If qemu supports multi function PCI device, the format of the PCI address passed
to qemu is "bus=pci.0,multifunction=on,addr=slot.function".
If qemu does not support multi function PCI device, the format of the PCI address
passed to qemu is "bus=pci.0,addr=slot".
Hot pluging/unpluging multi PCI device is not supported now. So the function
of hotplugged PCI device must be 0. When we hot unplug it, we should set release
all functions in the slot.
We save all used PCI address in the hash table. The key is generated by domain,
bus and slot now. We will support multi function PCI device, so the key should
be generated by domain, bus, slot and function.
We do not support to hot unplug multi function PCI device now. If the device is
one function of multi function PCI device, we shoul not allow to hot unplugg
it.
XenAPI session login can fail for a number of reasons, but currently no
specific
reason is displayed to the user, e.g.:
virsh -c XenAPI://citrix-xen.example.com/
Enter username for citrix-xen.example.com: root
Enter root's password for citrix-xen.example.com:
error: authentication failed: (null)
error: failed to connect to the hypervisor
This patch displays the session error description on failure.
Coverity complained about these intentional fallthrough cases, but
not about other cases that were explicitly marked with nice comments.
For some reason, Coverity doesn't seem smart enough to parse the
up-front English comment in virsh about intentional fallthrough :)
* tools/virsh.c (cmdVolSize): Mark fallthrough in a more typical
fashion.
* src/conf/nwfilter_conf.c (virNWFilterRuleDefDetailsFormat)
(virNWFilterRuleDetailsParse): Mark explicit fallthrough.
Detected by Coverity. The beginning of the function already filtered
out NULL objectContentList as invalid. Further investigation shows:
esxVI_RetrieveProperties is generated and returns a list of objects
that match the given propertyFilterSpec.
esxVI_LookupObjectContentByType then tests whether the result
corresponds to the expected occurrence and reports an error otherwise.
This simplifies the callers of esxVI_LookupObjectContentByType, but
due to the missing dereference the check was never performed because
the code thought that at least one item was obtained. NULL represents
an empty list. This is a potential segfault fix because callers of
esxVI_LookupObjectContentByType that specified "required" occurrence
assume *objectContentList to be non-NULL when
esxVI_LookupObjectContentByType succeeds.
* src/esx/esx_vi.c (esxVI_LookupObjectContentByType): Check
correct pointer.
Detected by Coverity. The only ways to get to the cleanup label
were by an early abort (list still unassigned) or after successfully
transferring list to dest, so there is no list to clean up.
* src/secret/secret_driver.c (loadSecrets): Kill dead code.
Detected by Coverity. All existing callers happen to be in
range, so this isn't too serious.
* src/qemu/qemu_cgroup.c (qemuCgroupControllerActive): Check
bounds before dereference.
Coverity already saw through a NULL dereference without these
annotations, and gcc is still too puny to do good NULL analysis.
But clang still benefits (and is easier to run than coverity),
not to mention that adding this bit of documentation to the code
may help future developers remember the constraints.
* src/util/uuid.h (virGetHostUUID, virUUIDFormat): Document
restrictions, for improved static analysis.
Detected by Coverity. Commit a98d8f0d tried to make uuid debugging
more robust, but missed some APIs. And on the APIs that it visited,
the mere act of preparing the debug message ends up dereferencing
uuid prior to the null check. Which means the APIs which are supposed
to gracefully reject NULL arguments now end up with SIGSEGV.
* src/libvirt.c (VIR_UUID_DEBUG): New macro.
(virDomainLookupByUUID, virDomainLookupByUUIDString)
(virNetworkLookupByUUID, virNetworkLookupByUUIDString)
(virStoragePoolLookupByUUID, virStoragePoolLookupByUUIDString)
(virSecretLookupByUUID, virSecretLookupByUUIDString)
(virNWFilterLookupByUUID, virNWFilterLookupByUUIDString): Avoid
null dereference.
Similar in nature to commit fd21ecfd, which shut up valgrind.
sigaction is apparently a nasty interface for analyzer tools,
at least for how many false positives it generates.
* src/util/command.c (virExecWithHook): Initialize entire var, since
coverity gripes about the (unused and non-standard) sa_restorer.
Detected by Coverity. The code was doing math on shifted unsigned
char (which promotes to int), then promoting that to unsigned long
during assignment to size. On 64-bit platforms, this risks sign
extending values of size > 2GiB. Bug present since commit
489fd3 (v0.6.0).
I'm not sure if a specially-crafted bogus qcow2 image could
exploit this, although it's probably not possible, since we
were already checking for the computed results being within
range of our fixed-size buffer.
* src/util/storage_file.c (qcowXGetBackingStore): Avoid sign
extension.
Add a simple handshake with the lxc_controller process so we can detect
process startup failures. We do this by adding a new --handshake cli arg
to lxc_controller for passing a file descriptor. If the process fails to
launch, we scrape all output from the logfile and report it to the user.
Seems reasonable to have all command wrappers in the same place
v2:
Dont move SetInherit
v3:
Comment spelling fix
Adjust WARN0 comment
Remove spurious #include movement
Don't include sys/types.h
Combine virExec enums
Signed-off-by: Cole Robinson <crobinso@redhat.com>
virGetVersion itself doesn't take a virConnectPtr, but in order to obtain
the hypervisor version against which libvirt was compiled it is used in
combination with virConnectGetType like this:
hvType = virConnectGetType(conn)
virGetVersion(&libVer, hvType, &typeVer)
When virConnectGetType is called on a remote connection then the remote
driver returns the type of the underlying driver on the server side, for
example QEMU. Then virGetVersion compares hvType to a set of strings that
depend on configure options and returns LIBVIR_VERSION_NUMBER in most
cases. Now this fails in case libvirt on the client side is just compiled
with the remote driver enabled only and the server side has the actual
driver such as the QEMU driver. It just happens to work when the actual
driver is enabled on client and server side. But that's not always true.
I noticed this on FreeBSD:
freebsd# virsh -c qemu+tcp://192.168.178.22/system version
Compiled against library: libvir 0.9.2
error: failed to get the library version
error: this function is not supported by the connection driver: virGetVersion
This is not FreeBSD specific, happens on Windows as well due to the
similar driver support configuration. The problem is that virConnectGetType
returns QEMU, but virGetVersion on the client side only accepts Remote
as hvType due to all other drivers being disabled on the client side.
Daniel P. Berrange suggested to get rid of all the conditional code in
virGetVersion, ignoring the hvType and always setting typeVer to
LIBVIR_VERSION_NUMBER. virConnectGetVersion is supposed to be used to
obtain the hypervisor version.
When peer-2-peer migration was invoked by a client supporting
v3, but where the target server only supported v2, we'd not
correctly shutdown the guest.
* src/qemu/qemu_migration.c: Ensure guest is shutdown in
v2 peer 2 peer migration
The v2 migration protocol doesn't use cookies, so we should not
be raising an error if the cookie parameters are NULL.
* src/qemu/qemu_migration.c: Don't raise error if cookie is NULL
The error code for virKillProcess is returned in the errno variable
not the return value. THis mistake caused the logs to be filled with
errors when shutting down QEMU processes
* src/qemu/qemu_process.c: Fix process kill check.
VirtualBox 4.0.8 changed the registry key layout. Before the version
number was in a Version key. Now the Version key contains %VER% and
the actual version number is in VersionExt now.
Move value lookup code into its own function: vboxLookupRegistryValue.
This commit is safe precisely because there has been no release
for any of the enum values being deleted (they were added post-0.9.1).
After the 0.9.2 release, we can then take advantage of
virDomainModificationImpact in more places.
* include/libvirt/libvirt.h.in (virDomainModificationImpact): New
enum.
(virDomainSchedParameterFlags, virMemoryParamFlags): Delete, since
these were never released, and the new enum works fine here.
* src/libvirt.c (virDomainGetMemoryParameters)
(virDomainSetMemoryParameters)
(virDomainGetSchedulerParametersFlags)
(virDomainSetSchedulerParametersFlags): Update documentation.
* src/qemu/qemu_driver.c (qemuDomainSetMemoryParameters)
(qemuDomainGetMemoryParameters, qemuSetSchedulerParametersFlags)
(qemuSetSchedulerParameters, qemuGetSchedulerParametersFlags)
(qemuGetSchedulerParameters): Adjust clients.
* tools/virsh.c (cmdSchedinfo, cmdMemtune): Likewise.
Based on ideas by Daniel Veillard and Hu Tao.
This fixes:
https://bugzilla.redhat.com/show_bug.cgi?id=702044https://bugzilla.redhat.com/show_bug.cgi?id=709454
Both of these complain of a failure to use an image file that resides
on a read-only NFS volume. The function in the DAC security driver
that chowns image files to the qemu user:group before using them
already has special cases to ignore failure of chown on read-only file
systems, and in a few other cases, but it hadn't been checking for
EINVAL, which is what is returned if the qemu user doesn't even exist
on the NFS server.
Since the explanation of EINVAL in the chown man page almost exactly
matches the log message already present for the case of EOPNOTSUPP,
I've just added EINVAL to that same conditional.
Coverity couldn't see that priv is NULL on failure. But on failure,
we might as well guarantee that callers don't try to free uninitialized
memory.
* src/remote/remote_driver.c (remoteGenericOpen): Even on failure,
pass priv back to caller.
Coverity complained that infd could be -1 at the point where it is
passed to write, when in reality, this code can only be reached if
infd is non-negative.
* src/util/command.c (virCommandProcessIO): Help out coverity.
Detected by Coverity. Bug introduced in 08106e2044 (unreleased).
* src/conf/domain_conf.c (virDomainChannelDefCheckABIStability):
Use correct sizeof operand.
Detected by Coverity. Introduced in commit aaf2b70, and turned into
a regression in the next few commits through 4e6e6672 (unreleased).
* src/conf/domain_event.c (virDomainEventStateFree): Free object,
per documentation.
Detected by Coverity. This leaked a cpumap on every iteration
of the loop. Leak introduced in commit 1cc4d02 (v0.9.0).
* src/qemu/qemu_process.c (qemuProcessSetVcpuAffinites): Plug
leak, and hoist allocation outside loop.
Spotted by coverity. Triggers on failed stat, although I'm not sure
how easy that condition is, so I'm not sure if this is a runtime
memory hog. Regression introduced in commit 8077d64 (unreleased).
* src/util/storage_file.c (virStorageFileGetMetadataFromFD):
Reduce need for malloc, avoiding a leak.
Coverity detected that options was being set by strdup but never
freed. But why even bother with an options variable? The options
parameter never changes! Leak present since commit 44948f5b (0.7.0).
This function could probably be rewritten to take better advantage
of virCommand, but that is more invasive.
* src/storage/storage_backend_fs.c
(virStorageBackendFileSystemMount): Avoid wasted strdup, and
guarantee proper cleanup on all paths.
Detected by Coverity. While it is possible on OOM condition, as
well as with bad code that passes binary == NULL, it is unlikely
to be encountered in the wild.
* src/util/command.c (virCommandNewArgList): Don't leak memory.
In v3 migration, once migration is completed, the VM needs
to be left in a paused state until after Finish3 has been
executed on the target. Only then will the VM be killed
off. When using non-JSON QEMU monitor though, we don't
receive any 'STOP' event from QEMU, so we need to manually
set our state offline & thus release lock manager leases.
It doesn't hurt to run this on the JSON case too, just in
case the event gets lost somehow
* src/qemu/qemu_migration.c: Explicitly set VM state to
paused when migration completes
The change 18c2a59206 caused
some regressions in behaviour of virDomainBlockStats
and virDomainBlockInfo in the QEMU driver.
The virDomainBlockInfo API stopped working for inactive
guests if querying a block device.
The virDomainBlockStats API did not promptly report
an error if the guest was not running in some cases.
* src/qemu/qemu_driver.c: Fix inactive guest handling
in BlockStats/Info APIs
The qemuAuditDisk calls in disk hotunplug operations were being
passed 'ret >= 0', but the code which sets ret to 0 was not yet
executed, and the error path had already jumped to the 'cleanup'
label. This meant hotunplug failures were never audited, and
hotunplug success was audited as a failure
* src/qemu/qemu_hotplug.c: Fix auditing of hotunplug
When virLockDriverAcquire is invoked during hotplug the state
parameter will be left as NULL.
* src/locking/lock_driver_nop.c,
src/locking/lock_driver_sanlock.c: Don't reference NULL state
parameter
Refactoring of the lock manager hotplug methods lost the
ret = 0 assignment for successful return path
* src/locking/domain_lock.c: Add missing ret = 0 assignments
Commit 4454a9efc7 introduced bad
behaviour on the VIR_EVENT_HANDLE_ERROR condition. This condition
is only hit when an invalid FD is used in poll() (typically due
to a double-close bug). The QEMU monitor code was treating this
condition as non-fatal, and thus libvirt would poll() in a fast
loop forever burning 100% CPU. VIR_EVENT_HANDLE_ERROR must be
handled in the same way as VIR_EVENT_HANDLE_HANGUP, killing the
QEMU instance.
* src/qemu/qemu_monitor.c: Treat VIR_EVENT_HANDLE_ERROR as EOF
In between fork and exec, a connection to sanlock is acquired
and the socket file descriptor is intionally leaked to the
child process. sanlock watches this FD for POLL_HANGUP to
detect when QEMU has exited. We don't want a rogus/compromised
QEMU from issuing sanlock RPC calls on the leaked FD though,
since that could be used to DOS other guests. By calling
sanlock_restrict() on the socket before exec() we can lock
it down.
* configure.ac: Check for sanlock_restrict API
* src/locking/domain_lock.c: Restrict lock acquired in
process startup phase
* src/locking/lock_driver.h: Add VIR_LOCK_MANAGER_ACQUIRE_RESTRICT
* src/locking/lock_driver_sanlock.c: Add call to sanlock_restrict
when requested by VIR_LOCK_MANAGER_ACQUIRE_RESTRICT flag
Based on the equivalent qemu driver code
* src/libxl/libxl_driver.c: refactor the Start save and restore
routines of the driver and adds the new entry points for
managed saves handling
Sanlock is a project that implements a disk-paxos locking
algorithm. This is suitable for cluster deployments with
shared storage.
* src/Makefile.am: Add dlopen plugin for sanlock
* src/locking/lock_driver_sanlock.c: Sanlock driver
* configure.ac: Check for sanlock
* libvirt.spec.in: Add a libvirt-lock-sanlock RPM
* src/conf/domain_conf.c, src/conf/domain_conf.h: APIs for
inserting/finding/removing virDomainLeaseDefPtr instances
* src/qemu/qemu_driver.c: Wire up hotplug/unplug for leases
* src/qemu/qemu_hotplug.h, src/qemu/qemu_hotplug.c: Support
for hotplug and unplug of leases
Some lock managers associate state with leases, allowing a process
to temporarily release its leases, and re-acquire them later, safe
in the knowledge that no other process has acquired + released the
leases in between.
This is already used between suspend/resume operations, and must
also be used across migration. This passes the lockstate in the
migration cookie. If the lock manager uses lockstate, then it
becomes compulsory to use the migration v3 protocol to get the
cookie support.
* src/qemu/qemu_driver.c: Validate that migration v2 protocol is
not used if lock manager needs state transfer
* src/qemu/qemu_migration.c: Transfer lock state in migration
cookie XML
The QEMU integrates with the lock manager instructure in a number
of key places
* During startup, a lock is acquired in between the fork & exec
* During startup, the libvirtd process acquires a lock before
setting file labelling
* During shutdown, the libvirtd process acquires a lock
before restoring file labelling
* During hotplug, unplug & media change the libvirtd process
holds a lock while setting/restoring labels
The main content lock is only ever held by the QEMU child process,
or libvirtd during VM shutdown. The rest of the operations only
require libvirtd to hold the metadata locks, relying on the active
QEMU still holding the content lock.
* src/qemu/qemu_conf.c, src/qemu/qemu_conf.h,
src/qemu/libvirtd_qemu.aug, src/qemu/test_libvirtd_qemu.aug:
Add config parameter for configuring lock managers
* src/qemu/qemu_driver.c: Add calls to the lock manager
To facilitate use of the locking plugins from hypervisor drivers,
introduce a higher level API for locking virDomainObjPtr instances.
In includes APIs targetted to VM startup, and hotplug/unplug
* src/Makefile.am: Add domain lock API
* src/locking/domain_lock.c, src/locking/domain_lock.h: High
level API for domain locking
To allow hypervisor drivers to assume that a lock driver impl
will be guaranteed to exist, provide a 'nop' impl that is
compiled into the library
* src/Makefile.am: Add nop driver
* src/locking/lock_driver_nop.c, src/locking/lock_driver_nop.h:
Nop lock driver implementation
* src/locking/lock_manager.c: Enable direct access of 'nop'
driver, instead of dlopen()ing it.
Define the basic framework lock manager plugins. The
basic plugin API for 3rd parties to implemented is
defined in
src/locking/lock_driver.h
This allows dlopen()able modules for alternative locking
schemes, however, we do not install the header. This
requires lock plugins to be in-tree allowing changing of
the lock manager plugin API in future.
The libvirt code for loading & calling into plugins
is in
src/locking/lock_manager.{c,h}
* include/libvirt/virterror.h, src/util/virterror.c: Add
VIR_FROM_LOCKING
* src/locking/lock_driver.h: API for lock driver plugins
to implement
* src/locking/lock_manager.c, src/locking/lock_manager.h:
Internal API for managing locking
* src/Makefile.am: Add locking code
A lock manager may operate in various modes. The direct mode of
operation is to obtain locks based on the resources associated
with devices in the XML. The indirect mode is where the app
creating the domain provides explicit leases for each resource
that needs to be locked. This XML extension allows for listing
resources in the XML
<devices>
...
<lease>
<lockspace>somearea</lockspace>
<key>thequickbrownfoxjumpsoverthelazydog</key>
<target path='/some/lease/path' offset='23432'/>
</lease>
...
</devices>
The 'lockspace' is a unique identifier for the lockspace which
the lease is associated
The 'key' is a unique identifier for the resource associated
with the lease.
The 'target' is the file on disk where the leases are held.
* docs/schemas/domain.rng: Add lease schema
* src/conf/domain_conf.c, src/conf/domain_conf.h: parsing and
formatting for leases
* tests/qemuxml2argvdata/qemuxml2argv-lease.args,
tests/qemuxml2argvdata/qemuxml2argv-lease.xml,
tests/qemuxml2xmltest.c: Test XML handling for leases
Allow the parent process to perform a bi-directional handshake
with the child process during fork/exec. The child process
will fork and do its initial setup. Immediately prior to the
exec(), it will stop & wait for a handshake from the parent
process. The parent process will spawn the child and wait
until the child reaches the handshake point. It will do
whatever extra setup work is required, before signalling the
child to continue.
The implementation of this is done using two pairs of blocking
pipes. The first pair is used to block the parent, until the
child writes a single byte. Then the second pair pair is used
to block the child, until the parent confirms with another
single byte.
* src/util/command.c, src/util/command.h,
src/libvirt_private.syms: Add APIs to perform a handshake
Regression introduced in commit d6623003 (v0.8.8) - using the
wrong sizeof operand meant that security manager private data
was overlaying the allowDiskFormatProbing member of struct
_virSecurityManager. This reopens disk probing, which was
supposed to be prevented by the solution to CVE-2010-2238.
* src/security/security_manager.c
(virSecurityManagerGetPrivateData): Use correct offset.
Commit 2d6adabd53 replaced qsorting disk
and controller devices with inserting them at the right position. That
was to fix unnecessary reordering of devices. However, when parsing
domain XML devices are just taken in the order in which they appear in
the XML since. Use the correct insertion algorithm to honor device
target.
Remove some special case code that took care of mapping hyper to the
correct C types.
As the list of procedures that is allowed to map hyper to long is fixed
put it in the generator instead annotations in the .x files. This
results in simpler .x file parsing code.
Use macros for hyper to long assignments that perform overflow checks
when long is smaller than hyper. Map hyper to long long by default.
Suggested by Eric Blake.
The gnutls_certificate_type_set_priority method is deprecated.
Since we already set the default gnutls priority, and do not
support OpenGPG credentials in any case, it was not serving
any useful purpose and can be removed
* src/remote/remote_driver.c: Remove src/remote/remote_driver.c
call
Convert openvzLocateConfFile to a replaceable function pointer to
allow testing the config file parsing without rewriting the whole
OpenVZ config parsing to a more testable structure.
Substitute VIR_ERR_NO_SUPPORT with VIR_ERR_INTERNAL_ERROR. Error
like following is not what user want to see.
error : pciDeviceIsAssignable:1487 : this function is not supported
by the connection driver: Device 0000:07:10.0 is behind a switch
lacking ACS and cannot be assigned
This function is also affected by getline conversion. But this
didn't result in a regression in general, because the difference
would only affect the behavior of the function when the line in
/proc/vz/vestat for the given vpsid wasn't found. Under normal
conditions this should not happen.
The regression fix in 3aab7f2d6b altered the error handling.
getline returns -1 on failure to read a line (including EOF). The
original openvzReadConfigParam function using openvz_readline only
treated EOF as not-found. The current getline version treats all
getline failures as not-found.
This patch fixes this and distinguishes EOF from other getline
failures.
Since directories can be used for <filesystem> passthrough, they are
basically storage volumes.
v2:
Skip ., .., lost+found dirs
v3:
Use gnulib last_component
v4:
Use gnulib "dirname.h", not system <dirname.h>
Don't skip lost+found
If spice graphics has no <channel> elements, the output graphics XML
is messed up. To prevent this, we need to end the <graphics> element
just before adding any compression selecting elements.
The virSysinfoIsEqual method was mistakenly inside a #ifndef WIN32
conditional.
The existing virSysinfoFormat is also stubbed out on Win32, even
though the code works without any trouble. This breaks XML output
on Win32, so the stub is removed.
virsh migrate mistakenly had some variables inside the conditional
* src/util/sysinfo.c: Build virSysinfoIsEqual on Win32 and remove
Win32 stub for virSysinfoFormat
* tools/virsh.c: Fix variable declaration on Win32
Update the qemuDomainMigrateBegin method so that it accepts
an optional incoming XML document. This will be validated
for ABI compatibility against the current domain config,
and if this check passes, will be passed back out for use
by the qemuDomainMigratePrepare method on the target
* src/qemu/qemu_domain.c, src/qemu/qemu_domain.h,
src/qemu/qemu_migration.c: Allow custom XML to be passed
To allow a client app to pass in custom XML during migration
of a guest it is neccessary to ensure the guest ABI remains
unchanged. The virDomainDefCheckABIStablity method accepts
two virDomainDefPtr structs and compares everything in them
that could impact the guest machine ABI
* src/conf/domain_conf.c, src/conf/domain_conf.h,
src/libvirt_private.syms: Add virDomainDefCheckABIStablity
* src/conf/cpu_conf.c, src/conf/cpu_conf.h: Add virCPUDefIsEqual
* src/util/sysinfo.c, src/util/sysinfo.h: Add virSysinfoIsEqual
The virDomainHostdevDef struct contains a 'char *target'
field. This is set to 'NULL' when parsing XML and never
used / set anywhere else. Clearly it is bogus & unused
* src/conf/domain_conf.c, src/conf/domain_conf.h: Remove
target from virDomainHostdevDef
This patch seperate the domain config loading just as qemu driver
does, first loading config of running or trasient domains, then
of persistent inactive domains. And only try to reconnect the
monitor of running domains, so that it won't always throws errors
saying can't connect to domain monitor.
And as "virDomainLoadConfig->virDomainAssignDef->virDomainObjAssignDef",
already do things like "vm->newDef = def", removed the codes
in "lxcReconnectVM" that does the same work.
Add support to set the maximum memory of the domain.
Also add support to change the memory of the current
state of the domain, which translates to a running
domain or the config of the domain.
Based on the code from the qemu driver.
v3:
* initialize xml pointer to avoid segfault
* throw error message if domain is paused as
libxenlight itself will pause it
v2:
* header is now padded and has a version field
* the correct restore function from libxl is used
* only create the restore event once in libxlVmStart
