus) would always have a pci-bridge added, so that there
would be a hotpluggable standard PCI slot available to plug in any PCI
devices that might be added. This patch removes the explicit add,
instead relying on the pci-bridge being auto-added during PCI address
assignment (it will add a pci-bridge if there are no free slots).
This doesn't eliminate the dmi-to-pci-bridge controller that is
explicitly added whether or not a standard PCI slot is required (and
that is almost never used as anything other than a converter between
pcie.0's PCIe slots and standard PCI). That will be done separately.
Previously there was no way to have a Q35 domain that didn't have
these two controllers. This patch skips their creation as long as
there are some other kinds of pci controllers at index 1 and 2
(e.g. some pcie-root-port controllers).
I'm hoping that soon we won't add them at all, plugging all devices
into auto-added pcie-*-port ports instead, but in the meantime this
makes it easier to experiment with alternative bus hierarchies.
The other two DomainHasBlockJob usage error messages don't contain
'an', so unify things to save translators some effort. Dropping
the 'an' is closer to the sentence structure in the errors from
qemuDomainDiskBlockJobIsActive as well
If the domain is not running, but for example the CPUs are stopped, the
ACPI event gets queued and resume of the domain will just shut it off.
https://bugzilla.redhat.com/show_bug.cgi?id=1216281
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Since obtaining a job can wait for another job to finish, the state
might change in the meantime. And checking it more than once is
pointless.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
The virQEMUDriverConfig object contains lists of
loader:nvram pairs to advertise firmwares supported by
by the driver, and qemu_conf.c contains code to populate
the lists, all of which is useful for other drivers too.
To avoid code duplication, introduce a virFirmware object
to encapsulate firmware details and switch the qemu driver
to use it.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
https://bugzilla.redhat.com/show_bug.cgi?id=1298070
We have the code for attaching redirdevs for ages now.
Unfortunately, our monitor code that handles talking to the qemu
process was missing a little piece of code that actually enabled
the feature.
BTW: it really is called "type" on the monitor, even though it's
called "name" on the cmd line. Don't ask.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Move all APIs with a virHostMEM name prefix out into new
util/virhostmem.h & util/virhostmem.c files
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Move all APIs with a virHostCPU name prefix out into new
util/virhostcpu.h & util/virhostcpu.c files
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
In preparation for moving all the CPU related APIs out of
the nodeinfo file, give them a virHostCPU name prefix.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
In preparation for moving all the memory related APIs out of
the nodeinfo file, give them a virHostMem name prefix.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Nearly all the methods in the nodeinfo file are given a
'const char *sysfs_prefix' parameter to override the
default sysfs path (/sys/devices/system). Every single
caller passes in NULL for this, except one use in the
unit tests. Furthermore this parameter is totally
Linux-specific, when the APIs are intended to be cross
platform portable.
This removes the sysfs_prefix parameter and instead gives
a new method linuxNodeInfoSetSysFSSystemPath for use by
the test suite.
For two of the methods this hardcodes use of the constant
SYSFS_SYSTEM_PATH, since the test suite does not need to
override the path for thos methods.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
If you want to set block device I/O tuning values that end with '_max'
and there is nothing else set, libvirt emits an error. In particular:
error: internal error: Unexpected error
That's an unknown error. That is because *_max values depend on their
respective non-_max values. QEMU even says that in the error message
sent as a response to the monitor command:
"error": {"class": "GenericError", "desc": "bps_max/iops_max require
corresponding bps/iops values"}
the problem was that we didn't know that and there was no check for it.
Adding such check makes sure that there will be less confused users.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
This new listen type is currently supported only by spice graphics.
It's introduced to make it easier and clearer specify to not listen
anywhere in order to start a guest with OpenGL support.
The old way to do this was set spice graphics autoport='no' and don't
specify any ports. The new way is to use <listen type='none'/>. In
order to be able to migrate to old libvirt the migratable XML will be
generated without the listen element and with autoport='no'. Also the
old configuration will be automatically converted to the this listen
type.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
VNC graphics already supports sockets but only via 'socket' attribute.
This patch coverts that attribute into listen type 'socket'.
For backward compatibility we need to handle listen type 'socket' and 'socket'
attribute properly to support old XMLs and new XMLs. If both are provided they
have to match, if only one of them is provided we need to be able to parse that
configuration too.
To not break migration back to old libvirt if the socket is provided by user we
need to generate migratable XML without the listen element and use only 'socket'
attribute.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
This moves the socket generation if "vnc_auto_unix_socket" is set.
It also fixes a bug with this config option that we should auto-generate
socket path only if listen type is address and there is no address
specified.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Even though it's auto-generated it's based on qemu.conf option and listen type
address already uses "fromConfig" to carry this information. Following commits
will convert the socket to listen element so this rename is required because
there will be also an option to get socket auto-generated independently on the
qemu.conf option.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Since commit 7140807917, qemu agent
channel cannot be plugged in because we won't generate its path
automatically. Let's not only fix that, but also add tests for it so
next time it's checked for.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1322210
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Put it into separate function called qemuDomainPrepareChannel() and call
it from the new qemuProcessPrepareDomain().
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
CPUID instruction normally takes its parameter from EAX, but sometimes
ECX is used as an additional parameter. This patch prepares the x86 CPU
driver code for the new 'ecx_in' CPUID parameter.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
virCPUData, virCPUx86Feature, and virCPUx86Model all contained a pointer
to virCPUx86Data, which was not very nice since the real CPUID data were
accessible by yet another pointer from virCPUx86Data. Moreover, using
virCPUx86Data directly will make static definitions of internal CPU
features a bit easier.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
This patch splits qemuMonitorJSONGetCPUx86Data in three functions:
- qemuMonitorJSONCheckCPUx86 checks if QEMU supports reporting CPUID
features for a guest CPU
- qemuMonitorJSONParseCPUx86Features parses CPUID features from a JSON
array
- qemuMonitorJSONGetCPUx86Data gets the requested guest CPU property
from QOM and uses qemuMonitorJSONParseCPUx86Features to parse it
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
CPUID instruction normally takes its parameter from EAX, but sometimes
ECX is used as an additional parameter. Let's rename 'function' to
'eax_in' in preparation for adding 'ecx_in'.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
The virConnectOpenInternal method opens the libvirt client
config file and uses it to resolve things like URI aliases.
There may be driver specific things that are useful to
store in the config file too, so rather than have them
re-parse the same file, pass the virConfPtr down to the
drivers.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
In commit 1e38ef72 the disk startup policy check was moved prior to the
call to virDomainObjSetDefTransient which dropped the disk from the
config rather than the def to be started which is a bug.
Additionally we'd not report the disk change event for this since the
disk aliases were not set at that point.
Finally 'volume' based disks would not work with startup policy too.
Fix it by moving it back after the definition is copied, aliases are
assigned and disk sources are translated.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1341415
qemuProcessStart does not unset the infrastructure that retrieves errors
from the qemu log file in case of migration. As this wasn't handled
properly in qemuDomainSaveImageStartVM we kept the logging context/fd
open for the lifetime of the VM rather than closing it after it's not
needed.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1325080
Use qemuDomainLogAppendMessage rather than attempting to open a new
logging context with file descriptors. The new approach allows to log
the message even if qemu is still running at that point which appens
during migration finish phase where qemuProcessStop is killing qemu.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1312188
Along with the virtlogd addition of the log file appending API implement
a helper for logging one-shot entries to the log file including the
fallback approach of using direct file access.
This will be used for noting the shutdown of the qemu proces and
possibly other actions such as VM migration and other critical VM
lifecycle events.
Since it will not be called from outside of conf we can unexport it too
if we move it to the appropriate place.
Test suite change is necessary since the error will be reported sooner
now.
Validation of qemu process startup requires to know whether the process
is used for a fresh VM or whether it's reloaded from a
snapshot/migration. Pass this information in via a flag rather than
calculating it from a bunch of bools.
Similarly to the domain definition validator add a device validator. The
change to the prototype of the domain validator is necessary as
virDomainDeviceInfoIterateInternal requires a non-const pointer.
Until now we weren't able to add checks that would reject configuration
once accepted by the parser. This patch adds a new callback and
infrastructure to add such checks. In this patch all the places where
rejecting a now-invalid configuration wouldn't be a good idea are marked
with a new parser flag.
Historically, we added heads=1 to videos, but for example for qxl, we
did not reflect that on the command line.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1283207
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Move the module from qemu_command.c to a new module virqemu.c and
rename the API to virQEMUBuildObjectCommandline.
This API will then be shareable with qemu-img and the need to build
a security object for luks support.
Signed-off-by: John Ferlan <jferlan@redhat.com>
Remove the live attribute and mark the definition as transient
whether the domain is runing or not.
There were only two callers left calling with live=false:
* testDomainStartState, where the domain already is active
because we assigned vm->def->id just a few lines above the call
* virDomainObjGetPersistentDef, which now only calls
virDomainObjSetDefTransient for an active domain
Commit b4a5fd95 introduced vram64 attribute for QXL video device but
there were two issues. Only function
qemuMonitorJSONUpdateVideoVram64Size should update the vram64 attribute
and also the value is in MiB, not in B.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
There's a bug in the function. We expect the following format for
the data we are parsing here:
key: value
So we use strchr() to find ':' and then see if it is followed by
space. But the check that does just that is slightly incorrect.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Commit ff2126225d changed the error message to be more
detailed about the failure at hand; however, while the new
error message claims that "bus must be <= index", the error
message is displayed if "idx <= addr->bus", ie. when bus
is larger than or *equal to* index.
Change the error message to report the correct constraint,
and format it in a way that mirrors the check exactly to
make it clearer to people reading the code. The new error
message reads "index must be larger than bus".
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1339900
Hand-entering indexes for 20 PCI controllers is not as tedious as
manually determining and entering their PCI addresses, but it's still
annoying, and the algorithm for determining the proper index is
incredibly simple (in all cases except one) - just pick the lowest
unused index.
The one exception is USB2 controllers because multiple controllers in
the same group have the same index. For these we look to see if 1) the
most recently added USB controller is also a USB2 controller, and 2)
the group *that* controller belongs to doesn't yet have a controller
of the exact model we're just now adding - if both are true, the new
controller gets the same index, but in all other cases we just assign
the lowest unused index.
With this patch in place and combined with the automatic PCI address
assignment, we can define a PCIe switch with several ports like this:
<controller type='pci' model='pcie-root-port'/>
<controller type='pci' model='pcie-switch-upstream-port'/>
<controller type='pci' model='pcie-switch-downstream-port'/>
<controller type='pci' model='pcie-switch-downstream-port'/>
<controller type='pci' model='pcie-switch-downstream-port'/>
<controller type='pci' model='pcie-switch-downstream-port'/>
<controller type='pci' model='pcie-switch-downstream-port'/>
...
These will each get a unique index, and PCI addresses that connect
them together appropriately with no pesky numbers required.
IS_USB2_CONTROLLER() is useful in more places aside from just when
assigning PCI addresses in QEMU, and is checking for enum values that
are all defined in conf/domain_conf.h anyway, so define it there
instead.
<os>
<acpi>
<table type="slic">/path/to/acpi/table/file</table>
</acpi>
</os>
will result in:
-acpitable sig=SLIC,file=/path/to/acpi/table/file
This option was introduced by QEMU commit 8a92ea2 in 2009.
https://bugzilla.redhat.com/show_bug.cgi?id=1327537
The libvirt internal bits can be changed for disks that don't otherwise
support changing media. Remove the switch statement and allow changes of
non-source data for all disks.
qemuDomainChangeDiskLive rolled back few changes to the disk definition
if changing of the media failed. This can be avoided by moving some code
around.
Based on some digital archaeology performed by jtomko, it's been determined
that the persistentAddrs variable is no longer necessary...
The variable was added by:
commit 141dea6bc7
CommitDate: 2010-02-12 17:25:52 +0000
Add persistence of PCI addresses to QEMU
Where it was set to 0 on domain startup if qemu did not support the
QEMUD_CMD_FLAG_DEVICE capability, to clear the addresses at shutdown,
because QEMU might make up different ones next time.
As of commit f5dd58a608
CommitDate: 2012-07-11 11:19:05 +0200
qemu: Extended qemuDomainAssignAddresses to be callable from
everywhere.
this was broken, when the persistentAddrs = 0 assignment was moved
inside qemuDomainAssignPCIAddresses and while it pretends to check
for !QEMU_CAPS_DEVICE, its parent qemuDomainAssignAddresses is only
called if QEMU_CAPS_DEVICE is present.
Since commit id '20a0fa8e' removed the QEMU_CAPS_DEVICE, Coverity notes
that it's no longer possible to have 'addrs' be NULL when checking for
a live domain since qemuDomainPCIAddressSetCreate would have jumped to
cleanup if addrs was NULL.
Use the detected tray presence flag to trigger the tray waiting code
only if the given storage device in qemu reports to have a tray.
This is necessary as the floppy device lost it's tray as of qemu commit:
commit abb3e55b5b718d6392441f56ba0729a62105ac56
Author: Max Reitz <mreitz@redhat.com>
Date: Fri Jan 29 20:49:12 2016 +0100
Revert "hw/block/fdc: Implement tray status"
Commit 1fad65d49a used a really big hammer
and overwrote the error message that might be reported by qemu if the
tray is locked. Fix it by reporting the error only if no error is
currently set.
Error after commit mentioned above:
error: internal error: timed out waiting for disk tray status update
New error:
error: internal error: unable to execute QEMU command 'eject': Tray of
device 'drive-ide0-0-0' is not open
Extract information for all disks and update tray state and source only
for removable drives. Additionally store whether a drive is removable
and whether it has a tray.
Extract whether a given drive has a tray and whether there is no image
inserted.
Negative logic for the image insertion is chosen so that the flag is set
only if we are certain of the fact.
Rather than only assigning a PCI address when no address is given at
all, also do it when the config says that the address type is 'pci',
but it gives no address (virDeviceInfoPCIAddressWanted()).
There are also several places after parsing but prior to address
assignment where code previously expected that any info with address
type='pci' would have a *valid* PCI address, which isn't always the
case - now we check not only for type='pci', but also for a valid
address (virDeviceInfoPCIAddressPresent()).
The test case added in this patch was directly copied from Cole's patch titled:
qemu: Wire up address type=pci auto_allocate
https://bugzilla.redhat.com/show_bug.cgi?id=1182074
If they're available and we need to pass secrets to qemu, then use the
qemu domain secret object in order to pass the secrets for RBD volumes
instead of passing the base64 encoded secret on the command line.
The goal is to make AES secrets the default and have no user interaction
required in order to allow using the AES mechanism. If the mechanism
is not available, then fall back to the current plain mechanism using
a base64 encoded secret.
New APIs:
qemu_domain.c:
qemuDomainGetSecretAESAlias:
Generate/return the secret object alias for an AES Secret Info type.
This will be called from qemuDomainSecretAESSetup.
qemuDomainSecretAESSetup: (private)
This API handles the details of the generation of the AES secret
and saves the pieces that need to be passed to qemu in order for
the secret to be decrypted. The encrypted secret based upon the
domain master key, an initialization vector (16 byte random value),
and the stored secret. Finally, the requirement from qemu is the IV
and encrypted secret are to be base64 encoded.
qemu_command.c:
qemuBuildSecretInfoProps: (private)
Generate/return a JSON properties object for the AES secret to
be used by both the command building and eventually the hotplug
code in order to add the secret object. Code was designed so that
in the future perhaps hotplug could use it if it made sense.
qemuBuildObjectSecretCommandLine (private)
Generate and add to the command line the -object secret for the
secret. This will be required for the subsequent RBD reference
to the object.
qemuBuildDiskSecinfoCommandLine (private)
Handle adding the AES secret object.
Adjustments:
qemu_domain.c:
The qemuDomainSecretSetup was altered to call either the AES or Plain
Setup functions based upon whether AES secrets are possible (we have
the encryption API) or not, we have secrets, and of course if the
protocol source is RBD.
qemu_command.c:
Adjust the qemuBuildRBDSecinfoURI API's in order to generate the
specific command options for an AES secret, such as:
-object secret,id=$alias,keyid=$masterKey,data=$base64encodedencrypted,
format=base64
-drive file=rbd:pool/image:id=myname:auth_supported=cephx\;none:\
mon_host=mon1.example.org\:6321,password-secret=$alias,...
where the 'id=' value is the secret object alias generated by
concatenating the disk alias and "-aesKey0". The 'keyid= $masterKey'
is the master key shared with qemu, and the -drive syntax will
reference that alias as the 'password-secret'. For the -drive
syntax, the 'id=myname' is kept to define the username, while the
'key=$base64 encoded secret' is removed.
While according to the syntax described for qemu commit '60390a21'
or as seen in the email archive:
https://lists.gnu.org/archive/html/qemu-devel/2016-01/msg04083.html
it is possible to pass a plaintext password via a file, the qemu
commit 'ac1d8878' describes the more feature rich 'keyid=' option
based upon the shared masterKey.
Add tests for checking/comparing output.
NB: For hotplug, since the hotplug code doesn't add command line
arguments, passing the encoded secret directly to the monitor
will suffice.
Move the logic from qemuDomainGenerateRandomKey into this new
function, altering the comments, variable names, and error messages
to keep things more generic.
NB: Although perhaps more reasonable to add soemthing to virrandom.c.
The virrandom.c was included in the setuid_rpc_client, so I chose
placement in vircrypto.
According to QEMU docs, the '-m' option for specifying RAM is by default
in MiB, and a suffix of "M" or "G" may be passed for values in MiB and
GiB respectively. This commit adds support and a test for the same.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=812295
Signed-off-by: Nishith Shah <nishithshah.2211@gmail.com>
Both VNC and SPICE requires the same code to resolve address for listen
type network. Remove code duplication and create a new function that
will be used in qemuProcessSetupGraphics().
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
For some disk types (SD), we want to emit the syntax
we used for disks before -device was available even
if QEMU supports -device.
Use the qemuDiskBusNeedsDeviceArg helper to figure out
whether to use the old or new syntax.
If the stats for a block device can't be acquired from qemu we've
fallen back to loading them from the file on the disk in libvirt.
If qemu is not cooperating due to being stuck on an inaccessible NFS
share we would then attempt to read the files and get stuck too with
the VM object locked. All other APIs would eventually get stuck waiting
on the VM lock.
Avoid this problem by skipping the block stats if the VM is online but
the monitor did not provide any stats.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1337073
Some Intel processor families (e.g. the Intel Xeon processor E5 v3
family) introduced some RDT (Resource Director Technology) features
to monitor or control shared resource. Among these features, MBM
(Memory Bandwidth Monitoring), which is build on the CMT (Cache
Monitoring Technology) infrastructure, provides OS/VMM a way to
monitor bandwidth from one level of cache to another.
With current perf framework, this patch adds support to perf event
for MBM.
Signed-off-by: Qiaowei Ren <qiaowei.ren@intel.com>
QEMU needs access to the /dev/dri/render* device for
virgl to work.
Allow access to all /dev/dri/* devices for domains with
<video>
<model type='virtio' heads='1' primary='yes'>
<acceleration accel3d='yes'/>
</model>
</video>
https://bugzilla.redhat.com/show_bug.cgi?id=1337290
All qemu versions we support have QEMU_CAPS_DEVICE, so checking
for it is redundant. Remove the usage.
The code diff isn't clear, but all that code is just inindented
with no other change.
Test cases that hit qemuDomainAssignAddresses but don't have
infrastructure for specifying qemuCaps values see lots of
churn, since now PCI addresses are in the XML output.
hotplug APIs with the AFFECT_CONFIG flag are essentially replicating
'insert <device> into XML document, and redefine XML'. Thinking of
it this way, it's natural that we call virDomainDefPostParse after
manually editing the XML here.
Not only does doing so allow us to drop a bunch of open coded calls
to qemuDomainAssignAddresses, but it also means we are going through
the standard channels for XML validation and potentially catching
errors in user submitted XML.
This wires up qemuDomainAssignAddresses into the new
virDomainDefAssignAddressesCallback, so it's always triggered
via virDomainDefPostParse. We are essentially doing this already
with open coded calls sprinkled about.
qemu argv parse output changes slightly since previously it wasn't
hitting qemuDomainAssignAddresses.
When the <gic/> element in not present in the domain XML, use the
domain capabilities to figure out what GIC version is usable and
choose that one automatically.
This allows guests to be created on hardware that only supports
GIC v3 without having to update virt-manager and similar tools.
Keep using the default GIC version if the <gic/> element has been
added to the domain XML but no version has been specified, as not
to break existing guests.
We support omitting listen attribute of graphics element so we should
also support omitting address attribute of listen element. This patch
also updates libvirt to always add a listen element into domain XML
except for VNC graphics if socket attribute is specified.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
