Although virtio-scsi supports SCSI PR (Persistent Reservations),
the device on host may do not support it. To avoid losing data,
Just like PCI and USB pass through devices, only one live guest
is allowed per SCSI host pass through device."
Signed-off-by: Han Cheng <hanc.fnst@cn.fujitsu.com>
VFIO device assignment requires a cgroup ACL to be setup for access to
the /dev/vfio/nn "group" device for any devices that will be assigned
to a guest. In the case of a host device that is allocated from a
pool, it was being allocated during qemuBuildCommandLine(), which is
called by qemuProcessStart() *after* the all-encompassing
qemuSetupCgroup() was called, meaning that the standard Cgroup ACL
setup wasn't creating ACLs for these devices allocated from pools.
One possible solution was to manually add a single ACL down inside
qemuBuildCommandLine() when networkAllocateActualDevice() is called,
but that has two problems: 1) the function that adds the cgroup ACL
requires a virDomainObjPtr, which isn't available in
qemuBuildCommandLine(), and 2) we really shouldn't be doing network
device setup inside qemuBuildCommandLine() anyway.
Instead, I've created a new function called
qemuNetworkPrepareDevices() which is called just before
qemuPrepareHostDevices() during qemuProcessStart() (explanation of
ordering in the comments), i.e. well before the call to
qemuSetupCgroup(). To minimize code churn in a patch that will be
backported to 1.0.5-maint, qemuNetworkPrepareDevices only does
networkAllocateActualDevice() and the bare amount of setup required
for type='hostdev network devices, but it eventually should do *all*
device setup for guest network devices.
Note that some of the code that was previously needed in
qemuBuildCommandLine() is no longer required when
networkAllocateActualDevice() is called earlier:
* qemuAssignDeviceHostdevAlias() is already done further down in
qemuProcessStart().
* qemuPrepareHostdevPCIDevices() is called by
qemuPrepareHostDevices() which is called after
qemuNetworkPrepareDevices() in qemuProcessStart().
As hinted above, this new function should be moved into a separate
qemu_network.c (or similarly named) file along with
qemuPhysIfaceConnect(), qemuNetworkIfaceConnect(), and
qemuOpenVhostNet(), and expanded to call those functions as well, then
the nnets loop in qemuBuildCommandLine() should be reduced to only
build the commandline string (which itself can be in a separate
qemuInterfaceBuilldCommandLine() function as suggested by
Michal). However, this will require storing away an array of tapfd and
vhostfd that are needed for the commandline, so I would rather do that
in a separate patch and leave this patch at the minimum to fix the
bug.
The source code base needs to be adapted as well. Some files
include virutil.h just for the string related functions (here,
the include is substituted to match the new file), some include
virutil.h without any need (here, the include is removed), and
some require both.
Commit eca3fdf inadvertantly caused a failure to start for any domain
with the following in its config:
<graphics type='spice' autoport='yes'/>
The problem is that when tlsPort == 0 and defaultMode == "any" (which
is the default for defaultMode), this would be flagged in the code as
"needTLSPort", and if there was then no spice tls config, the new
error+fail would happen.
This patch checks for the case of defaultMode == "any", and in that
case simply doesn't allocate a TLS port (since that's probably not
what the user wanted, and it would have failed later anyway.). It does
leave the error in place for cases when the user specifically asked to
use tls in one way or another, though.
When a user requests auto-allocation of the spice TLS port but spice TLS
is disabled in qemu.conf, we start the machine and let qemu fail instead
of erroring out sooner.
Add an error message so that this doesn't happen.
These were previously being set in a custom hook function, but now
that virCommand directly supports setting them, we can eliminate that
part of the hook and call the APIs directly.
Ensure that the driver struct field names match the public
API names. For an API virXXXX we must have a driver struct
field xXXXX. ie strip the leading 'vir' and lowercase any
leading uppercase letters.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Historically QEMU/LXC guests have been placed in a cgroup layout
that is
$LOCATION-OF-LIBVIRTD/libvirt/{qemu,lxc}/$VMNAME
This is bad for a number of reasons
- The cgroup hierarchy gets very deep which seriously
impacts kernel performance due to cgroups scalability
limitations.
- It is hard to setup cgroup policies which apply across
services and virtual machines, since all VMs are underneath
the libvirtd service.
To address this the default cgroup location is changed to
be
/system/$VMNAME.{lxc,qemu}.libvirt
This puts virtual machines at the same level in the hierarchy
as system services, allowing consistent policy to be setup
across all of them.
This also honours the new resource partition location from the
XML configuration, for example
<resource>
<partition>/virtualmachines/production</partitions>
</resource>
will result in the VM being placed at
/virtualmachines/production/$VMNAME.{lxc,qemu}.libvirt
NB, with the exception of the default, /system, path which
is intended to always exist, libvirt will not attempt to
auto-create the partitions in the XML. It is the responsibility
of the admin/app to configure the partitions. Later libvirt
APIs will provide a way todo this.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Instead of calling virCgroupForDomain every time we need
the virCgrouPtr instance, just do it once at Vm startup
and cache a reference to the object in qemuDomainObjPrivatePtr
until shutdown of the VM. Removing the virCgroupPtr from
the QEMU driver state also means we don't have stale mount
info, if someone mounts the cgroups filesystem after libvirtd
has been started
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
To support "shareable" for volume type disk, we have to translate
the source before trying to add the shared disk entry. To achieve
the goal, this moves the helper qemuTranslateDiskSourcePool into
src/qemu/qemu_conf.c, and introduce an internal only member (voltype)
for struct _virDomainDiskSourcePoolDef, to record the underlying
volume type for use when building the drive string.
Later patch will support "shareable" volume type disk.
This patch is the result of running:
for i in $(git ls-files | grep -v html | grep -v \.po$ ); do
sed -i -e "s/virDomainXMLConf/virDomainXMLOption/g" -e "s/xmlconf/xmlopt/g" $i
done
and a few manual tweaks.
The VIR_ERR_NO_SUPPORT error code is reserved for cases where an
API is not implemented in a driver. It definitely should not be
used when an API execution fails due to unsupported operation.
Intend to reduce the redundant code,use virNumaSetupMemoryPolicy
to replace virLXCControllerSetupNUMAPolicy and
qemuProcessInitNumaMemoryPolicy.
This patch also moves the numa related codes to the
file virnuma.c and virnuma.h
Signed-off-by: Gao feng <gaofeng@cn.fujitsu.com>
qemuGetNumadAdvice will be used by LXC driver, rename
it to virNumaGetAutoPlacementAdvice and move it to virnuma.c
Signed-off-by: Gao feng <gaofeng@cn.fujitsu.com>
Commit 82d5fe5437
qemu: check backing chains even when cgroup is omitted
added backing file checks just before the code that removes optional
disks if they are not present. However, the backing chain code fails in
case the disk file does not exist, which makes qemuProcessStart fail
regardless on configured startupPolicy.
Note that startupPolicy implementation is still wrong after this patch
since it only check the first file in a possible chain. It should rather
check the complete backing chain. But this is an existing limitation
that can be solved later. After all, startupPolicy is most useful for
CDROM images and they won't make use of backing files in most cases.
This commit adds the QEMU driver support for CCW addresses. The
current QEMU only allows virtio devices to be attached to the
CCW bus. We named the new capability indicating that support
QEMU_CAPS_VIRTIO_CCW accordingly.
The fact that CCW devices can only be assigned to domains with a
machine type of s390-ccw-virtio requires a few extra checks for
machine type in qemu_command.c on top of querying
QEMU_CAPS_VIRTIO_{CCW|S390}.
The majority of the new functions deals with CCW address generation
and management.
Signed-off-by: Viktor Mihajlovski <mihajlov@linux.vnet.ibm.com>
The virCaps structure gathered a ton of irrelevant data over time that.
The original reason is that it was propagated to the XML parser
functions.
This patch aims to create a new data structure virDomainXMLConf that
will contain immutable data that are used by the XML parser. This will
allow two things we need:
1) Get rid of the stuff from virCaps
2) Allow us to add callbacks to check and add driver specific stuff
after domain XML is parsed.
This first attempt removes pointers to private data allocation functions
to this new structure and update all callers and function that require
them.
The current QEMU code for skipping log messages only skips over
'debug' message, switch to virLogProbablyLogMessage to make sure
it skips over all of them
This reverts the hack done in
commit 568a6cda27
Author: Jiri Denemark <jdenemar@redhat.com>
Date: Fri Feb 15 15:11:47 2013 +0100
qemu: Avoid deadlock in autodestroy
since we now have a fix which avoids the deadlock scenario
entirely
When the auto-destroy callback runs it is supposed to return
NULL if the virDomainObjPtr is no longer valid. It was not
doing this for transient guests, so we tried to virObjectUnlock
a mutex which had been freed. This often led to a crash.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Currently we call virSetDeviceUnprivSGIO with val == 0 if a block device
has an sgio attribute. But for sgio='filtered', we know that a
kernel with no unpriv_sgio support will always behave as the user
wanted. In this case, there is no need to call the function and
report a (bogus) error.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
We need to start NBD server and feed it with all non-<shared/>,
RW and source-full disks. Moreover, with new virPortAllocator we
must ensure the borrowed port for NBD server will be returned if
either migration completes or qemu process is torn down.
https://bugzilla.redhat.com/show_bug.cgi?id=896685 points out
a regression caused by commit 38c4a9c - libvirt only labels
the backing chain if the backing chain cache is populated, but
the code to populate the cache was only conditionally performed
if cgroup labeling was necessary.
* src/qemu/qemu_cgroup.c (qemuSetupCgroup): Hoist cache setup...
* src/qemu/qemu_process.c (qemuProcessStart): ...earlier into
caller, where it is now unconditional.
To avoid having to hold the qemu driver lock while iterating through
close callbacks and calling them. This fixes a real deadlock when a
domain which is being migrated from another host gets autodestoyed as a
result of broken connection to the other host.
The hash entry is changed from "ref" to {ref, @domains}. With this, the
caller can simply call qemuRemoveSharedDisk, without afraid of removing
the entry belongs to other domains. qemuProcessStart will obviously
benifit from it on error codepath (which calls qemuProcessStop to do
the cleanup).
Based on moving various checking into qemuAddSharedDisk, this
avoids the caller using it in wrong ways. Also this adds two
new checking for qemuCheckSharedDisk (disk device not 'lun'
and kernel doesn't support unpriv_sgio simply returns 0).
The qemu driver had been calling virSecurityManagerSetProcessLabel()
from a "pre-exec hook" function that is run after the child is forked,
but before exec'ing qemu. This is problematic because the uid and gid
of the child are set by the security driver, but capabilities are
dropped by virCommand - such separation doesn't work; the two
operations must be done together or the capabilities do not transfer
properly to the child process.
This patch switches to using virSecurityManagerSetChildProcessLabel(),
which is called prior to virCommandRun() (rather than being called
*during* virCommandrun() by the hook function), and doesn't set the
UID/GID/security label directly, but instead merely informs virCommand
what it should set them all to when the time is appropriate.
This lets virCommand choose to do the uid/gid and caps dropping all at
the same time if it wants (it does *want* to, but isn't doing so yet;
that's for an upcoming patch).
With the majority of fields in the virQEMUDriverPtr struct
now immutable or self-locking, there is no need for practically
any methods to be using the QEMU driver lock. Only a handful
of helper APIs in qemu_conf.c now need it
The hook scripts used by virCommand must be careful wrt
accessing any mutexes that may have been held by other
threads in the parent process. With the recent refactoring
there are 2 potential flaws lurking, which will become real
deadlock bugs once the global QEMU driver lock is removed.
Remove use of the QEMU driver lock from the hook function
by passing in the 'virQEMUDriverConfigPtr' instance directly.
Add functions to the virSecurityManager to be invoked before
and after fork, to ensure the mutex is held by the current
thread. This allows it to be safely used in the hook script
in the child process.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Currently the APIs for managing the shared disk list take
a virHashTablePtr as the primary argument. This is bad
because it requires the caller to deal with locking of
the QEMU driver. Switch the APIs to take the full
virQEMUDriverPtr instance
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The 'driver->caps' pointer can be changed on the fly. Accessing
it currently requires the global driver lock. Isolate this
access in a single helper, so a future patch can relax the
locking constraints.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
To avoid confusion between 'virCapsPtr' and 'qemuCapsPtr'
do some renaming of various fucntions/variables. All
instances of 'qemuCapsPtr' are renamed to 'qemuCaps'. To
avoid that clashing with the 'qemuCaps' typedef though,
rename the latter to virQEMUCaps.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
As a step towards making virDomainObjList thread-safe turn it
into an opaque virObject, preventing any direct access to its
internals.
As part of this a new method virDomainObjListForEach is
introduced to replace all existing usage of virHashForEach
Currently the virQEMUDriverPtr struct contains an wide variety
of data with varying access needs. Move all the static config
data into a dedicated virQEMUDriverConfigPtr object. The only
locking requirement is to hold the driver lock, while obtaining
an instance of virQEMUDriverConfigPtr. Once a reference is held
on the config object, it can be used completely lockless since
it is immutable.
NB, not all APIs correctly hold the driver lock while getting
a reference to the config object in this patch. This is safe
for now since the config is never updated on the fly. Later
patches will address this fully.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
This will allow storing additional topology data in the NUMA topology
definition.
This patch changes the storage type and fixes fallout of the change
across the drivers using it.
This patch also changes semantics of adding new NUMA cell information.
Until now the data were re-allocated and copied to the topology
definition. This patch changes the addition function to steal the
pointer to a pre-allocated structure to simplify the code.
https://bugzilla.redhat.com/show_bug.cgi?id=892079
One of my previous patches (f2a4e5f176) tried to fix crashing
libvirtd on domain detroy. However, we need to copy pattern from
qemuProcessHandleMonitorEOF() instead of decrementing reference
counter. The rationale for this is, if qemu process is dying due
to domain being destroyed, we obtain EOF on both the monitor and
agent sockets. However, if the exit is expected, qemuProcessStop
is called, which cleans both agent and monitor sockets up. We
want qemuAgentClose() to be called iff the EOF is not expected,
so we don't leak an FD and memory. Moreover, there could be race
with qemuProcessHandleMonitorEOF() which could have already
closed the agent socket, in which case we don't want to do
anything.
