The initpid will be required long term to enable LXC to
implement various hotplug operations. Thus it needs to be
persisted in the domain status XML. LXC has not used the
domain status XML before, so this introduces use of the
helpers.
Currently the lxcContainerSetupMounts method uses the
virSecurityManagerPtr instance to obtain the mount options
string and then only passes the string down into methods
it calls. As functionality in LXC grows though, those
methods need to have direct access to the virSecurityManagerPtr
instance. So push the code down a level.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The impls of virSecurityManagerGetMountOptions had no way to
return errors, since the code was treating 'NULL' as a success
value. This is somewhat pointless, since the calling code did
not want NULL in the first place and has to translate it into
the empty string "". So change the code so that the impls can
return "" directly, allowing use of NULL for error reporting
once again
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=876828
Commit 38c4a9cc introduced a regression in hot unplugging of disks
from qemu, where cgroup device ACLs were no longer being revoked
(thankfully not a security hole: cgroup ACLs only prevent open()
of the disk; so reverting the ACL prevents future abuse but doesn't
stop abuse from an fd that was already opened before the ACL change).
The actual regression is due to a latent bug. The hot unplug code
was computing the set of files needing cgroup ACL revocation based
on the XML passed in by the user, rather than based on the domain's
details on which disk was being deleted. As long as the revoke
path was always recomputing the backing chain, this didn't really
matter; but now that we want to compute the chain exactly once and
remember that computation, we need to hang on to the backing chain
until after the revoke has happened.
* src/qemu/qemu_hotplug.c (qemuDomainDetachPciDiskDevice):
Transfer backing chain before deletion.
This patch introduces the RNG schema and updates necessary data strucutures
to allow various hypervisors to make use of Gluster protocol as one of the
supported network disk backend. Next patch will add support to make use of
this feature in Qemu since it now supports Gluster protocol as one of the
network based storage backend.
Two new optional attributes for <host> element are introduced - 'transport'
and 'socket'. Valid transport values are tcp, unix or rdma. If none specified,
tcp is assumed. If transport is unix, socket specifies path to unix socket.
This patch allows users to specify disks on gluster backends like this:
<disk type='network' device='disk'>
<driver name='qemu' type='raw'/>
<source protocol='gluster' name='Volume1/image'>
<host name='example.org' port='6000' transport='tcp'/>
</source>
<target dev='vda' bus='virtio'/>
</disk>
<disk type='network' device='disk'>
<driver name='qemu' type='raw'/>
<source protocol='gluster' name='Volume2/image'>
<host transport='unix' socket='/path/to/sock'/>
</source>
<target dev='vdb' bus='virtio'/>
</disk>
Signed-off-by: Harsh Prateek Bora <harsh@linux.vnet.ibm.com>
Although we require various C99 features, we don't yet require a
complete C99 compiler. On RHEL 5, compilation complained:
qemu/qemu_command.c: In function 'qemuBuildGraphicsCommandLine':
qemu/qemu_command.c:4688: error: 'for' loop initial declaration used outside C99 mode
* src/qemu/qemu_command.c (qemuBuildGraphicsCommandLine): Declare
variable sooner.
* src/qemu/qemu_process.c (qemuProcessInitPasswords): Likewise.
The patch refactors the current ESX storage driver due to following reasons:
1. Given most of the public APIs exposed by the storage driver in Libvirt
remains same, ESX storage driver should not implement logic specific
for only one supported format (current implementation only supports VMFS).
2. Decoupling interface from specific storage implementation gives us an
extensible design to hook implementation for other supported storage
formats.
This patch refactors the current driver to implement it as a facade pattern i.e.
the driver exposes all the public libvirt APIs, but uses backend drivers to get
the required task done. The backend drivers provide implementation specific to
the type of storage device.
File changes:
------------------
esx_storage_driver.c ----> esx_storage_driver.c (base storage driver)
|
|---> esx_storage_backend_vmfs.c (VMFS backend)
When no security driver is specified libvirt_lxc segfaults as a debug
message tries to access security labels for the container that are not
present.
This problem was introduced in commit 6c3cf57d6c.
Early jumps to the cleanup label caused a crash of the libvirt_lxc
container helper as the cleanup section called
virLXCControllerDeleteInterfaces(ctrl) without checking the ctrl argument
for NULL. The argument was de-referenced soon after.
$ /usr/libexec/libvirt_lxc
/usr/libexec/libvirt_lxc: missing --name argument for configuration
Segmentation fault
This will simplify the refactoring of the ESX storage driver to support
a VMFS and an iSCSI backend.
One of the tasks the storage driver needs to do is to decide which backend
driver needs to be invoked for a given request. This approach extends
virStoragePool and virStorageVol to store extra parameters:
1. privateData: stores pointer to respective backend storage driver.
2. privateDataFreeFunc: stores cleanup function pointer.
virGetStoragePool and virGetStorageVol are modfied to accept these extra
parameters as user params. virStoragePoolDispose and virStorageVolDispose
checks for cleanup operation if available.
The private data pointer allows the ESX storage driver to store a pointer
to the used backend with each storage pool and volume. This avoids the need
to detect the correct backend in each storage driver function call.
The new model supports following features in addition to those supported
by SandyBridge:
fma, pcid, movbe, fsgsbase, bmi1, hle, avx2, smep, bmi2, erms, invpcid,
rtm
Commit 258e06c removed setting of the volume type to
VIR_STORAGE_VOL_BLOCK, which leads to failures in
storageVolumeCreateXMLFrom.
The type (and target.format) of the volume was set to zero. In
virStorageBackendGetBuildVolFromFunction, this gets interpreted as
VIR_STORAGE_FILE_NONE and the qemu-img tool is called with unknown
"none" format.
Bug: https://bugzilla.redhat.com/show_bug.cgi?id=879780
bridge_driver.h: silence gcc warnings:
statement with no effect [-Wunused-value]
unused variable 'net' [-Wunused-variable]
virdrivermoduletest.c: don't require network driver module
if it hasn't been built.
The virLXCControllerClientCloseHook method was mistakenly
assuming that the private data associated with the network
client was the virLXCControllerPtr. In fact it was just a
dummy int, so we were derefencing a bogus struct. The
frequent result of this was that we would never quit, because
we tried to arm a non-existant timer.
Fix the code by removing the dummy private data and just
using the virLXCControllerPtr instance as private data
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
It is possible for there to be deleted timers when we
calculate the next timeout, and they must be skipped.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The event code is a no-op if requested to update a non-existent
timer/handle watch. This makes it hard to detect bugs in the
caller who have passed bogus data. Add a VIR_WARN output in
such cases, since the API does not allow for return errors.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The docs for virDiskNameToIndex claim it ignores partition
numbers. In actual fact though, a code ordering bug means
that a partition number will cause the code to accidentally
multiply the result by 26.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Commit e0c469e58b that fixes the detection
of image chain wasn't complete. Iteration through the backing image
chain has to stop at the last existing image if some of the images are
missing otherwise the backing chain that is cached contains entries with
paths being set to NULL resulting to:
error: Unable to allow access for disk path (null): Bad address
Fortunately stat() is kind enough not to crash when it's presented with
a NULL argument. At least on Linux.
The error "... but the cause is unknown" appeared for XMLs similar to
this:
<disk type='file' device='cdrom'>
<driver name='qemu' type='raw'/>
<source file='/dev/zero'/>
<target dev='sr0'/>
</disk>
Notice unsupported disk type (for the driver), but also no address
specified. The first part is not a problem and we should not abort
immediately because of that, but the combination with the address
unknown was causing an unspecified error.
While fixing this, I added an error to one place where this return
value was not managed properly.
Fixes this error when building with -Werror on Alpine Linux:
util/processinfo.c: In function 'virProcessInfoSetAffinity':
util/processinfo.c:52:5: error: implicit declaration of function 'malloc' [-Werror=implicit-function-declaration]
Signed-off-by: Natanael Copa <ncopa@alpinelinux.org>
Currently the LXC driver logs audit messages when a container
is started or stopped. These audit messages, however, contain
the PID of the libvirt_lxc supervisor process. To enable
sysadmins to correlate with audit messages generated by
processes /inside/ the container, we need to include the
container init process PID.
We can't do this in the main 'start' audit message, since
the init PID is not available at that point. Instead we output
a completely new audit record, that lists both PIDs.
type=VIRT_CONTROL msg=audit(1353433750.071:363): pid=20180 uid=0 auid=501 ses=3 subj=unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023 msg='virt=lxc op=init vm="busy" uuid=dda7b947-0846-1759-2873-0f375df7d7eb vm-pid=20371 init-pid=20372 exe="/home/berrange/src/virt/libvirt/daemon/.libs/lt-libvirtd" hostname=? addr=? terminal=pts/6 res=success'
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The LXC controller code currently directly invokes the
libvirt main loop code. The problem is that this misses
the cleanup of virNetServerClient connections that
virNetServerRun takes care of.
The result is that when libvirtd is stopped, the
libvirt_lxc controller process gets stuck in a I/O loop.
When libvirtd is then started again, it fails to connect
to the controller and thus kills off the entire domain.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Regression introduced by commit 258e06c85b, "ret" could be set to 1
or 0 by virStorageBackendFileSystemIsMounted before goto cleanup.
This could mislead the callers (up to the public API
virStoragePoolDestroy) to return success even the underlying umount
command fails.
I have been testing libvirt v1.0.0 for deployment within my
organization, and in the process discovered what appears to be a bug
that breaks virsh attach-device, when attaching an RBD volume to an
instance. First, here is the error presented, with v1.0.0 (this worked
in v0.10.2):
[root@host ~]# virsh attach-device W5APQ8 G84VV1.xml
error: Failed to attach device from G84VV1.xml
error: cannot open file 'dc3-1-test/G84VV1': No such file or directory
Using git bisect, I narrowed the problem down to this as the first
commit to break this setup:
4d34c92947 is the first bad commit
Commit a4c19459aa only added the
QEMU capability flag, command line option and added the boot element
for redirdev's in the XML schema.
This patch adds support for parsing and writing the XML with redirdevs
with the boot flag. It also ignores unknown XML elements in redirdev
instead of failing with:
"error: An error occurred, but the cause is unknown"
Bug: https://bugzilla.redhat.com/show_bug.cgi?id=805414
Upcoming patches for revert-and-clone branching of snapshots need
to be able to copy a domain definition; make this step reusable.
* src/conf/domain_conf.h (virDomainDefCopy): New prototype.
* src/conf/domain_conf.c (virDomainObjCopyPersistentDef): Split...
(virDomainDefCopy): ...into new function.
(virDomainObjSetDefTransient): Use it.
* src/libvirt_private.syms (domain_conf.h): Export it.
* src/qemu/qemu_driver.c (qemuDomainRevertToSnapshot): Use it.
Relatively straight-forward. And since qemu was already using
VIR_DOMAIN_SNAPSHOT_FILTERS_ALL, with 6 different APIs all calling
into this common code, I've instantly added all 5 flags to 6 APIs.
* src/conf/snapshot_conf.h (VIR_DOMAIN_SNAPSHOT_FILTERS_ALL):
Enable new filters.
* src/conf/snapshot_conf.c (virDomainSnapshotObjListGetNames):
Prep the new flags.
(virDomainSnapshotObjListCopyNames): Actually do the filtering.
As we enable more modes of snapshot creation, it becomes more important
to be able to quickly filter based on snapshot properties. This patch
introduces new filter flags; subsequent patches will introduce virsh
back-compat filtering, as well as actual libvirt filtering.
* include/libvirt/libvirt.h.in (virDomainSnapshotListFlags): Add
five new flags in two new groups.
* src/libvirt.c (virDomainSnapshotNum, virDomainSnapshotListNames)
(virDomainListAllSnapshots, virDomainSnapshotNumChildren)
(virDomainSnapshotListChildrenNames)
(virDomainSnapshotListAllChildren): Document them.
* src/conf/snapshot_conf.h (VIR_DOMAIN_SNAPSHOT_FILTERS_STATUS)
(VIR_DOMAIN_SNAPSHOT_FILTERS_LOCATION): Add new convenience filter
collection macros.
* tools/virsh-snapshot.c (cmdSnapshotList): Add 5 new flags.
* tools/virsh.pod (snapshot-list): Document them.
This resolves:
https://bugzilla.redhat.com/show_bug.cgi?id=873134
The reported problem is that an attempt to restore a saved domain that
was configured with <currentMemory> and <memory> set to some (same for
both) number that's not a multiple of 4096KiB results in an error like
this:
error: Failed to start domain libvirt_test_api
error: XML error: current memory '4001792k' exceeds maximum '4000768k'
(in this case, currentMemory was set to 4000000KiB).
The reason for this failure is:
1) a saved image contains the "live xml" of the domain at the time of
the save.
2) the live xml of a running domain gets its currentMemory
(a.k.a. cur_balloon) directly from the qemu monitor rather than from
the configuration of the domain.
3) the value reported by qemu is (sometimes) not exactly what was
originally given to qemu when the domain was started, but is rounded
up to [some indeterminate granularity] - in some versions of qemu that
granularity is apparently 1MiB, and in others it is 4MiB.
4) When the XML is parsed to setup the state of the restored domain,
the XML parser for <currentMemory> compares it to <memory> (which is
the maximum allowed memory size for the domain) and if <currentMemory>
is greater than the next 1024KiB boundary above <memory>, it spits out
an error and fails.
For example (from the BZ) if you start qemu on RHEL6 with both
<currentMemory> and <memory> of 4000000 (this number is in KiB),
libvirt's dominfo or dumpxml will report "4001792" back (rounded up to
next 4MiB) for 10-20 seconds after the start, then revert to reporting
"4000000". On Fedora 16 (which uses qemu-1.0), it will instead report
"4000768" (rounded up to next 1MiB). On Fedora 17 (qemu-1.2), it seems
to always report "4000000". ("4000000" is of course okay, and
"4000768" is also okay since that's the next 1024KiB boundary above
"4000000" and the parser was already allowing for that. But "4001792
is *not* okay and produces the error message.)
This patch solves the problem by changing the allowed "fudge factor"
when parsing from 1024KiB to 4096KiB to match the maximum up-rounding
that could be done in qemu.
(I had earlier thought to fix this by up-rounding <memory> in the
dumpxml that's put into the saved image, but that wouldn't have fixed
the case where the save image was produced by an "unfixed"
libvirtd.)
Prior to this patch, 'virsh nodecpumap' on older kernels reported:
error: Unable to get cpu map
error: out of memory
* src/nodeinfo.c (linuxParseCPUmax): Don't overwrite error.
(nodeGetCPUBitmap): Provide backup implementation.
On RHEL 5, I was getting a segfault trying to start libvirtd,
because we were failing virNodeParseSocket but not checking
for errors, and then calling CPU_SET(-1, &sock_map) as a result.
But if you don't have a topology/physical_package_id file,
then you can just assume that the cpu belongs to socket 0.
* src/nodeinfo.c (virNodeGetCpuValue): Change bool into
default_value.
(virNodeParseSocket): Allow for default value when file is missing,
different from fatal error on reading file.
(virNodeParseNode): Update call sites to fail on error.
For disk snapshots, the user could request an external snapshot
but not supply a filename; later on, we would check this condition
and generate a suitable name if possible, or gracefully error out
when not possible (such as when the original file was a block
device). But unless we come up with a suitable way to generate
external memory file names, we have no later code point that was
checking for NULL, so we should forbid this up front.
* src/conf/snapshot_conf.c (virDomainSnapshotDefParseString):
Avoid NULL deref, since we don't generate names yet.
It may take some time for sanlock to add a lockspace. And if user
restart libvirtd service meanwhile, the fresh daemon can fail adding
the same lockspace with EINPROGRESS. Recent sanlock has
sanlock_inq_lockspace() function which should block until lockspace
changes state. If we are building against older sanlock we should
retry a few times before claiming an error. This issue can be easily
reproduced:
for i in {1..1000} ; do echo $i; service libvirtd restart; sleep 2; done
20
Stopping libvirtd daemon: [FAILED]
Starting libvirtd daemon: [ OK ]
21
Stopping libvirtd daemon: [ OK ]
Starting libvirtd daemon: [ OK ]
22
Stopping libvirtd daemon: [ OK ]
Starting libvirtd daemon: [ OK ]
error : virLockManagerSanlockSetupLockspace:334 : Unable to add
lockspace /var/lib/libvirt/sanlock/__LIBVIRT__DISKS__: Operation now in
progress
Since /sys/devices/system/cpu/present is not available on
older kernels like on RHEL 5.x nodeGetCPUCount will
fail there. The fallback implemented is to scan for
/sys/devices/system/cpu/cpuNN entries.
Signed-off-by: Viktor Mihajlovski <mihajlov@linux.vnet.ibm.com>
This simplifies the top-level code, at the cost of using a little more
stack space. The primary benefit is being able to send more fields
without knowing in advance how many of them, and of which types, these
fields will be, and without having to individually add buffer variables.
The code imposes an upper limit on the total number of iovs/buffers
used, and fields that wouldn't fit are silently dropped. This is not
significant in this patch, but will affect the following one.
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
... and update all users. No change in functionality, the parameter
will be used later.
The metadata representation is as minimal as possible, but requires
the caller to allocate an array on stack explicitly.
The alternative of using varargs in the virLogMessage() callers:
* Would not allow the caller to optionally omit some metadata elements,
except by having two calls to virLogMessage.
* Would not be as type-safe (e.g. using int vs. size_t), and the compiler
wouldn't be able to do type checking
* Depending on parameter order:
a) virLogMessage(..., message format, message params...,
metadata..., NULL)
can not be portably implemented (parse_printf_format() is a glibc
function)
b) virLogMessage(..., metadata..., NULL,
message format, message params...)
would prevent usage of ATTRIBUTE_FMT_PRINTF and the associated
compiler checking.
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
The "restart" function for locks allocates a new array according to
and pre-sets its length, then reads the owner pids from a JSON
document in a loop. Rather than adding each owner at a different
index, though, it repeatedly overwrites the last element of the array
with all the owners.
This patch adds a helper to determine if snapshots are external and uses
the helper to fix detection of those in snapshot deletion code.
Snapshots are external if they have an external memory image or if the
disk locations are external. As mixed snapshots are forbidden for now
we need to check just one disk to know.
Lately there were a few reports of the output of the virsh nodeinfo
command being inaccurate. This patch tries to avoid that by checking if
the topology actually makes sense. If it doesn't we then report a
synthetic topology that indicates to the user that the host capabilities
should be checked for the actual topology.
Currently, if user calls virDomainAbortJob we just issue
'migrate_cancel' and hope for the best. However, if user calls
the API in wrong phase when migration hasn't been started yet
(perform phase) the cancel request is just ignored. With this
patch, the request is remembered and as soon as perform phase
starts, migration is cancelled.
For S390, the default console target type cannot be of type 'serial'.
It is necessary to at least interpret the 'arch' attribute
value of the os/type element to produce the correct default type.
Therefore we need to extend the signature of defaultConsoleTargetType
to account for architecture. As a consequence all the drivers
supporting this capability function must be updated.
Despite the amount of changed files, the only change in behavior is
that for S390 the default console target type will be 'virtio'.
N.B.: A more future-proof approach could be to to use hypervisor
specific capabilities to determine the best possible console type.
For instance one could add an opaque private data pointer to the
virCaps structure (in case of QEMU to hold capsCache) which could
then be passed to the defaultConsoleTargetType callback to determine
the console target type.
Seems to be however a bit overengineered for the use case...
Signed-off-by: Viktor Mihajlovski <mihajlov@linux.vnet.ibm.com>
When the libvirt daemon is restarted it tries to reconnect to running
qemu domains. Since commit d38897a5d4 the
re-connection code runs in separate threads. In the original
implementation the maximum of domain ID's (that is used as an
initializer for numbering guests created next) while libvirt was
reconnecting to the guest.
With the threaded implementation this opens a possibility for race
conditions with the thread that is autostarting guests. When there's a
guest running with id 1 and the daemon is restarted. The autostart code
is reached first and spawns the first guest that should be autostarted
as id 1. This results into the following unwanted situation:
# virsh list
Id Name State
----------------------------------------------------
1 guest1 running
1 guest2 running
This patch extracts the detection code before the re-connection threads
are started so that the maximum id of the guests being reconnected to is
known.
The only semantic change created by this is if the guest with greatest ID
quits before we are able to reconnect it's ID is used anyway as the
greatest one as without this patch the greatest ID of a process we could
successfuly reconnect to would be used.
82507838 refactored the code to keep both the raw and canonicalized form
of the backingStore, which breaks badly when the storage pool contains a
storage volume, which is missing its backing store file:
# ./daemon/libvirtd -l
2012-11-07 12:43:33.279+0000: 22175: info : libvirt version: 1.0.0
2012-11-07 12:43:33.279+0000: 22175: error : absolutePathFromBaseFile:542 : Can't canonicalize path '/var/lib/libvirt/images/base.qcow2': No such file or directory
2012-11-07 12:43:33.280+0000: 22175: error : storageDriverAutostart:115 : Failed to autostart storage pool 'default': Can't canonicalize path '/var/lib/libvirt/images/base.qcow2': No such file or directory
This is because virStorageFileGetMetadataFromBuf() aborts with -1 if the
filename of the backingStore can not be canonicalized:
#0 absolutePathFromBaseFile () at util/storage_file.c:541
#1 virStorageFileGetMetadataFromBuf () at util/storage_file.c:728
#2 virStorageFileGetMetadataFromFD () at util/storage_file.c:932
#3 virStorageBackendProbeTarget () at storage/storage_backend_fs.c:94
#4 virStorageBackendFileSystemRefresh () at storage/storage_backend_fs.c:849
#5 storagePoolStart () at storage/storage_driver.c:700
#6 virStoragePoolCreate () at libvirt.c:12471
...
Treat files which miss their backing file as standalone files.
Signed-off-by: Philipp Hahn <hahn@univention.de>
This patch adds support for external disk snapshots of inactive domains.
The snapshot is created by calling using qemu-img by calling:
qemu-img create -f format_of_snapshot -o
backing_file=/path/to/src,backing_fmt=format_of_backing_image
/path/to/snapshot
in case the backing image format is known or probing is allowed and
otherwise:
qemu-img create -f format_of_snapshot -o backing_file=/path/to/src
/path/to/snapshot
on each of the disks selected for snapshotting. This patch also modifies
the snapshot preparing function to support creating external snapshots
and to sanitize arguments. For now the user isn't able to mix external
and internal snapshots but this restriction might be lifted in the
future.
Some operations, APIs needs domain to be paused prior operation can be
performed, e.g. (managed-) save of a domain. The processors should be
restored in the end. However, if 'cont' fails for some reason, we log a
message but this is not sufficient as an event should be emitted as
well. Mgmt application can then decide what to do.
The code that was split out into the qemuDomainSaveMemory expands the
pointer containing the XML description of the domain that it gets from
higher layers. If the pointer changes the old one is invalid and the
upper layer function tries to free it causing an abort.
This patch changes the expansion of the original string to a new
allocation and copy of the contents.
After the connection to ESX 5.1 being broken since g1e7cd39, the fix
in bab7752c helped a bit, but still missed a spot, so the connection
is now successful, but some APIs (for example defineXML) don't work.
Two cases missing are added in this patch to avoid that.
qemu is sensitive to the order of arguments passed. Hence, if a
device requires a controller, the controller cmd string must
precede device cmd string. The same apply for controllers, when
for instance ccid controller requires usb controller. So
controllers create partial ordering in which they should be added
to qemu cmd line.
Some FDs may not implement fdatasync() functionality,
e.g. pipes. In that case EINVAL or EROFS is returned.
We don't want to fail then nor report any error.
Reported-by: Christophe Fergeau <cfergeau@redhat.com>
Some of the pre-snapshot check have restrictions wired in regarding
configuration options that influence taking of external checkpoints.
This patch removes restrictions that would inhibit taking of such a
snapshot.
This patch adds support to take external system checkpoints.
The functionality is layered on top of the previous disk-only snapshot
code. When the checkpoint is requested the domain memory is saved to the
memory image file using migration to file. (The user may specify to
take the memory image while the guest is live with the
VIR_DOMAIN_SNAPSHOT_CREATE_LIVE flag.)
The memory save image shares format with the image created by
virDomainSave() API.
Before now, libvirt supported only internal snapshots for active guests.
This patch renames this function to qemuDomainSnapshotCreateActiveInternal
to prepare the grounds for external active snapshots.
The new external system checkpoints will require an async job while the
snapshot is taken. This patch adds QEMU_ASYNC_JOB_SNAPSHOT to track this
job type.
The default behavior while creating external checkpoints is to pause the
guest while the memory state is captured. We want the users to sacrifice
space saving for creating the memory save image while the guest is live
to minimize downtime.
This patch adds a flag that causes the guest not to be paused before
taking the snapshot.
*include/libvirt/libvirt.h.in:
- add new paused reason: VIR_DOMAIN_PAUSED_SNAPSHOT
- add new flag for taking snapshot: VIR_DOMAIN_SNAPSHOT_CREATE_LIVE
*tools/virsh-domain-monitor.c:
- add string representation for VIR_DOMAIN_PAUSED_SNAPSHOT
*tools/virsh-snapshot.c:
- add support for VIR_DOMAIN_SNAPSHOT_CREATE_LIVE
*tools/virsh.pod:
- add docs for --live option added to use
VIR_DOMAIN_SNAPSHOT_CREATE_LIVE flag
The code that saves domain memory by migration to file can be reused
while doing external checkpoints of a machine. This patch extracts the
common code and places it in a separate function.
When pausing the guest while migration is running (to speed up
convergence) the virDomainSuspend API checks if the migration job is
active before entering the job. This could cause a possible race if the
virDomainSuspend is called while the job is active but ends before the
Suspend API enters the job (this would require that the migration is
aborted). This would cause a incorrect event to be emitted.
Both system checkpoint snapshots and disk snapshots were iterating
over all disks, doing a final sanity check before doing any work.
But since future patches will allow offline snapshots to be either
external or internal, it makes sense to share the pass over all
disks, and then relax restrictions in that pass as new modes are
implemented. Future patches can then handle external disks when
the domain is offline, then handle offline --disk-snapshot, and
finally, combine with migration to file to gain a complete external
system checkpoint snapshot of an active domain without using 'savevm'.
* src/qemu/qemu_driver.c (qemuDomainSnapshotDiskPrepare)
(qemuDomainSnapshotIsAllowed): Merge...
(qemuDomainSnapshotPrepare): ...into one function.
(qemuDomainSnapshotCreateXML): Update caller.
Now that the XML supports listing internal snapshots, it is worth
always populating the <memory> and <disks> element to match.
* src/qemu/qemu_driver.c (qemuDomainSnapshotCreateXML): Always
parse disk info and set memory info.
There were not previous callers with require_match set to true.
I originally implemented this bool with the intent of supporting
ESX snapshot semantics, where the choice of internal vs. external
vs. non-checkpointable must be made at domain start, but as ESX
has not been wired up to use it yet, we might as well fix it to
work with our next qemu patch for now, and worry about any further
improvements (changing the bool to a flags argument) if the ESX
driver decides to use this function in the future.
* src/conf/snapshot_conf.c (virDomainSnapshotAlignDisks): Alter
logic when require_match is true to deal with new XML.
Each <domainsnapshot> can now contain an optional <memory>
element that describes how the VM state was handled, similar
to disk snapshots. The new element will always appear in
output; for back-compat, an input that lacks the element will
assume 'no' or 'internal' according to the domain state.
Along with this change, it is now possible to pass <disks> in
the XML for an offline snapshot; this also needs to be wired up
in a future patch, to make it possible to choose internal vs.
external on a per-disk basis for each disk in an offline domain.
At that point, using the --disk-only flag for an offline domain
will be able to work.
For some examples below, remember that qemu supports the
following snapshot actions:
qemu-img: offline external and internal disk
savevm: online internal VM and disk
migrate: online external VM
transaction: online external disk
=====
<domainsnapshot>
<memory snapshot='no'/>
...
</domainsnapshot>
implies that there is no VM state saved (mandatory for
offline and disk-only snapshots, not possible otherwise);
using qemu-img for offline domains and transaction for online.
=====
<domainsnapshot>
<memory snapshot='internal'/>
...
</domainsnapshot>
state is saved inside one of the disks (as in qemu's 'savevm'
system checkpoint implementation). If needed in the future,
we can also add an attribute pointing out _which_ disk saved
the internal state; maybe disk='vda'.
=====
<domainsnapshot>
<memory snapshot='external' file='/path/to/state'/>
...
</domainsnapshot>
This is not wired up yet, but future patches will allow this to
control a combination of 'virsh save /path/to/state' plus disk
snapshots from the same point in time.
=====
So for 1.0.1 (and later, as needed), I plan to implement this table
of combinations, with '*' designating new code and '+' designating
existing code reached through new combinations of xml and/or the
existing DISK_ONLY flag:
domain memory disk disk-only | result
-----------------------------------------
offline omit omit any | memory=no disk=int, via qemu-img
offline no omit any |+memory=no disk=int, via qemu-img
offline omit/no no any | invalid combination (nothing to snapshot)
offline omit/no int any |+memory=no disk=int, via qemu-img
offline omit/no ext any |*memory=no disk=ext, via qemu-img
offline int/ext any any | invalid combination (no memory to save)
online omit omit off | memory=int disk=int, via savevm
online omit omit on | memory=no disk=default, via transaction
online omit no/ext off | unsupported for now
online omit no on | invalid combination (nothing to snapshot)
online omit ext on | memory=no disk=ext, via transaction
online omit int off |+memory=int disk=int, via savevm
online omit int on | unsupported for now
online no omit any |+memory=no disk=default, via transaction
online no no any | invalid combination (nothing to snapshot)
online no int any | unsupported for now
online no ext any |+memory=no disk=ext, via transaction
online int/ext any on | invalid combination (disk-only vs. memory)
online int omit off |+memory=int disk=int, via savevm
online int no/ext off | unsupported for now
online int int off |+memory=int disk=int, via savevm
online ext omit off |*memory=ext disk=default, via migrate+trans
online ext no off |+memory=ext disk=no, via migrate
online ext int off | unsupported for now
online ext ext off |*memory=ext disk=ext, via migrate+transaction
* docs/schemas/domainsnapshot.rng (memory): New RNG element.
* docs/formatsnapshot.html.in: Document it.
* src/conf/snapshot_conf.h (virDomainSnapshotDef): New fields.
* src/conf/domain_conf.c (virDomainSnapshotDefFree)
(virDomainSnapshotDefParseString, virDomainSnapshotDefFormat):
Manage new fields.
* tests/domainsnapshotxml2xmltest.c: New test.
* tests/domainsnapshotxml2xmlin/*.xml: Update existing tests.
* tests/domainsnapshotxml2xmlout/*.xml: Likewise.
The libvirt coding standard is to use 'function(...args...)'
instead of 'function (...args...)'. A non-trivial number of
places did not follow this rule and are fixed in this patch.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
When assigning the new persistent definition for a transient network
(thus making it persistent) the network needs to be marked persistent
before actually atempting to assign the definition.
Until now, the network undefine API was able to undefine only inactive
networks. The restriction doesn't make sense any more so this patch
implements changing networks to transient.
When a transient network was created some of the checks weren't run on
the definition allowing to start invalid networks.
This patch splits out code to the network validation function and
re-uses that code when creating transient networks.
The network driver didn't care about config files when a network was
destroyed, just when it was undefined leaving behind files for transient
networks.
This patch splits out the cleanup code to a helper function that handles
the cleanup if the inactive network object is being removed and re-uses
this code when getting rid of inactive networks.
The hosts file was created in the network definition function. This
patch moves the place the file is being created to the point where
dnsmasq is being started.
With our fix of mkostemp (pushed as 2b435c15) we define a macro
to compile with uclibc. However, this definition is conditional
and thus needs to be properly indented. Moreover, with this definition
sc_prohibit_mkstemp syntax-check rule keeps yelling:
src/util/logging.c:63:# define mkostemp(x,y) mkstemp(x)
maint.mk: use mkostemp with O_CLOEXEC instead of mkstemp
Therefore we should ignore this file for this rule.
BZ:https://bugzilla.redhat.com/show_bug.cgi?id=871273
when using virsh qemu-attach to attach an existing qemu process,
if it misses the -M option in qemu command line, libvirtd crashed
because the NULL value of def->os.machine in later use.
Example:
/usr/libexec/qemu-kvm -name foo \
-cdrom /var/lib/libvirt/images/boot.img \
-monitor unix:/tmp/demo,server,nowait \
error: End of file while reading data: Input/output error
error: Failed to reconnect to the hypervisor
This patch tries to set default machine type if the value of
def->os.machine is still NULL after qemu command line parsing.
* configure.ac docs/news.html.in libvirt.spec.in: update for the new release
* po/*.po*: update from transifex, a lot of added support e.g. Indian
languages, and regenerate
It turns out that calling virNodeGetCPUMap(conn, NULL, NULL, 0)
is both useful, and with Viktor's patches, common enough to
optimize. Since this interface hasn't been released yet, we
can change the RPC call.
A bit more background on the optimization - learning the cpu count
is a single file read (/sys/devices/system/cpu/possible), but
learning the number of online cpus can possibly trigger a file
read per cpu, depending on the age of the kernel, and all wasted
if the caller passed NULL for both arguments.
* src/nodeinfo.c (nodeGetCPUMap): Avoid bitmap when not needed.
* src/remote/remote_protocol.x (remote_node_get_cpu_map_args):
Supply two separate flags for needed arguments.
* src/remote/remote_driver.c (remoteNodeGetCPUMap): Update
caller.
* daemon/remote.c (remoteDispatchNodeGetCPUMap): Likewise.
* src/remote_protocol-structs: Regenerate.
Per the code comment in qemuCapsInitQMPBasic() and commit 43e23c7, we
should only use QMP for capabilities probing starting with 1.2 and
newer. The old code had dead logic that probed on 1.0 and newer.
Signed-off-by: Eric Blake <eblake@redhat.com>
This needs to be done before the container starts. Turning
off the mknod capability is noticed by systemd, which will
no longer attempt to create device nodes.
This eliminates SELinux AVC messages and ugly failure messages in the journal.
The string comparison logic was inverted and matched the first drive
that does *not* have the name we search for.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The QEMU -drive id= begins with libvirt's QEMU host drive prefix
("drive-"), which is stripped off in several places two convert between
host ("-drive") and guest ("-device") device names.
In the case of BlkIoTune it is unnecessary to strip the QEMU host drive
prefix because we operate on "info block"/"query-block" output that uses
host drive names.
Stripping the prefix incorrectly caused string comparisons to fail since
we were comparing the guest device name against the host device name.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Currently, when we are doing (managed) save, we insert the
iohelper between the qemu and OS. The pipe is created, the
writing end is passed to qemu and the reading end to the
iohelper. It reads data and write them into given file. However,
with write() being asynchronous data may still be in OS
caches and hence in some (corner) cases, all migration data
may have been read and written (not physically though). So
qemu will report success, as well as iohelper. However, with
some non local filesystems, where ENOSPACE is polled every X
time units, we may get into situation where all operations
succeeded but data hasn't reached the disk. And in fact will
never do. Therefore we ought sync caches to make sure data
has reached the block device on remote host.
QEMU uses 'i386' for its 32-bit x86 architecture, but libvirt
wants that to be 'i686', so we must fix it up
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
virPidFileReadPathIfAlive passed in an 'int *' where a 'pid_t *'
was expected, which breaks on Mingw64 targets. Also a few places
were using '%d' for formatting pid_t, change them to '%lld' and
force a cast to the longer type as done elsewhere in the same
file.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=871756
Commit cd1e8d1 assumed that systems new enough to have journald
also have mkostemp; but this is not true for uclibc.
For that matter, use of mkstemp[s] is unsafe in a multi-threaded
program. We should prefer mkostemp[s] in the first place.
* bootstrap.conf (gnulib_modules): Add mkostemp, mkostemps; drop
mkstemp and mkstemps.
* cfg.mk (sc_prohibit_mkstemp): New syntax check.
* tools/virsh.c (vshEditWriteToTempFile): Adjust caller.
* src/qemu/qemu_driver.c (qemuDomainScreenshot)
(qemudDomainMemoryPeek): Likewise.
* src/secret/secret_driver.c (replaceFile): Likewise.
* src/vbox/vbox_tmpl.c (vboxDomainScreenshot): Likewise.
https://bugzilla.redhat.com/show_bug.cgi?id=871312
Recent fixes made almost all the right steps to make emulator pinned
to the cpuset of the whole domain in case <emulatorpin> isn't
specified, but qemudDomainGetEmulatorPinInfo still reports all the
CPUs even when cpuset is specified. This patch fixes that.
There are multiple reasons canonicalize_file_name() used in
absolutePathFromBaseFile helper can fail. This patch enhances error
reporting from that helper.
When there is no 'qemu-kvm' binary and the emulator used for a machine
is, for example, 'qemu-system-x86_64' that, by default, runs without
kvm enabled, libvirt still supplies '-no-kvm' option to this process,
even though it does not recognize such option (making the start of a
domain fail in that case).
This patch fixes building a command-line for QEMU machines without KVM
acceleration and is based on following assumptions:
- QEMU_CAPS_KVM flag means that QEMU is running KVM accelerated
machines by default (without explicitly requesting that using a
command-line option). It is the closest to the truth according to
the code with the only exception being the comment next to the
flag, so it's fixed in this patch as well.
- QEMU_CAPS_ENABLE_KVM flag means that QEMU is, by default, running
without KVM acceleration and in case we need KVM acceleration it
needs to be explicitly instructed to do so. This is partially
true for the past (this option essentially means that QEMU
recognizes the '-enable-kvm' option, even though it's almost the
same).
Three FORWARD chain rules are added and two INPUT chain rules
are added when a network is started but only the FORWARD chain
rules are removed when the network is destroyed.
I noticed this while answering a list question about Java bindings
of volume creation. All other functions that take xml logged xmlDesc.
* src/libvirt.c (virStorageVolCreateXML)
(virStorageVolCreateXMLFrom): Use consistent spelling of xmlDesc,
and log the argument.
This patch resolves: https://bugzilla.redhat.com/show_bug.cgi?id=871201
If libvirt is restarted after updating the dnsmasq or radvd packages,
a subsequent "virsh net-destroy" will fail to kill the dnsmasq/radvd
process.
The problem is that when libvirtd restarts, it re-reads the dnsmasq
and radvd pidfiles, then does a sanity check on each pid it finds,
including checking that the symbolic link in /proc/$pid/exe actually
points to the same file as the path used by libvirt to execute the
binary in the first place. If this fails, libvirt assumes that the
process is no longer alive.
But if the original binary has been replaced, the link in /proc is set
to "$binarypath (deleted)" (it literally has the string " (deleted)"
appended to the link text stored in the filesystem), so even if a new
binary exists in the same location, attempts to resolve the link will
fail.
In the end, not only is the old dnsmasq/radvd not terminated when the
network is stopped, but a new dnsmasq can't be started when the
network is later restarted (because the original process is still
listening on the ports that the new process wants).
The solution is, when the initial "use stat to check for identical
inodes" check for identity between /proc/$pid/exe and $binpath fails,
to check /proc/$pid/exe for a link ending with " (deleted)" and if so,
truncate that part of the link and compare what's left with the
original binarypath.
A twist to this problem is that on systems with "merged" /sbin and
/usr/sbin (i.e. /sbin is really just a symlink to /usr/sbin; Fedora
17+ is an example of this), libvirt may have started the process using
one path, but /proc/$pid/exe lists a different path (indeed, on F17
this is the case - libvirtd uses /sbin/dnsmasq, but /proc/$pid/exe
shows "/usr/sbin/dnsmasq"). The further bit of code to resolve this is
to call virFileResolveAllLinks() on both the original binarypath and
on the truncated link we read from /proc/$pid/exe, and compare the
results.
The resulting code still succeeds in all the same cases it did before,
but also succeeds if the binary was deleted or replaced after it was
started.
After separating 5.x and 5.1 versions of ESX, we forgot to add 5.1
into the list of allowed connections, so connections to 5.1 fail since
v1.0.0-rc1-5-g1e7cd39
Ever since commit eefb881, ATTRIBUTE_NONNULL has normally been a
no-op under gcc (since it tends to cause more bugs than it cures
given gcc's current lame implementation of the attribute). However,
the macro is still useful to Coverity and other static-analysis
tools, but only if we use it correctly. Coverity follows gcc's lead
in accepting function declarations with attributes at the end, but
function bodies must attach attributes to the return type. That is,
these are valid:
void foo(void *arg) ATTRIBUTE_NONNULL(1);
void ATTRIBUTE_NONNULL(1) foo(void *arg);
void ATTRIBUTE_NONNULL(1) foo(void *arg) {}
but this is not:
void foo(void *arg) ATTRIBUTE_NONNULL(1) {}
even though you don't get a compile failure until you do static
analysis. Bug introduced in commit 80533ca, with these symptoms:
nodeinfo.c:206: error: expected ',' or ';' before '{' token
cc1: warning: unrecognized command line option "-Wno-suggest-attribute=const"
cc1: warning: unrecognized command line option "-Wno-suggest-attribute=pure"
make[3]: *** [libvirt_driver_la-nodeinfo.lo] Error 1
* src/nodeinfo.c (virNodeParseNode): Fix syntax error when
non-null attribute is in use.
Commit 34e8f63a3 altered virfile.o to drag in additional symbols,
which in turn led to pulling in other .o files and eventually causing
a link failure when systemtap probes are enabled, such as:
./.libs/libvirt_util.a(libvirt_util_la-event_poll.o): In function `virEventPollRunOnce':
/home/dummy/libvirt/src/util/event_poll.c:614: undefined reference to `libvirt_event_poll_run_semaphore'
./.libs/libvirt_util.a(libvirt_util_la-event_poll.o):(.note.stapsdt+0x24): undefined reference to `libvirt_event_poll_add_handle_semaphore'
Even though libvirt_iohelper and libvirt_parthelper don't directly
use the portion of virfile.o that drags in probing, it was easier
to satisfy the linker and get the build back up, than to figure out
whether it is even possible or worth trying to disentangle the mess.
* src/Makefile.am (libvirt_iohelper_LDADD)
(libvirt_parthelper_LDADD): Use libvirt_probes.lo when needed.
Currently, we use iohelper when saving/restoring a domain.
However, if there's some kind of error (like I/O) it is not
propagated to libvirt. Since it is not qemu who is doing
the actual write() it will not get error. The iohelper does.
Therefore we should check for iohelper errors as it makes
libvirt more user friendly.
In the XML warning, we print a virsh command line that can be used to
edit that XML. This patch prints UUIDs if the entity name contains
special characters (like shell metacharacters, or "--" that would break
parsing of the XML comment). If the entity doesn't have a UUID, just
print the virsh command that can be used to edit it.
When using block copy to pivot over to a new chain, the backing files
for the new chain might still need labeling (particularly if the user
passes --reuse-ext with a relative backing file name). Relabeling a
file that is already labeled won't hurt, so this just labels the entire
chain at the point of the pivot. Doing the relabel of the chain uses
the fact that we already safely probed the file type of an external
file at the start of the block copy.
* src/qemu/qemu_driver.c (qemuDomainBlockPivot): Relabel chain before
asking qemu to pivot.
Use the recent addition of qemuDomainPrepareDiskChainElement to
obtain locking manager lease, permit a block device through cgroups,
and set the SELinux label; then audit the fact that we hand a new
file over to qemu. Alas, releasing the lease and label at the end
of the mirroring is a trickier prospect (we would have to trace the
backing chain of both source and destination, and be sure not to
revoke rights to any part of the chain that is shared), so for now,
virDomainBlockJobAbort still leaves things with additional access
granted (as block-pull and block-commit have the same problem of
not clamping access after completion, a future cleanup would cover
all three commands).
* src/qemu/qemu_driver.c (qemuDomainBlockCopy): Set up labeling.
Support the REUSE_EXT flag, in part by copying sanity checks from
snapshot code. This code introduces a case of probing an external
file for its type; such an action would be a security risk if the
existing file is supposed to be raw but the contents resemble some
other format; however, since the virDomainBlockRebase API has a
flag to force treating the file as raw rather than probe, we can
assume that probing is safe in all other instances. Besides, if
we don't probe or force raw, then qemu will.
* src/qemu/qemu_driver.c (qemuDomainBlockRebase): Allow REUSE_EXT
flag.
(qemuDomainBlockCopy): Wire up flag, and add some sanity checks.
Minimal patch to wire up all the pieces in the previous patches
to actually enable a block copy job. By minimal, I mean that
qemu creates the file (that is, no REUSE_EXT flag support yet),
SELinux must be disabled, a lock manager is not informed, and the
audit logs aren't updated. But those will be added as
improvements in future patches.
This patch is designed so that if we ever add a future API
virDomainBlockCopy with more bells and whistles (such as letting
the user specify a destination image format different than the
source), where virDomainBlockRebase is a wrapper around the
simpler portions of the new functionality, then the new API can
just reuse the new qemuDomainBlockCopy function and already
support _SHALLOW and _REUSE_EXT flags. Also note that libvirt.c
already filtered the new flags if _COPY is not present, so that
we are not impacting the case of BlockRebase being a wrapper
around BlockPull.
* src/qemu/qemu_driver.c (qemuDomainBlockCopy): New function.
(qemuDomainBlockRebase): Call it when appropriate.
Since libvirt drops locks between issuing a monitor command and
getting a response, it is possible for libvirtd to be restarted
before getting a response on a block-job-complete command; worse, it
is also possible for the guest to shut itself down during the window
while libvirtd is down, ending the qemu process. A management app
needs to know if the pivot happened (and the destination file
contains guest contents not in the source) or failed (and the source
file contains guest contents not in the destination), but since
the job is finished, 'query-block-jobs' no longer tracks the
status of the job, and if the qemu process itself has disappeared,
even 'query-block' cannot be checked to ask qemu its current state.
At the time of this patch, the design for persistent bitmap has not
been clarified, so a followup patch will be needed once qemu
actually figures out how to expose it, and we figure out how to use
it. In the meantime, we have a solution that avoids the worst of
the problem. [This problem was first analyzed with the RHEL 6.3
__com.redhat_drive-reopen command; which partly explains why
upstream qemu 1.3 ditched the drive-reopen idea and went with
block-job-complete plus persistent bitmap instead.]
If we surround 'drive-reopen' with a pause/resume pair, then we can
guarantee that the guest cannot modify either source or destination
files in the window of libvirtd uncertainty, and the management app
is guaranteed that either libvirt knows the outcome and reported it
correctly; or that on libvirtd restart, the guest will still be
paused and that the qemu process cannot have disappeared due to
guest shutdown; and use that as a clue that the management app must
implement recovery protocol, with both source and destination files
still being in sync and with 'query-block' still being an option as
part of that recovery. My testing shows that the pause window will
typically be only a fraction of a second.
* src/qemu/qemu_driver.c (qemuDomainBlockPivot): Pause around
drive-reopen.
(qemuDomainBlockJobImpl): Update caller.
This is the bare minimum to end a copy job (of course, until a
later patch adds the ability to start a copy job, this patch
doesn't do much in isolation; I've just split the patches to
ease the review).
This patch intentionally avoids SELinux, lock manager, and audit
actions. Also, if libvirtd restarts at the exact moment that a
'block-job-complete' is in flight, the proposed proper way to
detect the outcome of that would be with a persistent bitmap and
some additional query commands when libvirtd restarts. This
patch is enough to test the common case of success when used
correctly, while saving the subtleties of proper cleanup for
worst-case errors for later.
When a mirror job is started, cancelling the job safely reverts back
to the source disk, regardless of whether the destination is in
phase 1 (streaming, in which case the destination is worthless) or
phase 2 (mirroring, in which case the destination is synced up to
the source at the time of the cancel). Our existing code does just
fine in either phase, other than some bookkeeping cleanup; this
implements live block copy.
Ideas for future enhancements via new flags:
Depending on when persistent bitmap support is added, it may be
worth adding a VIR_DOMAIN_REBASE_COPY_ATOMIC flag that fails up
front if we detect an older qemu with risky pivot operation.
Interesting side note: while snapshot-create --disk-only creates a
copy of the disk at a point in time by moving the domain on to a
new file (the copy is the file now in the just-extended backing
chain), blockjob --abort of a copy job creates a copy of the disk
while keeping the domain on the original file. There may be
potential improvements to the snapshot code to exploit block copy
over multiple disks all at one point in time. And, if
'block-job-cancel' were made part of 'transaction', you could
copy multiple disks at the same point in time without pausing
the domain. This also implies we may want to add a --quiesce flag
to virDomainBlockJobAbort, so that when breaking a mirror (whether
by cancel or pivot), the side of the mirror that we are abandoning
is at least in a stable state with regards to guest I/O.
* src/qemu/qemu_driver.c (qemuDomainBlockJobAbort): Accept new flag.
(qemuDomainBlockPivot): New helper function.
(qemuDomainBlockJobImpl): Implement it.
Handle the new type of block copy event and info. Of course,
this patch does nothing until a later patch actually allows the
creation/abort of a block copy job.
* include/libvirt/libvirt.h.in (VIR_DOMAIN_BLOCK_JOB_READY): New
block job status.
* src/libvirt.c (virDomainBlockRebase): Document the event.
* src/qemu/qemu_monitor_json.c (eventHandlers): New event.
(qemuMonitorJSONHandleBlockJobReady): New function.
(qemuMonitorJSONGetBlockJobInfoOne): Translate new job type.
(qemuMonitorJSONHandleBlockJobImpl): Handle new event and job type.
* src/qemu/qemu_process.c (qemuProcessHandleBlockJob): Recognize
the event to minimize snooping.
* src/qemu/qemu_driver.c (qemuDomainBlockJobImpl): Snoop a successful
info query to save effort on a pivot request.
For now, disk migration via block copy job is not implemented in
libvirt. But when we do implement it, we have to deal with the
fact that qemu does not yet provide an easy way to re-start a qemu
process with mirroring still intact. Paolo has proposed an idea
for a persistent dirty bitmap that might make this possible, but
until that design is complete, it's hard to say what changes
libvirt would need. Even something like 'virDomainSave' becomes
hairy, if you realize the implications that 'virDomainRestore'
would be stuck with recreating the same mirror layout.
But if we step back and look at the bigger picture, we realize that
the initial client of live storage migration via disk mirroring is
oVirt, which always uses transient domains, and that if a transient
domain is destroyed while a mirror exists, oVirt can easily restart
the storage migration by creating a new domain that visits just the
source storage, with no loss in data.
We can make life a lot easier by being cowards for now, forbidding
certain operations on a domain. This patch guarantees that we
never get in a state where we would have to restart a domain with
a mirroring block copy, by preventing saves, snapshots, migration,
hot unplug of a disk in use, and conversion to a persistent domain
(thankfully, it is still relatively easy to 'virsh undefine' a
running domain to temporarily make it transient, run tests on
'virsh blockcopy', then 'virsh define' to restore the persistence).
Later, if the qemu design is enhanced, we can relax our code.
The change to qemudDomainDefine looks a bit odd for undoing an
assignment, rather than probing up front to avoid the assignment,
but this is because of how virDomainAssignDef combines both a
lookup and assignment into a single function call.
* src/conf/domain_conf.h (virDomainHasDiskMirror): New prototype.
* src/conf/domain_conf.c (virDomainHasDiskMirror): New function.
* src/libvirt_private.syms (domain_conf.h): Export it.
* src/qemu/qemu_driver.c (qemuDomainSaveInternal)
(qemuDomainSnapshotCreateXML, qemuDomainRevertToSnapshot)
(qemuDomainBlockJobImpl, qemudDomainDefine): Prevent dangerous
actions while block copy is already in action.
* src/qemu/qemu_hotplug.c (qemuDomainDetachDiskDevice): Likewise.
* src/qemu/qemu_migration.c (qemuMigrationIsAllowed): Likewise.
Upstream qemu 1.3 is adding two new monitor commands, 'drive-mirror'
and 'block-job-complete'[1], which can drive live block copy and
storage migration. [Additionally, RHEL 6.3 had backported an earlier
version of most of the same functionality, but under the names
'__com.redhat_drive-mirror' and '__com.redhat_drive-reopen' and with
slightly different JSON arguments, and has been using patches similar
to these upstream patches for several months now.]
The libvirt API virDomainBlockRebase as already committed for 0.9.12
is flexible enough to expose the basics of block copy, but some
additional features in the 'drive-mirror' qemu command, such as
setting error policy, setting granularity, or using a persistent
bitmap, may later require a new libvirt API virDomainBlockCopy. I
will wait to add that API until we know more about what qemu 1.3
will finally provide.
This patch caters only to the upstream qemu 1.3 interface, although
I have proven that the changes for RHEL 6.3 can be isolated to
just qemu_monitor_json.c, and the rest of this series will
gracefully handle either interface once the JSON differences are
papered over in a downstream patch.
For consistency with other block job commands, libvirt must handle
the bandwidth argument as MiB/sec from the user, even though qemu
exposes the speed argument as bytes/sec; then again, qemu rounds
up to cluster size internally, so using MiB hides the worst effects
of that rounding if you pass small numbers.
[1]https://lists.gnu.org/archive/html/qemu-devel/2012-10/msg04123.html
* src/qemu/qemu_capabilities.h (QEMU_CAPS_DRIVE_MIRROR)
(QEMU_CAPS_DRIVE_REOPEN): New bits.
* src/qemu/qemu_capabilities.c (qemuCaps): Name them.
* src/qemu/qemu_monitor_json.c (qemuMonitorJSONCheckCommands): Set
them.
(qemuMonitorJSONDriveMirror, qemuMonitorDrivePivot): New functions.
* src/qemu/qemu_monitor_json.h (qemuMonitorJSONDriveMirror)
(qemuMonitorDrivePivot): Declare them.
* src/qemu/qemu_monitor.c (qemuMonitorDriveMirror)
(qemuMonitorDrivePivot): New passthroughs.
* src/qemu/qemu_monitor.h (qemuMonitorDriveMirror)
(qemuMonitorDrivePivot): Declare them.
This resolves:
https://bugzilla.redhat.com/show_bug.cgi?id=862515
which describes inconsistencies in dealing with duplicate mac
addresses on network devices in a domain.
(at any rate, it resolves *almost* everything, and prints out an
informative error message for the one problem that isn't solved, but
has a workaround.)
A synopsis of the problems:
1) you can't do a persistent attach-interface of a device with a mac
address that matches an existing device.
2) you *can* do a live attach-interface of such a device.
3) you *can* directly edit a domain and put in two devices with
matching mac addresses.
4) When running virsh detach-device (live or config), only MAC address
is checked when matching the device to remove, so the first device
with the desired mac address will be removed. This isn't always the
one that's wanted.
5) when running virsh detach-interface (live or config), the only two
items that can be specified to match against are mac address and model
type (virtio, etc) - if multiple netdevs match both of those
attributes, it again just finds the first one added and assumes that
is the only match.
Since it is completely valid to have multiple network devices with the
same MAC address (although it can cause problems in many cases, there
*are* valid use cases), what is needed is:
1) remove the restriction that prohibits doing a persistent add of a
netdev with a duplicate mac address.
2) enhance the backend of virDomainDetachDeviceFlags to check for
something that *is* guaranteed unique (but still work with just mac
address, as long as it yields only a single results.
This patch does three things:
1) removes the check for duplicate mac address during a persistent
netdev attach.
2) unifies the searching for both live and config detach of netdevices
in the subordinate functions of qemuDomainModifyDeviceFlags() to use the
new function virDomainNetFindIdx (which matches mac address and PCI
address if available, checking for duplicates if only mac address was
specified). This function returns -2 if multiple matches are found,
allowing the callers to print out an appropriate message.
Steps 1 & 2 are enough to fully fix the problem when using virsh
attach-device and detach-device (which require an XML description of
the device rather than a bunch of commandline args)
3) modifies the virsh detach-interface command to check for multiple
matches of mac address and show an error message suggesting use of the
detach-device command in cases where there are multiple matching mac
addresses.
Later we should decide how we want to input a PCI address on the virsh
commandline, and enhance detach-interface to take a --address option,
eliminating the need to use detach-device
* src/conf/domain_conf.c
* src/conf/domain_conf.h
* src/libvirt_private.syms
* added new virDomainNetFindIdx function
* removed now unused virDomainNetIndexByMac and
virDomainNetRemoveByMac
* src/qemu/qemu_driver.c
* remove check for duplicate max from qemuDomainAttachDeviceConfig
* use virDomainNetFindIdx/virDomainNetRemove instead
of virDomainNetRemoveByMac in qemuDomainDetachDeviceConfig
* use virDomainNetFindIdx instead of virDomainIndexByMac
in qemuDomainUpdateDeviceConfig
* src/qemu/qemu_hotplug.c
* use virDomainNetFindIdx instead of a homespun loop in
qemuDomainDetachNetDevice.
* tools/virsh-domain.c: modified detach-interface command as described
above
It turns out that the cpuacct results properly account for offline
cpus, and always returns results for every possible cpu, not just
the online ones. So there is no need to check the map of online
cpus in the first place, merely only a need to know the maximum
possible cpu. Meanwhile, virNodeGetCPUBitmap had a subtle change
from returning the maximum id to instead returning the width of
the bitmap (one larger than the maximum id) in commit 2f4c5338,
which made this code encounter some off-by-one logic leading to
bad error messages when a cpu was offline:
$ virsh cpu-stats dom
error: Failed to virDomainGetCPUStats()
error: An error occurred, but the cause is unknown
Cleaning this up unraveled a chain of other unused variables.
* src/qemu/qemu_driver.c (qemuDomainGetPercpuStats): Drop
pointless check for cpumap changes, and use correct number of
cpus. Simplify signature.
(qemuDomainGetCPUStats): Adjust caller.
* src/nodeinfo.h (nodeGetCPUCount): New prototype.
(nodeGetCPUBitmap): Drop unused parameter.
* src/nodeinfo.c (nodeGetCPUBitmap): Likewise.
(nodeGetCPUMap): Adjust caller.
(nodeGetCPUCount): New function.
* src/libvirt_private.syms (nodeinfo.h): Export it.
Commit 246143b fixed a warning on older gcc, but caused a warning
on newer gcc.
../../src/rpc/virnetserverservice.c: In function 'virNetServerServiceNewPostExecRestart':
../../src/rpc/virnetserverservice.c:277:41: error: pointer targets in passing argument 3 of 'virJSONValueObjectGetNumberUint' differ in signedness [-Werror=pointer-sign]
* src/rpc/virnetserverservice.c: Use correct types.
With older gcc and 64-bit size_t, the compiler issues a real warning:
rpc/virnetserverservice.c:277: warning: dereferencing type-punned pointer will break strict-aliasing rules [-Wstrict-aliasing]
Introduced in commit 0cc79255. Depending on machine endianness,
this warning represents a real bug that could mis-interpret the
value by a factor of 2^32. I don't know why I couldn't get newer
gcc to report the same warning message.
* src/rpc/virnetserverservice.c
(virNetServerServiceNewPostExecRestart): Use temporary instead.
Found this when building on RHEL5:
parallels/parallels_storage.c: In function 'parallelsStorageOpen':
parallels/parallels_storage.c:180: error: 'for' loop initial declaration used outside C99 mode
(and similar error in parallels_driver.c). This was in spite of
configuring with "-Wno-error".
This commit changes the behavior of LIBVIRT_DEBUG=1 libvirtd:
$ git show 7022b09111
commit 7022b09111
Author: Daniel P. Berrange <berrange@redhat.com>
Date: Thu Sep 27 13:13:09 2012 +0100
Automatically enable systemd journal logging
Probe to see if the systemd journal is accessible, and if
so enable logging to the journal by default, rather than
stderr (current default under systemd).
Previously 'LIBVIRT_DEBUG=1 /usr/sbin/libvirtd' would show all debug
output to stderr, now it send debug output to the journal.
Only use the journal by default if running in daemon mode, or
if stdin is _not_ a tty. This should make libvirtd launched from
systemd use the journal, but preserve the old behavior in most
situations.
This was found during testing of the fix for:
https://bugzilla.redhat.com/show_bug.cgi?id=868483
networkValidate was supposed to check for the existence of multiple
portgroups and report an error if this was encountered. It did, but
there were two problems:
1) even though it logged an error, it still returned success, allowing
the operation to continue.
2) It could exit the portgroup checking loop early (or possibly not
even do it once) if a vlan tag was supplied in the base network config
or one of the portgroups.
This patch fixes networkValidate to return failure in addition to
logging the error, and also changes it to not exit the portgroup
checking loop early. The logic was a bit off in the checking for vlan
anyway, and it's intertwined with fixing the early loop exit, so I
fixed that as well. Now it correctly checks for combinations where a
<virtualport> is specified in the base network def and <vlan> is given
in a portgroup, as well as the opposite (<vlan> in base network def
and <virtualport> in portgroup), and ignores the case of a disallowed
vlan when using *no* portgroup if there is a default portgroup (since
in that case there is no way to not use any portgroup).
Added an implemention of virNodeGetCPUMap to nodeinfo.c,
(nodeGetCPUMap) which can be used by all drivers for a Linux
hypervisor host.
Signed-off-by: Viktor Mihajlovski <mihajlov@linux.vnet.ibm.com>
Callers should not need to know what the name of the file to
be read in the Linux-specific version of nodeGetCPUmap;
furthermore, qemu cares about online cpus, not present cpus,
when determining which cpus to skip.
While at it, I fixed the fact that we were computing the maximum
online cpu id by doing a slow iteration, when what we really want
to know is the max available cpu.
* src/nodeinfo.h (nodeGetCPUmap): Rename...
(nodeGetCPUBitmap): ...and simplify signature.
* src/nodeinfo.c (linuxParseCPUmax): New function.
(linuxParseCPUmap): Simplify and alter signature.
(nodeGetCPUBitmap): Change implementation.
* src/libvirt_private.syms (nodeinfo.h): Reflect rename.
* src/qemu/qemu_driver.c (qemuDomainGetPercpuStats): Update
caller.
Sometimes it's handy to know how many bits are set.
* src/util/bitmap.h (virBitmapCountBits): New prototype.
(virBitmapNextSetBit): Use correct type.
* src/util/bitmap.c (virBitmapNextSetBit): Likewise.
(virBitmapSetAll): Maintain invariant of clear tail bits.
(virBitmapCountBits): New function.
* src/libvirt_private.syms (bitmap.h): Export it.
* tests/virbitmaptest.c (test2): Test it.
Also remove warnings for upcoming versions. There hadn't been any
compatibility problems with new ESX version over the whole lifetime
of the ESX driver, so I don't expect any in the future.
Update documentation to mention vSphere 5.x support.
Qemu has added some new feature flags. This patch adds them to libvirt.
The new features are for the cpuid function 0x7 that takes an argument
in the ecx register. Currently only 0x0 is used as the argument so I was
lazy and I just clear the registers to 0 before calling cpuid. In future
when there maybe will be some other possible arguments, we will need to
improve the cpu detection code to take this into account.
On one hand, numad probably will manage the affinity of domain process
dynamically in future. On the other hand, even numad won't manage it,
it still could confusion. Let's make things simpler enough to avoid
the lair for now.
When the cpu placement model is "auto", it sets the affinity for
domain process with the advisory nodeset from numad, however,
creating cgroup for the domain process (called emulator thread
in some contexts) later overrides that with pinning it to all
available pCPUs.
How to reproduce:
* Configure the domain with "auto" placement for <vcpu>, e.g.
<vcpu placement='auto'>4</vcpu>
* % virsh start dom
* % cat /proc/$dompid/status
Though the emulator cgroup cause conflicts, but we can't simply
prohibit creating it, as other tunables are still useful, such
as "emulator_period", which is used by API
virDomainSetSchedulerParameter. So this patch doesn't prohibit
creating the emulator cgroup, but inherit the nodeset from numad,
and reset the affinity for domain process.
* src/qemu/qemu_cgroup.h: Modify definition of qemuSetupCgroupForEmulator
to accept the passed nodenet
* src/qemu/qemu_cgroup.c: Set the affinity with the passed nodeset
Abstract the codes to prepare cpumap into a helper a function,
which can be used later.
* src/qemu/qemu_process.h: Declare qemuPrepareCpumap
* src/qemu/qemu_process.c: Implement qemuPrepareCpumap, and use it.
- Defined the wire protocol format for virNodeGetCPUMap and its
arguments
- Implemented remote method invocation (remoteNodeGetCPUMap)
- Implemented method dispatcher (remoteDispatchNodeGetCPUMap)
Signed-off-by: Viktor Mihajlovski <mihajlov@linux.vnet.ibm.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
Adding a new API to obtain information about the
host node's present, online and offline CPUs.
int virNodeGetCPUMap(virConnectPtr conn,
unsigned char **cpumap,
unsigned int *online,
unsigned int flags);
The function will return the number of CPUs present on the host
or -1 on failure;
If cpumap is non-NULL virNodeGetCPUMap will allocate an array
containing a bit map representation of the online CPUs. It's
the callers responsibility to deallocate cpumap using free().
If online is non-NULL, the variable pointed to will contain
the number of online host node CPUs.
The variable flags has been added to support future extensions
and must be set to 0.
Extend the driver structure by nodeGetCPUMap entry in support of the
new API virNodeGetCPUMap.
Added implementation of virNodeGetCPUMap to libvirt.c
Signed-off-by: Viktor Mihajlovski <mihajlov@linux.vnet.ibm.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
Transport Open vSwitch per-port data during live
migration by using the utility functions
virNetDevOpenvswitchGetMigrateData() and
virNetDevOpenvswitchSetMigrateData().
Signed-off-by: Kyle Mestery <kmestery@cisco.com>
Add utility functions for Open vSwitch to both save
per-port data before a live migration, and restore the
per-port data after a live migration.
Signed-off-by: Kyle Mestery <kmestery@cisco.com>
Add the ability for the Qemu V3 migration protocol to
include transporting network configuration. A generic
framework is proposed with this patch to allow for the
transfer of opaque data.
Signed-off-by: Kyle Mestery <kmestery@cisco.com>
Signed-off-by: Laine Stump <laine@laine.org>
In commit 371ddc98, I mistakenly added the check for sysctl
version 9 after setting the hypercall version to 1, which will
fail with
error : xenHypervisorDoV1Op:967 : Unable to issue hypervisor
ioctl 3166208: Function not implemented
This check should be included along with the others that use
hypercall version 2.
When restoring selinux labels after a VM is stopped, any non-standard
path that doesn't have a default selinux label causes the process
to stop and exit early. This isn't really an error condition IMO.
Of course the selinux API could be erroring for some other reason
but hopefully that's rare enough to not need explicit handling.
Common example here is storing disk images in a non-standard location
like under /mnt.
We put a comment containing "virsh edit <domain_name>" at the start of
the XML. W3C recommendation forbids the use of "--" in comments [1] and
libvirt can't parse it either. This patch omits the domain name if it
contains a double hyphen.
[1] http://www.w3.org/TR/REC-xml/#sec-comments
Rename the 'wait' parameter to 'loop'.
This silences the warning:
storage/storage_backend.c:1348:34: error: declaration of 'wait' shadows
a global declaration [-Werror=shadow]
and fixes the build with -Werror.
--
Note: loop is pool backwards.
The snapshot code when reusing an existing file had hard-to-read
logic, as well as a missing sanity check: REUSE_EXT should require
the destination to already be present.
* src/qemu/qemu_driver.c (qemuDomainSnapshotDiskPrepare): Require
destination on REUSE_EXT, rename variable for legibility.
Fixes a build failure on cygwin:
cc1: warnings being treated as errors
security/security_dac.c: In function 'virSecurityDACSetProcessLabel':
security/security_dac.c:862:5: error: format '%u' expects type 'unsigned int', but argument 7 has type 'uid_t' [-Wformat]
security/security_dac.c:862:5: error: format '%u' expects type 'unsigned int', but argument 8 has type 'gid_t' [-Wformat]
* src/security/security_dac.c (virSecurityDACSetProcessLabel)
(virSecurityDACGenLabel): Use proper casts.
virStorageVolLookupByPath is an API call that virt-manager uses
quite a bit when dealing with storage. This call use BackendStablePath
which has several usleep() heuristics that can be tripped up
and hang virt-manager for a while.
Current example: an empty mpath pool pointing to /dev/mapper makes
_any_ calls to virStorageVolLookupByPath take 5 seconds.
The sleep heuristics are actually only needed in certain cases
when we are waiting for new storage to appear, so let's skip the
timeout steps when calling from LookupByPath.
Currently it's assumed that qemu always supports VNC, however it is
definitely possible to compile qemu without VNC support so we should at
the very least check for it and handle that correctly.
Yet another instance of where using plain open() mishandles files
that live on root-squash NFS, and where improving the API can
improve the chance of a successful probe.
* src/util/storage_file.h (virStorageFileProbeFormat): Alter
signature.
* src/util/storage_file.c (virStorageFileProbeFormat): Use better
method for opening file.
* src/qemu/qemu_driver.c (qemuDomainGetBlockInfo): Update caller.
* src/storage/storage_backend_fs.c (virStorageBackendProbeTarget):
Likewise.
In v2 migration protocol, XML is obtained by calling domainGetXMLDesc.
This includes the default USB controller in XML, which breaks migration
to older libvirt (before 0.9.2).
Commit 409b5f5495
qemu: Emit compatible XML when migrating a domain
only fixed this for v3 migration.
This patch uses the new VIR_DOMAIN_XML_MIGRATABLE flag (detected by
VIR_DRV_FEATURE_XML_MIGRATABLE) to obtain XML without the default controller,
enabling backward v2 migration.
As we switched to setting capabilities based on QMP communication,
qemu seamless-migration capability was not set. In the -help output
this knob is called seamless-migration=[on|off]. The equivalent in
QMP world is SPICE_MIGRATE_COMPLETED event (qemu upstream commit
2fdd16e2).
This resolves: https://bugzilla.redhat.com/show_bug.cgi?id=868483
virNetworkUpdate, virNetworkDefine, and virNetworkCreate all three
allow network definitions to contain multiple <portgroup> elements
with default='yes'. Only a single default portgroup should be allowed
for each network.
This patch updates networkValidate() (called by both
virNetworkCreate() and virNetworkDefine()) and
virNetworkDefUpdatePortGroup (called by virNetworkUpdate() to not
allow multiple default portgroups.
This fixes the problem reported in:
https://bugzilla.redhat.com/show_bug.cgi?id=868389
Previously, the dnsmasq hosts file (used for static dhcp entries, and
addnhosts file (used for additional dns host entries) were only
created/referenced on the dnsmasq commandline if there was something
to put in them at the time the network was started. Once we can update
a network definition while it's active (which is now possible with
virNetworkUpdate), this is no longer a valid strategy - if there were
0 dhcp static hosts (resulting in no reference to the hosts file on the
commandline), then one was later added, the commandline wouldn't have
linked dnsmasq up to the file, so even though we create it, dnsmasq
doesn't pay any attention.
The solution is to just always create these files and reference them
on the dnsmasq commandline (almost always, anyway). That way dnsmasq
can notice when a new entry is added at runtime (a SIGHUP is sent to
dnsmasq by virNetworkUdpate whenever a host entry is added or removed)
The exception to this is that the dhcp static hosts file isn't created
if there are no lease ranges *and* no static hosts. This is because in
this case dnsmasq won't be setup to listen for dhcp requests anyway -
in that case, if the count of dhcp hosts goes from 0 to 1, dnsmasq
will need to be restarted anyway (to get it listening on the dhcp
port). Likewise, if the dhcp hosts count goes from 1 to 0 (and there
are no dhcp ranges) we need to restart dnsmasq so that it will stop
listening on port 67. These special situations are handled in the
bridge driver's networkUpdate() by checking for ((bool)
nranges||nhosts) both before and after the update, and triggering a
dnsmasq restart if the before and after don't match.
https://bugzilla.redhat.com/show_bug.cgi?id=866364
pointed out a crash due to virNetworkObjAssignDef free'ing
network->newDef without NULLing it afterward. A fix for this is in
upstream commit b7e9202401. While the
NULLing of newDef was a legitimate fix, newDef should have already
been empty (NULL) anyway (as indicated in the comment that was deleted
by that commit).
The reason that newDef had a non-NULL value (i.e. the root cause) was
that networkStartNetwork() had failed after populating
network->newDef, but then neglected to free/NULL newDef in the
cleanup.
(A bit of background here: network->newDef should contain the
persistent config of a network when a network is active (and of course
only when it is persisten), and NULL at all other times. There is also
a network->def which should contain the persistent definition of the
network when it is inactive, and the current live state at all other
times. The idea is that you can make changes to network->newDef which
will take effect the next time the network is restarted, but won't
mess with the current state of the network (virDomainObj has a similar
pair of virDomainDefs that behave in the same fashion). Personally I
think there should be a network->live and network->config, and the
location of the persistent config should *always* be in
network->config, but that's for a later cleanup).
Since I love things to be symmetric, I created a new function called
virNetworkObjUnsetDefTransient(), which reverses the effects of
virNetworkObjSetDefTransient(). I don't really like the name of the
new function, but then I also didn't really like the name of the old
one either (it's just named that way to match a similar function in
the domain conf code).
Gcc with optimization warns:
../../src/qemu/qemu_driver.c: In function 'qemuDomainBlockCommit':
../../src/qemu/qemu_driver.c:12813:46: error: 'disk' may be used uninitialized in this function [-Werror=maybe-uninitialized]
../../src/qemu/qemu_driver.c:12698:25: note: 'disk' was declared here
cc1: all warnings being treated as errors
so obviously I had only been testing with optimization off.
* src/qemu/qemu_driver.c (qemuDomainBlockCommit): Guard cleanup.
I finally have all the pieces in place to perform a block-commit with
SELinux enforcing. There's still missing cleanup work when the commit
completes, but doing that requires tracking both the backing chain and
the base and top files within that chain in domain XML across libvirtd
restarts. Furthermore, from a security standpoint, once you have
granted access, you must assume any damage that can be done will be
done; later revoking access is nice to minimize the window of damage,
but less important as it does not affect the fact that damage can be
done in the first place. Therefore, deferring the revoke efforts until
we have better XML tracking of what chain operations are in effect,
including across a libvirtd restart, is reasonable.
* src/qemu/qemu_driver.c (qemuDomainBlockCommit): Label disks as
needed.
(qemuDomainPrepareDiskChainElement): Cast away const.
Previously, snapshot code did its own permission granting (lock
manager, cgroup device controller, and security manager labeling)
inline. But now that we are adding block-commit and block-copy
which also have to change permissions, it's better to reuse
common code for the task. While snapshot should fall back to
no access if read-write access failed, block-commit will want to
fall back to read-only access. The common code doesn't know
whether failure to grant read-write access should revert to no
access (snapshot, block-copy) or read-only access (block-commit).
This code can also be used to revoke access to unused files after
block-pull.
It might be nice to clean things up in a future patch by adding
new functions to the lock manager, cgroup manager, and security
manager that takes a single file name and applies context of a
disk to that file, rather than the current semantics of applying
context to the entire chain already associated to a disk. That
way, we could avoid the games this patch plays of temporarily
swapping out the disk->src and related fields of the disk. But
that would involve more code changes, so this patch really is
the smallest hack for doing the necessary work; besides, this
patch is more or less code motion (the hack was already employed
by the snapshot creation code, we are just making it reusable).
* src/qemu/qemu_driver.c (qemuDomainSnapshotCreateSingleDiskActive)
(qemuDomainSnapshotUndoSingleDiskActive): Refactor labeling hacks...
(qemuDomainPrepareDiskChainElement): ...into new function.
Now that we can crawl the chain of backing files, we can do
argument validation and implement the 'shallow' flag. In
testing this, I discovered that it can be handy to pass the
shallow flag and an explicit base, as a means of validating
that the base is indeed the file we expected.
* src/qemu/qemu_driver.c (qemuDomainBlockCommit): Crawl through
chain to implement shallow flag.
* src/libvirt.c (virDomainBlockCommit): Relax API.
This is the bare minimum to kick off a block commit. In particular,
flags support is missing (shallow requires us to crawl the backing
chain to determine the file name to pass to the qemu monitor command;
delete requires us to track what needs to be deleted at the time
the completion event fires). Also, we are relying on qemu to do
error checking (such as validating 'top' and 'base' as being members
of the backing chain), including the fact that the current qemu code
does not support committing the active layer (although it is still
planned to add that before qemu 1.3). Since the active layer won't
change, we have it easy and do not have to alter the domain XML.
Additionally, this will fail if SELinux is enforcing, because we fail
to grant qemu proper read/write access to the files it will modify.
* src/qemu/qemu_driver.c (qemuDomainBlockCommit): New function.
(qemuDriver): Register it.
qemu 1.3 will be adding a 'block-commit' monitor command, per
qemu.git commit ed61fc1. It matches nicely to the libvirt API
virDomainBlockCommit.
* src/qemu/qemu_capabilities.h (QEMU_CAPS_BLOCK_COMMIT): New bit.
* src/qemu/qemu_capabilities.c (qemuCapsProbeQMPCommands): Set it.
* src/qemu/qemu_monitor.h (qemuMonitorBlockCommit): New prototype.
* src/qemu/qemu_monitor_json.h (qemuMonitorJSONBlockCommit):
Likewise.
* src/qemu/qemu_monitor.c (qemuMonitorBlockCommit): Implement it.
* src/qemu/qemu_monitor_json.c (qemuMonitorJSONBlockCommit):
Likewise.
(qemuMonitorJSONHandleBlockJobImpl)
(qemuMonitorJSONGetBlockJobInfoOne): Handle new event type.
We used to walk the backing file chain at least twice per disk,
once to set up cgroup device whitelisting, and once to set up
security labeling. Rather than walk the chain every iteration,
which possibly includes calls to fork() in order to open root-squashed
NFS files, we can exploit the cache of the previous patch.
* src/conf/domain_conf.h (virDomainDiskDefForeachPath): Alter
signature.
* src/conf/domain_conf.c (virDomainDiskDefForeachPath): Require caller
to supply backing chain via disk, if recursion is desired.
* src/security/security_dac.c
(virSecurityDACSetSecurityImageLabel): Adjust caller.
* src/security/security_selinux.c
(virSecuritySELinuxSetSecurityImageLabel): Likewise.
* src/security/virt-aa-helper.c (get_files): Likewise.
* src/qemu/qemu_cgroup.c (qemuSetupDiskCgroup)
(qemuTeardownDiskCgroup): Likewise.
(qemuSetupCgroup): Pre-populate chain.
Technically, we should not be re-probing any file that qemu might
be currently writing to. As such, we should cache the backing
file chain prior to starting qemu. This patch adds the cache,
but does not use it until the next patch.
Ultimately, we want to also store the chain in domain XML, so that
it is remembered across libvirtd restarts, and so that the only
kosher way to modify the backing chain of an offline domain will be
through libvirt API calls, but we aren't there yet. So for now, we
merely invalidate the cache any time we do a live operation that
alters the chain (block-pull, block-commit, external disk snapshot),
as well as tear down the cache when the domain is not running.
* src/conf/domain_conf.h (_virDomainDiskDef): New field.
* src/conf/domain_conf.c (virDomainDiskDefFree): Clean new field.
* src/qemu/qemu_domain.h (qemuDomainDetermineDiskChain): New
prototype.
* src/qemu/qemu_domain.c (qemuDomainDetermineDiskChain): New
function.
* src/qemu/qemu_driver.c (qemuDomainAttachDeviceDiskLive)
(qemuDomainChangeDiskMediaLive): Pre-populate chain.
(qemuDomainSnapshotCreateSingleDiskActive): Uncache chain before
snapshot.
* src/qemu/qemu_process.c (qemuProcessHandleBlockJob): Update
chain after block pull.
In order to temporarily label files read/write during a commit
operation, we need to crawl the backing chain and find the absolute
file name that needs labeling in the first place, as well as the
name of the file that owns the backing file.
* src/util/storage_file.c (virStorageFileChainLookup): New
function.
* src/util/storage_file.h: Declare it.
* src/libvirt_private.syms (storage_file.h): Export it.
In order to search for a backing file name as literally present
in a chain, we need to remember if the chain had relative names.
Also, searching for absolute names is easier if we only have
to canonicalize once, rather than on every iteration.
* src/util/storage_file.h (_virStorageFileMetadata): Add field.
* src/util/storage_file.c (virStorageFileGetMetadataFromBuf):
(virStorageFileFreeMetadata): Manage it
(absolutePathFromBaseFile): Store absolute names in canonical form.
Requiring pre-allocation was an unusual idiom. It allowed iteration
over the backing chain to use fewer mallocs, but made one-shot
clients harder to read. Also, this makes it easier for a future
patch to move away from opening fds on every iteration over the chain.
* src/util/storage_file.h (virStorageFileGetMetadataFromFD): Alter
signature.
* src/util/storage_file.c (virStorageFileGetMetadataFromFD): Allocate
return value.
(virStorageFileGetMetadata): Update clients.
* src/conf/domain_conf.c (virDomainDiskDefForeachPath): Likewise.
* src/qemu/qemu_driver.c (qemuDomainGetBlockInfo): Likewise.
* src/storage/storage_backend_fs.c (virStorageBackendProbeTarget):
Likewise.
Previously, no one was using virStorageFileGetMetadata, and for good
reason - it couldn't support root-squash NFS. Change the signature
and make it useful to future patches, including enhancing the metadata
to recursively track the entire chain.
* src/util/storage_file.h (_virStorageFileMetadata): Add field.
(virStorageFileGetMetadata): Alter signature.
* src/util/storage_file.c (virStorageFileGetMetadata): Rewrite.
(virStorageFileGetMetadataRecurse): New function.
(virStorageFileFreeMetadata): Handle recursion.
Backing chains can end on a network protocol, such as nbd:xxx; we
should not attempt to probe the file system in this case.
* src/storage/storage_backend_fs.c (virStorageBackendProbeTarget):
Only probe files.
This is the last use of raw strings for disk formats throughout
the src/conf directory.
* src/conf/snapshot_conf.h (_virDomainSnapshotDiskDef): Store enum
rather than string for disk type.
* src/conf/snapshot_conf.c (virDomainSnapshotDiskDefClear)
(virDomainSnapshotDiskDefParseXML, virDomainSnapshotDefFormat):
Adjust users.
* src/qemu/qemu_driver.c (qemuDomainSnapshotDiskPrepare)
(qemuDomainSnapshotCreateSingleDiskActive): Likewise.
Express the default disk type as an enum, for easier handling.
* src/conf/capabilities.h (_virCaps): Store enum rather than
string for disk type.
* src/conf/domain_conf.c (virDomainDiskDefParseXML): Adjust
clients.
* src/qemu/qemu_driver.c (qemuCreateCapabilities): Likewise.
We have historically allowed 'aio' as a synonym for 'raw' for
back-compat to xen, but since a future patch will move to using
an enum value, we have to pick one to be our preferred output
name. This is a slight change in the output XML, but the sexpr
and xm outputs should still be identical, and the input XML can
still use either form.
* src/conf/domain_conf.c (virDomainDiskDefForeachPath): Move aio
back-compat...
(virDomainDiskDefParseXML): ...to parse time.
* src/xenxs/xen_sxpr.c (xenParseSxprDisks, xenFormatSxprDisk): ...and
to output time.
* src/xenxs/xen_xm.c (xenParseXM, xenFormatXMDisk): Likewise.
* tests/sexpr2xmldata/sexpr2xml-*.xml: Update tests.
When an image has no backing file, using VIR_STORAGE_FILE_AUTO
for its type is a bit confusing. Additionally, a future patch
would like to reserve a default value for the case of no file
type specified in the XML, but different from the current use
of -1 to imply probing, since probing is not always safe.
Also, a couple of file types were missing compared to supported
code: libxl supports 'vhd', and qemu supports 'fat' for directories
passed through as a file system.
* src/util/storage_file.h (virStorageFileFormat): Add
VIR_STORAGE_FILE_NONE, VIR_STORAGE_FILE_FAT, VIR_STORAGE_FILE_VHD.
* src/util/storage_file.c (virStorageFileMatchesVersion): Match
documentation when version probing not supported.
(cowGetBackingStore, qcowXGetBackingStore, qcow1GetBackingStore)
(qcow2GetBackingStoreFormat, qedGetBackingStore)
(virStorageFileGetMetadataFromBuf)
(virStorageFileGetMetadataFromFD): Take NONE into account.
* src/conf/domain_conf.c (virDomainDiskDefForeachPath): Likewise.
* src/qemu/qemu_driver.c (qemuDomainGetBlockInfo): Likewise.
* src/conf/storage_conf.c (virStorageVolumeFormatFromString): New
function.
(poolTypeInfo): Use it.
Relabeling tapfd right after the tap device is created.
qemuPhysIfaceConnect is common function called both for static
netdevs and for hotplug netdevs.
Having hostuuid in migration cookie is a nice bonus since it provides an
easy way of detecting migration to the same host. However, requiring it
breaks backward compatibility with older libvirt releases.
Recently, patches were added support for (managed)saving, restoring, and
migrating domains with host USB devices. However, qemu driver would
still forbid migration of such domains because qemuMigrationIsAllowed
was not updated.
If we can't probe the architecture from QMP we parse the architecture
from the qemu binaries name. This results in the architecture being i386
instead of i686 which then results in QEMU_CAPS_PCI_MULTIBUS being unset
which gives a broken qemu command line.
This probably didn't show up earlier since most of the time there's also
a /usr/bin/qemu around which results in i686 capabilities.
which frees all allocated memory but doesn't set the passed pointer to
NULL. Therefore, we must do it ourselves. This is causing actual
libvirtd crash: Basically, when doing 'virsh net-edit' the newDef should
be dropped. And the memory is freed, indeed. However, the pointer is
not set to NULL but kept instead. And the next duo of calls 'virsh
net-start' and 'virsh net-destroy' starts the disaster. The latter one
does the same as 'virsh destroy'; it sees that newDef is nonNULL so it
replaces def with newDef (which has been freed already as said a few
lines above). Therefore any subsequent call accessing def will hit the ground.
