Pavel Hrdina [Wed, 19 Jul 2017 15:01:56 +0000 (17:01 +0200)]
qemu: switch QEMU capabilities to use virFileCache
The switch contains considerable amount of changes:
virQEMUCapsRememberCached() is removed because this is now handled
by virFileCacheSave().
virQEMUCapsInitCached() is removed because this is now handled by
virFileCacheLoad().
virQEMUCapsNewForBinary() is split into two functions,
virQEMUCapsNewData() which creates new data if there is nothing
cached and virQEMUCapsLoadFile() which loads the cached data.
This is now handled by virFileCacheNewData().
virQEMUCapsCacheValidate() is removed because this is now handled by
virFileCacheValidate().
virQEMUCapsCacheFree() is removed because it's no longer required.
Add virCapsPtr into virQEMUCapsCachePriv because for each call of
virFileCacheLookup*() we need to use current virCapsPtr.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com> Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Pavel Hrdina [Fri, 21 Jul 2017 13:09:54 +0000 (15:09 +0200)]
qemu: pass only host arch instead of the whole virCaps
This is a preparation for following patches where we switch to
virFileCache for QEMU capabilities cache
The host arch will always remain the same but virCaps may change. Now
the host arch is stored while creating new qemu capabilities cache.
It removes the need to pass virCaps into virQEMUCapsCache*() functions.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com> Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Pavel Hrdina [Fri, 21 Jul 2017 12:24:51 +0000 (14:24 +0200)]
tests: rewrite host CPU mocking
Move all the host CPU data into a separate file and rewrite qemucpumock
to not use passed @caps. This is preparation for following patch which
will replace virCaps argument with virArch.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com> Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Pavel Hrdina [Fri, 7 Apr 2017 15:34:53 +0000 (17:34 +0200)]
util: introduce virFileCache
The new virFileCache will nicely handle the caching logic for any data
that we would like to cache. For each type of data we will just need
to implement few handlers that will take care of creating, validating,
loading and saving the cached data.
The cached data must be an instance of virObject.
Currently we cache QEMU capabilities which will start using
virFileCache.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com> Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Scott Garfinkle [Tue, 25 Jul 2017 14:33:50 +0000 (09:33 -0500)]
Generate unique socket file
It's possible to have more than one unnamed virtio-serial unix channel.
We need to generate a unique name for each channel. Currently, we use
".../unknown.sock" for all of them. Better practice would be to specify
an explicit target path name; however, in the absence of that, we need
uniqueness in the names we generate internally.
Before the changes we'd get /var/lib/libvirt/qemu/channel/target/unknown.sock
for each instance of
<channel type='unix'>
<source mode='bind'/>
<target type='virtio'/>
</channel>
Now, we get vioser-00-00-01.sock, vioser-00-00-02.sock, etc.
Move machineName generation from virsystemd into domain_conf
It is more related to a domain as we might use it even when there is
no systemd and it does not use any dbus/systemd functions. In order
not to use code from conf/ in util/ pass machineName in cgroups code
as a parameter. That also fixes a leak of machineName in the lxc
driver and cleans up and de-duplicates some code.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
John Ferlan [Thu, 1 Jun 2017 16:43:06 +0000 (12:43 -0400)]
secret: Handle object list removal and deletion properly
Rather than rely on virSecretObjEndAPI to make the final virObjectUnref
after the call to virSecretObjListRemove, be more explicit by calling
virObjectUnref and setting @obj to NULL for secretUndefine and in
the error path of secretDefineXML. Calling EndAPI will end up calling
Unlock on an already unlocked object which has indeteriminate results
(usually an ignored error).
The virSecretObjEndAPI will both Unref and Unlock the object; however,
the virSecretObjListRemove would have already Unlock'd the object so
calling Unlock again is incorrect. Once the virSecretObjListRemove
is called all that's left is to Unref our interest since that's the
corrollary to the virSecretObjListAdd which returned our ref interest
plus references for each hash table in which the object resides. In math
terms, after an Add there's 2 refs on the object (1 for the object and
1 for the list). After calling Remove there's just 1 ref on the object.
For the Add callers, calling EndAPI removes the ref for the object and
unlocks it, but since it's in a list the other 1 remains.
John Ferlan [Tue, 25 Jul 2017 13:02:09 +0000 (09:02 -0400)]
secret: Fix memory leak in virSecretLoad
If the virSecretLoadValue fails, the code jumped to cleanup without
setting @ret = obj, thus calling virSecretObjListRemove which only
accounts for the object reference related to adding the object to
the list during virSecretObjListAdd, but does not account for the
reference to the object itself as the return of @ret would be NULL
so the caller wouldn't call virSecretObjEndAPI on the object recently
added thus reducing the refcnt to zero.
This patch will perform the ObjListRemove in the failure path of
virSecretLoadValue and Unref @obj in order to perform clean up
and return @obj as NULL. The @def will be freed as part of the
virObjectUnref.
John Ferlan [Thu, 1 Jun 2017 12:17:52 +0000 (08:17 -0400)]
secret: Properly handle @def after virSecretObjAdd in driver
Since the virSecretObjListAdd technically consumes @def on success,
the secretDefineXML should set @def = NULL immediately and process
the remaining calls using a new @objDef variable. We can use use
VIR_STEAL_PTR since we know the Add function just stores @def in
obj->def.
Because we steal @def into @objDef, if we jump to restore_backup:
and @backup is set, then we need to ensure the @def would be
free'd properly, so we'll steal it back from @objDef. For the other
condition this fixes a double free of @def if the code had jumped to
@backup == NULL thus calling virSecretObjListRemove without setting
@def = NULL. In this case, the subsequent call to DefFree would
succeed and free @def; however, the call to EndAPI would also
call DefFree because the Unref done would be the last one for
the @obj meaning the obj->def would be used to call DefFree,
but it's already been free'd because @def wasn't managed right
within this error path.
As it turns out, this is not portable and very Linux & glibc
specific. Worse, this may lead to not starving writers on Linux
but everywhere else. Revert this and if the starvation occurs
resolve it.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Daniel P. Berrange <berrange@redhat.com>
Andrea Bolognani [Wed, 19 Jul 2017 08:37:04 +0000 (10:37 +0200)]
conf: Fix backwards migration of pSeries guests
Recent commits made it so that pci-root controllers for
pSeries guests are automatically assigned the
spapr-pci-host-bridge model name; however, that prevents
guests to migrate to older versions of libvirt which don't
know about that model name at all, which at the moment is
all of them :)
To avoid the issue, just strip the model name from PHBs
when formatting the migratable XML; guests that use more
than one PHB are not going to be migratable anyway.
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
If the parent provided for the storage pool adapter is not vHBA
capable, then issue a configuration error even though the provided
wwnn/wwpn were found.
It is a configuration error to provide a mismatched parent to
the wwnn/wwpn. The @parent is optional and is used as a means to
perform duplicate pool source checks.
Commit id '106930aaa' altered the order of checking for an existing
vHBA (e.g something created via nodedev-create functionality outside
of the storage pool logic) which inadvertantly broke the code to
decide whether to alter/force the fchost->managed field to be 'yes'
because the storage pool will be managing the created vHBA in order
to ensure when the storage pool is destroyed that the vHBA is also
destroyed.
This patch moves the check (and checkParent helper) for an existing
vHBA back into the createVport in storage_backend_scsi. It also
adjusts the checkParent logic to more closely follow the intentions
prior to commit id '79ab0935'. The changes made by commit id '08c0ea16f'
are only necessary to run the virStoragePoolFCRefreshThread when
a vHBA was really created because there's a timing lag such that
the refreshPool call made after a startPool from storagePoolCreate*
wouldn't necessarily find LUNs, but the thread would. For an already
existing vHBA, using the thread is unnecessary since the vHBA already
exists and the lag to configure the LUNs wouldn't exist.
John Ferlan [Thu, 29 Jun 2017 20:51:54 +0000 (16:51 -0400)]
nodedev: Remove driver locks around object list mgmt code
Since virnodedeviceobj now has a self-lockable hash table, there's no
need to lock the table from the driver for processing. Thus remove the
locks from the driver for NodeDeviceObjList mgmt.
John Ferlan [Thu, 29 Jun 2017 19:50:06 +0000 (15:50 -0400)]
nodedev: Convert virNodeDeviceObjListPtr to use hash tables
Rather than use a forward linked list of elements, it'll be much more
efficient to use a hash table to reference the elements by unique name
and to perform hash searches.
This patch does all the heavy lifting of converting the list object to
use a self locking list that contains the hash table. Each of the FindBy
functions that do not involve finding the object by it's key (name) is
converted to use virHashSearch in order to find the specific object.
When searching for the key (name), it's possible to use virHashLookup.
For any of the list perusal functions that are required to evaluate
each object, the virHashForEach function is used.
John Ferlan [Thu, 20 Jul 2017 13:07:52 +0000 (09:07 -0400)]
nodedev: Alter node device deletion logic
Alter the node device deletion logic to make use of the parent field
from the obj->def rather than call virNodeDeviceObjListGetParentHost.
As it turns out the saved @def won't have parent_wwnn/wwpn or
parent_fabric_wwn, so the only logical path would be to call
virNodeDeviceObjListGetParentHostByParent which we can accomplish
directly via virNodeDeviceObjListFindByName.
Michal Privoznik [Wed, 19 Jul 2017 10:11:05 +0000 (12:11 +0200)]
virdomainobjlist: Use virObjectRWLockable
There is no reason why two threads trying to look up two domains
should mutually exclude each other. Utilize new
virObjectRWLockable that was just introduced.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Michal Privoznik [Wed, 19 Jul 2017 07:31:50 +0000 (09:31 +0200)]
virobject: Introduce virObjectRWLockable
Up until now we only had virObjectLockable which uses mutexes for
mutually excluding each other in critical section. Well, this is
not enough. Future work will require RW locks so we might as well
have virObjectRWLockable which is introduced here.
Moreover, polymorphism is introduced to our code for the first
time. Yay! More specifically, virObjectLock will grab a write
lock, virObjectLockRead will grab a read lock then (what a
surprise right?). This has great advantage that an object can be
made derived from virObjectRWLockable in a single line and still
continue functioning properly (mutexes can be viewed as grabbing
write locks only). Then just those critical sections that can
grab a read lock need fixing. Therefore the resulting change is
going to be way smaller.
In order to avoid writer starvation, the object initializes RW
lock that prefers writers.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Michal Privoznik [Wed, 19 Jul 2017 10:54:23 +0000 (12:54 +0200)]
virthread: Introduce virRWLockInitPreferWriter
We already have virRWLockInit. But this uses pthread defaults
which prefer reader to initialize the RW lock. This may lead to
writer starvation. Therefore we need to have the counterpart that
prefers writers. Now, according to the
pthread_rwlockattr_setkind_np() man page setting
PTHREAD_RWLOCK_PREFER_WRITER_NP attribute is no-op. Therefore we
need to use PTHREAD_RWLOCK_PREFER_WRITER_NONRECURSIVE_NP
attribute. So much for good enum value names.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
qemu_capabilities: Fix the formatting with a space
It was observed while adding new property that there should be a space
before closing a curly brace in intel-iommu object property definition.
Fixing it as a separate patch.
Signed-off-by: Shivaprasad G Bhat <sbhat@linux.vnet.ibm.com>
Michal Privoznik [Sat, 22 Jul 2017 07:12:36 +0000 (09:12 +0200)]
docs: Span cells if there's not doc text for enum val
When generating HTML documentation we put enum values into a
table so that we can display the value's name, numerical value
and description (if it has one). Now the last part is problem. If
the value doesn't have description the table row has just two
cells and if it has one the row counts three cells. This makes
HTML engines render the description into very little space - for
instance see:
Michal Privoznik [Sat, 22 Jul 2017 06:05:12 +0000 (08:05 +0200)]
apibuild.py: Handle enum comments properly
After f4cb85c6aff7c1d90 we only have two options for placing enum
values descriptions. It's either:
typedef enum {
/* Some long description. Therefore it's placed before
* the value. */
VIR_ENUM_A_VAL = 1,
} virEnumA;
or:
typedef enum {
VIR_ENUM_B_VAL = 1, /* Some short description */
} virEnumB;
However, our apibuild.py script is not able to deal with the
former one. It messes up comments. To fix this couple of things
needs to be done:
a) DO NOT reset self.comment in parseEnumBlock(). This is a
result from our tokenizer. Upon calling token() if it finds a
comment block it stores it in self.comment and returns the next
token (which is not comment). Therefore, if we reset self.comment
we might lose the first comment in the enum block.
b) we need a variable to track if the current enum block uses
value descriptions before or after values. That is if it's type
virEnumA or virEnumB. Depending on that, it we're dealing with
virEnumA type and the current token is a comma ',' we can add the
value into the list as we already have everything needed:
comment, name and value.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Michal Privoznik [Sat, 22 Jul 2017 08:30:19 +0000 (10:30 +0200)]
qemu_capabilities: Honour caps values formatting
So the way we format this huge virQEMUCaps enum is we group the
values in groups of five. And then at the beginning of each group
we have a small comment that says what's the number of the first
item in the group. Well, the last commit of 11b2ebf3e152 does not
follow this formatting.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
John Ferlan [Tue, 9 May 2017 11:00:41 +0000 (07:00 -0400)]
storage: Alter volume num, name, and export API's to just take obj
Alter the virStoragePoolObjNumOfVolumes, virStoragePoolObjVolumeGetNames,
and virStoragePoolObjVolumeListExport APIs to take a virStoragePoolObjPtr
instead of the &obj->volumes and obj->def.
John Ferlan [Fri, 5 May 2017 21:31:36 +0000 (17:31 -0400)]
test: Cleanup exit/failure paths of some storage pool APIs
Rework some of the test driver API's to remove the need to return
failure when testStoragePoolObjFindByName returns NULL rather than
going to cleanup. This removes the need for check for "if (obj)" and in
some instances the need to for a cleanup label and a local ret variable.
dann frazier [Thu, 20 Jul 2017 19:56:55 +0000 (13:56 -0600)]
qemu: Add AAVMF32 to the list of known UEFIs
Add a path for UEFI VMs for AArch32 VMs, based on the path Debian is using.
libvirt is the de facto canonical location for defining where distros
should place these firmware images, so let's define this path here to try
and minimize distro fragmentation.
Andrea Bolognani [Tue, 27 Jun 2017 06:30:58 +0000 (08:30 +0200)]
conf: Move some virDomainDeviceInfo functions
The virDomainDeviceInfo struct is defined in device_conf,
so generic functions that operate on it should also be
defined there rather than in domain_conf.
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
This patch addresses the same aspects on PPC the bug 1103314 addressed
on x86.
PCI expander bus creates multiple primary PCI busses, where each of these
busses can be assigned a specific NUMA affinity, which, on x86 is
advertised through ACPI on a per-bus basis.
For SPAPR, a PHB's NUMA affinities are assigned on a per-PHB basis, and
there is no mechanism for advertising NUMA affinities to a guest on a
per-bus basis. So, even if qemu-ppc manages to get some sort of multi-bus
topology working using PXB, there is no way to expose the affinities
of these busses to the guest. It can only be exposed on a per-PHB/per-domain
basis.
So patch enables NUMA node tag in pci-root controller on PPC.
The way to set the NUMA node is through the numa_node option of
spapr-pci-host-bridge device. However for the implicit PHB, the only way
to set the numa_node is from the -global option. The -global option applies
to all the PHBs unless explicitly specified with the option on the
respective PHB of CLI. The default PHB has the emulated devices only, so
the patch prevents setting the NUMA node for the default PHB.
Signed-off-by: Shivaprasad G Bhat <sbhat@linux.vnet.ibm.com> Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Andrea Bolognani [Fri, 21 Jul 2017 10:09:58 +0000 (12:09 +0200)]
qemu: Clean up firmware list initialization
Instead of going through two completely different code paths,
one of which repeats the same hardcoded bit of information
three times in rapid succession, depending on whether or not
a firmware list has been provided at configure time, just
provide a reasonable default value and remove the extra code.
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Michal Privoznik [Fri, 21 Jul 2017 10:46:18 +0000 (12:46 +0200)]
libvirt-domain.h: Fix enum description placement
There are only two acceptable places for describing enum values.
It's either:
typedef enum {
/* Some long description. Therefore it's placed before
* the value. */
VIR_ENUM_A_VAL = 1,
} virEnumA;
or:
typedef enum {
VIR_ENUM_B_VAL = 1, /* Some short description */
} virEnumB;
However, during review of a patch sent upstream I realized that
is not always the case. I went through all the public header
files and identified all the offenders. Luckily there were just
two of them.
Yes, this makes our HTML generated documentation broken, but
that's bug of the generator. Our header files shouldn't be forced
to use something we don't want to.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Peter Krempa [Wed, 12 Jul 2017 11:59:35 +0000 (13:59 +0200)]
qemu: process: Don't put memoryless NUMA nodes into autoNodeset
'numad' may return a nodeset which contains NUMA nodes without memory
for certain configurations. Since cgroups code will not be happy using
nodes without memory we need to store only numa nodes with memory in
autoNodeset.
On the other hand autoCpuset should contain cpus also for nodes which
do not have any memory.
Antoine Millet [Mon, 17 Jul 2017 15:49:00 +0000 (17:49 +0200)]
Handle hotplug change on VLAN configuration using OVS
A new function virNetDevOpenvswitchUpdateVlan has been created to instruct
OVS of the changes. qemuDomainChangeNet has been modified to handle the
update of the VLAN configuration for a running guest and rely on
virNetDevOpenvswitchUpdateVlan to do the actual update if needed.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
qemu: shared disks with cache=directsync should be safe for migration
At present shared disks can be migrated with either readonly or cache=none. But
cache=directsync should be safe for migration, because both cache=directsync and cache=none
don't use the host page cache, and cache=direct write through qemu block layer cache.
Signed-off-by: Peng Hao <peng.hao2@zte.com.cn> Reviewed-by: Wang Yechao <wang.yechao255@zte.com.cn>
Peter Krempa [Tue, 11 Jul 2017 15:45:12 +0000 (17:45 +0200)]
qemu: blockcopy: Refactor logic checking the target storage file
Use virStorageSource accessors to check the file and call
virStorageFileAccess before even attempting to stat the target. This
will be helpful once we try to add network destinations for block copy,
since there will be no need to stat them.
Peter Krempa [Tue, 11 Jul 2017 06:23:38 +0000 (08:23 +0200)]
qemu: blockcopy: Explicitly assert 'reuse' for block devices
When copying to a block device, the block device will already exist. To
allow users using a block device without any preparation, they need to
use the block copy without VIR_DOMAIN_BLOCK_COPY_REUSE_EXT.
This means that if the target is an existing block device we don't need
to prepare it, but we can't reject it as being existing.
To avoid breaking this feature, explicitly assume that existing block
devices will be reused even without that flag explicitly specified,
while skipping attempts to create it.
qemuMonitorDriveMirror still needs to honor the flag as specified by the
user, since qemu overwrites the metadata otherwise.
Peter Krempa [Mon, 3 Jul 2017 13:16:02 +0000 (15:16 +0200)]
tests: virjson: Remove spaces from 'very-hard' parsing example
The example is rather long and upcomming patch will check whether the
string can be formatted back. As the formatted string lacks spaces and
adding the 'expect' string with spaces would be rather long, just drop
spaces from this test case.
There are other test cases which do contain spaces.
projects / libvirt.git / log