When setting up mount namespace for a qemu domain the following
steps are executed:
1) get list of mountpoints under /dev/
2) move them to /var/run/libvirt/qemu/$domName.ext
3) start constructing new device tree under /var/run/libvirt/qemu/$domName.dev
4) move the mountpoint of the new device tree to /dev
5) restore original mountpoints from step 2)
Note the problem with this approach is that if some device in step
3) requires access to a mountpoint from step 2) it will fail as
the mountpoint is not there anymore. For instance consider the
following domain disk configuration:
<disk type='file' device='disk'>
<driver name='qemu' type='raw'/>
<source file='/dev/shm/vhostmd0'/>
<target dev='vdb' bus='virtio'/>
<address type='pci' domain='0x0000' bus='0x00' slot='0x0a' function='0x0'/>
</disk>
In this case operation fails as we are unable to create vhostmd0
in the new device tree because after step 2) there is no /dev/shm
anymore. Leave aside fact that we shouldn't try to create devices
living in other mountpoints. That's a separate bug that will be
addressed later.
Currently, the order described above is rearranged to:
1) get list of mountpoints under /dev/
2) start constructing new device tree under /var/run/libvirt/qemu/$domName.dev
3) move them to /var/run/libvirt/qemu/$domName.ext
4) move the mountpoint of the new device tree to /dev
5) restore original mountpoints from step 3)
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Cedric Bosdonnat <cbosdonnat@suse.com>
This is a USB3 controller and it's a better choice than piix3-uhci.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Acked-by: Andrea Bolognani <abologna@redhat.com>
The new logic will set the piix3-uhci if available regardless of
any architecture and it will be updated to better model based on
architecture and device existence.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Acked-by: Andrea Bolognani <abologna@redhat.com>
This patch maps /domain/cpu/cache element into -cpu parameters:
- <cache mode='passthrough'/> is translated to host-cache-info=on
- <cache level='3' mode='emulate'/> is transformed into l3-cache=on
- <cache mode='disable'/> is turned in host-cache-info=off,l3-cache=off
Any other <cache> element is forbidden.
The tricky part is detecting whether QEMU supports the CPU properties.
The 'host-cache-info' property is introduced in v2.4.0-1389-ge265e3e480,
earlier QEMU releases enabled host-cache-info by default and had no way
to disable it. If the property is present, it defaults to 'off' for any
QEMU until at least 2.9.0.
The 'l3-cache' property was introduced later by v2.7.0-200-g14c985cffa.
Earlier versions worked as if l3-cache=off was passed. For any QEMU
until at least 2.9.0 l3-cache is 'off' by default.
QEMU 2.9.0 was the first release which supports probing both properties
by running device-list-properties with typename=host-x86_64-cpu. Older
QEMU releases did not support device-list-properties command for CPU
devices. Thus we can't really rely on probing them and we can just use
query-cpu-model-expansion QMP command as a witness.
Because the cache property probing is only reliable for QEMU >= 2.9.0
when both are already supported for quite a few releases, we let QEMU
report an error if a specific cache mode is explicitly requested. The
other mode (or both if a user requested CPU cache to be disabled) is
explicitly turned off for QEMU >= 2.9.0 to avoid any surprises in case
the QEMU defaults change. Any older QEMU already turns them off so not
doing so explicitly does not make any harm.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Not all async jobs are visible via virDomainGetJobStats (either they are
too fast or getting the stats is not allowed during the job), but
forcing all of them to advertise the operation is easier than hunting
the jobs for which fetching statistics is allowed. And we won't need to
think about this when we add support for getting stats for more jobs.
https://bugzilla.redhat.com/show_bug.cgi?id=1441563
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
We are currently parsing only rx/frames/max because that's the only
value that makes sense for us. The tun device just added support for
this one and the others are only supported by hardware devices which
we don't need to worry about as the only way we'd pass those to the
domain is using <hostdev/> or <interface type='hostdev'/>. And in
those cases the guest can modify the settings itself.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
The history of USB controller for ppc64 guest is complex and goes
back to libvirt 1.3.1 where the fun started.
Prior Libvirt 1.3.1 if no model for USB controller was specified
we've simply passed "-usb" on QEMU command line.
Since Libvirt 1.3.1 there is a patch (8156493d8d) that fixes this
issue by using "-device pci-ohci,..." but it breaks migration with
older Libvirts which was agreed that's acceptable. However this
patch didn't reflect this change in the domain XML and the model
was still missing.
Since Libvirt 2.2.0 there is a patch (f55eaccb0c) that fixes the
issue with not setting the USB model into domain XML which we need
to know about to not break the migration and since the default
model was *pci-ohci* it was used as default in this patch as well.
This patch tries to take all the previous changes into account and
also change the default for newly defined domains that don't specify
any model for USB controller.
The VIR_DOMAIN_DEF_PARSE_ABI_UPDATE is set only if new domain is
defined or new device is added into a domain which means that in
all other cases we will use the old *pci-ohci* model instead of the
better and not broken *nec-usb-xhci* model.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1373184
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Introduce new wrapper functions without *Machine* in the function
name that take the whole virDomainDef structure as argument and
call the existing functions with *Machine* in the function name.
Change the arguments of existing functions to *machine* and *arch*
because they don't need the whole virDomainDef structure and they
could be used in places where we don't have virDomainDef.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
This way qemuDomainLogContextRef() and qemuDomainLogContextFree() is
no longer needed. The naming qemuDomainLogContextFree() was also
somewhat misleading. Additionally, it's easier to turn
qemuDomainLogContext in a self-locking object.
Signed-off-by: Marc Hartmayer <mhartmay@linux.vnet.ibm.com>
Reviewed-by: Bjoern Walk <bwalk@linux.vnet.ibm.com>
Depending on the architecture, requirements for ACPI and UEFI can
be different; more specifically, while on x86 UEFI requires ACPI,
on aarch64 it's the other way around.
Enforce these requirements when validating the domain, and make
the error message more accurate by mentioning that they're not
necessarily applicable to all architectures.
Several aarch64 test cases had to be tweaked because they would
have failed the validation step otherwise.
Currently, if we want to zero out disk source (e,g, due to
startupPolicy when starting up a domain) we use
virDomainDiskSetSource(disk, NULL). This works well for file
based storage (storage type file, dir, or block). But it doesn't
work at all for other types like volume and network.
So imagine that you have a domain that has a CDROM configured
which source is a volume from an inactive pool. Because it is
startupPolicy='optional', the CDROM is empty when the domain
starts. However, the source element is not cleared out in the
status XML and thus when the daemon restarts and tries to
reconnect to the domain it refreshes the disks (which fails - the
storage pool is still not running) and thus the domain is killed.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
When idx is 0 virStorageFileChainLookup returns the base (bottom) of the
backing chain rather than the top. This is expected by the callers of
qemuDomainGetStorageSourceByDevstr.
Add a special case for idx == 0
For guests that use <memoryBacking><locked>, our only option
is to remove the memory locking limit altogether.
Partially-resolves: https://bugzilla.redhat.com/1431793
Instead of having a separate function, we can simply return
zero from the existing qemuDomainGetMemLockLimitBytes() to
signal the caller that the memory locking limit doesn't need
to be set for the guest.
Having a single function instead of two makes it less likely
that we will use the wrong value, which is exactly what
happened when we started applying the limit that was meant
for VFIO-using guests to <memoryBacking><locked>-using
guests.
This reverts commit c2e60ad0e5.
Turns out this check is excessively strict: there are ways
other than <memtune><hard_limit> to raise the memory locking
limit for QEMU processes, one prominent example being
tweaking /etc/security/limits.conf.
Partially-resolves: https://bugzilla.redhat.com/1431793
Since mdevs are just another type of VFIO devices, we should increase
the memory locking limit the same way we do for VFIO PCI devices.
Signed-off-by: Erik Skultety <eskultet@redhat.com>
As goes for all the other hostdev device types, grant the qemu process
access to /dev/vfio/<mediated_device_iommu_group>.
Signed-off-by: Erik Skultety <eskultet@redhat.com>
A mediated device will be identified by a UUID (with 'model' now being
a mandatory <hostdev> attribute to represent the mediated device API) of
the user pre-created mediated device. We also need to make sure that if
user explicitly provides a guest address for a mdev device, the address
type will be matching the device API supported on that specific mediated
device and error out with an incorrect XML message.
The resulting device XML:
<devices>
<hostdev mode='subsystem' type='mdev' model='vfio-pci'>
<source>
<address uuid='c2177883-f1bb-47f0-914d-32a22e3a8804'>
</source>
</hostdev>
</devices>
Signed-off-by: Erik Skultety <eskultet@redhat.com>
The code is currently simple, but if we later add node names, it will be
necessary to generate the names based on the node name. Add a helper so
that there's a central point to fix once we add self-generated node
names.
If the migration flags indicate this migration will be using TLS,
then set up the destination during the prepare phase once the target
domain has been started to add the TLS objects to perform the migration.
This will create at least an "-object tls-creds-x509,endpoint=server,..."
for TLS credentials and potentially an "-object secret,..." to handle the
passphrase response to access the TLS credentials. The alias/id used for
the TLS objects will contain "libvirt_migrate".
Once the objects are created, the code will set the "tls-creds" and
"tls-hostname" migration parameters to signify usage of TLS.
During the Finish phase we'll be sure to attempt to clear the
migration parameters and delete those objects (whether or not they
were created). We'll also perform the same reset during recovery
if we've reached FINISH3.
If the migration isn't using TLS, then be sure to check if the
migration parameters exist and clear them if so.
So, majority of the code is just ready as-is. Well, with one
slight change: differentiate between dimm and nvdimm in places
like device alias generation, generating the command line and so
on.
Speaking of the command line, we also need to append 'nvdimm=on'
to the '-machine' argument so that the nvdimm feature is
advertised in the ACPI tables properly.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
NVDIMM is new type of memory introduced into QEMU 2.6. The idea
is that we have a Non-Volatile memory module that keeps the data
persistent across domain reboots.
At the domain XML level, we already have some representation of
'dimm' modules. Long story short, NVDIMM will utilize the
existing <memory/> element that lives under <devices/> by adding
a new attribute 'nvdimm' to the existing @model and introduce a
new <path/> element for <source/> while reusing other fields. The
resulting XML would appear as:
<memory model='nvdimm'>
<source>
<path>/tmp/nvdimm</path>
</source>
<target>
<size unit='KiB'>523264</size>
<node>0</node>
</target>
<address type='dimm' slot='0'/>
</memory>
So far, this is just a XML parser/formatter extension. QEMU
driver implementation is in the next commit.
For more info on NVDIMM visit the following web page:
http://pmem.io/
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=1431112
Yeah, that's right. A mount point doesn't have to be a directory.
It can be a file too. However, the code that tries to preserve
mount points under /dev for new namespace for qemu does not count
with that option.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=1430634
If a qemu process has died, we get EOF on its monitor. At this
point, since qemu process was the only one running in the
namespace kernel has already cleaned the namespace up. Any
attempt of ours to enter it has to fail.
This really happened in the bug linked above. We've tried to
attach a disk to qemu and while we were in the monitor talking to
qemu it just died. Therefore our code tried to do some roll back
(e.g. deny the device in cgroups again, restore labels, etc.).
However, during the roll back (esp. when restoring labels) we
still thought that domain has a namespace. So we used secdriver's
transactions. This failed as there is no namespace to enter.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
When libvirtd is started we call qemuDomainRecheckInternalPaths
to detect whether a domain has VNC socket path generated by libvirt
based on option from qemu.conf. However if we are parsing status XML
for running domain the existing socket path can be generated also if
the config XML uses the new <listen type='socket'/> element without
specifying any socket.
The current code doesn't make difference how the socket was generated
and always marks it as "fromConfig". We need to store the
"autoGenerated" value in the status XML in order to preserve that
information.
The difference between "fromConfig" and "autoGenerated" is important
for migration, because if the socket is based on "fromConfig" we don't
print it into the migratable XML and we assume that user has properly
configured qemu.conf on both hosts. However if the socket is based
on "autoGenerated" it means that a new feature was used and therefore
we need to leave the socket in migratable XML to make sure that if
this feature is not supported on destination the migration will fail.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Now that we have some qemuSecurity wrappers over
virSecurityManager APIs, lets make sure everybody sticks with
them. We have them for a reason and calling virSecurityManager
API directly instead of wrapper may lead into accidentally
labelling a file on the host instead of namespace.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
The functions in virCommand() after fork() must be careful with regard
to accessing any mutexes that may have been locked by other threads in
the parent process. It is possible that another thread in the parent
process holds the lock for the virQEMUDriver while fork() is called.
This leads to a deadlock in the child process when
'virQEMUDriverGetConfig(driver)' is called and therefore the handshake
never completes between the child and the parent process. Ultimately
the virDomainObjectPtr will never be unlocked.
It gets much worse if the other thread of the parent process, that
holds the lock for the virQEMUDriver, tries to lock the already locked
virDomainObject. This leads to a completely unresponsive libvirtd.
It's possible to reproduce this case with calling 'virsh start XXX'
and 'virsh managedsave XXX' in a tight loop for multiple domains.
This commit fixes the deadlock in the same way as it is described in
commit 61b52d2e38.
Signed-off-by: Marc Hartmayer <mhartmay@linux.vnet.ibm.com>
Reviewed-by: Boris Fiuczynski <fiuczy@linux.vnet.ibm.com>
When enabling virgl, qemu opens /dev/dri/render*. So far, we are
not allowing that in devices CGroup nor creating the file in
domain's namespace and thus requiring users to set the paths in
qemu.conf. This, however, is suboptimal as it allows access to
ALL qemu processes even those which don't have virgl configured.
Now that we have a way to specify render node that qemu will use
we can be more cautious and enable just that.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
So far, qemuDomainGetHostdevPath has no knowledge of the reasong
it is called and thus reports /dev/vfio/vfio for every VFIO
backed device. This is suboptimal, as we want it to:
a) report /dev/vfio/vfio on every addition or domain startup
b) report /dev/vfio/vfio only on last VFIO device being unplugged
If a domain is being stopped then namespace and CGroup die with
it so no need to worry about that. I mean, even when a domain
that's exiting has more than one VFIO devices assigned to it,
this function does not clean /dev/vfio/vfio in CGroup nor in the
namespace. But that doesn't matter.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
So far, we are allowing /dev/vfio/vfio in the devices cgroup
unconditionally (and creating it in the namespace too). Even if
domain has no hostdev assignment configured. This is potential
security hole. Therefore, when starting the domain (or
hotplugging a hostdev) create & allow /dev/vfio/vfio too (if
needed).
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Since these two functions are nearly identical (with
qemuSetupHostdevCgroup actually calling virCgroupAllowDevicePath)
we can have one function call the other and thus de-duplicate
some code.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
The bare fact that mnt namespace is available is not enough for
us to allow/enable qemu namespaces feature. There are other
requirements: we must copy all the ACL & SELinux labels otherwise
we might grant access that is administratively forbidden or vice
versa.
At the same time, the check for namespace prerequisites is moved
from domain startup time to qemu.conf parser as it doesn't make
much sense to allow users to start misconfigured libvirt just to
find out they can't start a single domain.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
mknod() is affected my the current umask, so we're not
guaranteed the newly-created device node will have the
right permissions.
Call chmod(), which is not affected by the current umask,
immediately afterwards to solve the issue.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1421036
Just like we need wrappers over other virSecurityManager APIs, we
need one for virSecurityManagerSetImageLabel and
virSecurityManagerRestoreImageLabel. Otherwise we might end up
relabelling device in wrong namespace.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Firstly, instead of checking for next->path the
virStorageSourceIsEmpty() function should be used which also
takes disk type into account.
Secondly, not every disk source passed has the correct type set
(due to our laziness). Therefore, instead of checking for
virStorageSourceIsBlockLocal() and also S_ISBLK() the former can
be refined to just virStorageSourceIsLocalStorage().
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Again, one missed bit. This time without this commit there is no
/dev entry in the namespace of the qemu process when doing disk
snapshots or block-copy.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
These functions do not need to see the whole virDomainDiskDef.
Moreover, they are going to be called from places where we don't
have access to the full disk definition. Sticking with
virStorageSource is more than enough.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
