Erik Skultety [Thu, 28 May 2015 15:00:01 +0000 (17:00 +0200)]
storage: Don't update volume objs list before we successfully create one
We do update pool volume object list before we actually create any
volume. If buildVol fails, we then try to delete the volume in the
storage as well as remove it from our structures. The problem is, that
any backend that supports both buildVol and deleteVol would fail in this
case which is completely unnecessary. This patch causes the update to
take place after we know a volume has been created successfully, thus no
removal in case of a buildVol failure is necessary.
Erik Skultety [Tue, 2 Jun 2015 07:25:04 +0000 (09:25 +0200)]
nwfilter: Fix sscanf off-by-one error in virNWFilterSnoopLeaseFileLoad
We allocate 16 bytes for IPv4 address and 55 bytes for interface
key, therefore we should read up to 15/54 bytes and let the last byte
reserved for terminating null byte in sscanf.
Eric Blake [Tue, 26 May 2015 23:34:10 +0000 (17:34 -0600)]
maint: update to latest gnulib
Incorporates fixes for cross-compiling to mingw on rawhide, where
gcc 5.1 changes detection of how to properly determine PRIdMAX.
Also picks up some improvements for compilation on Mac OS X.
Jim Fehlig [Fri, 1 May 2015 18:39:30 +0000 (12:39 -0600)]
spec: fixes for recently introduced libxl conf files
Commit 198cc1d3 introduced libxl-lockd and libxl-sanlock config
files but forgot to add them to the spec file. Follow-up commit 62b18d98 added the files to daemon-driver-libxl, but missed adding
them to the daemon package when configuring libvirt
--without-driver-modules. In addition, commit 62b18d98 added
libxl-sanlock to daemon-driver-libxl, but it should be included
in lock-sanlock when libvirt is configured --with-sanlock.
Ján Tomko [Wed, 27 May 2015 14:46:01 +0000 (16:46 +0200)]
virsh: make negative values with vol-resize more convenient
When shrinking a volume by a certain size, instead of typing
vol-resize volume 1G --delta --shrink
we allow the convience of specifying a negative value:
vol-resize volume -1G --delta --shrink
getting the same results with one more character.
A negative value only makes sense as a delta. Imply the
--delta parameter if the value is negative.
Still require --shrink, because the operation is potentially
destructive.
Andrea Bolognani [Thu, 28 May 2015 14:39:13 +0000 (16:39 +0200)]
qemu: Automatically add <panic> element for pSeries guests.
The guest firmware provides the same functionality as the pvpanic
device, and the relevant element should always be present in the
domain XML to reflect this fact, so add it after parsing the
definition if it wasn't there already.
Andrea Bolognani [Thu, 28 May 2015 14:39:12 +0000 (16:39 +0200)]
qemu: Allow panic device for pSeries guests
The guest firmware provides the same functionality as the pvpanic
device, which is not available in QEMU on pSeries, so the domain
XML should be allowed to contain the <panic> element.
On the other hand, unlike the pvpanic device, the guest firmware
can't be configured, so report an error if an address has been
provided in the XML.
commit id 'bd00e00e' neglected to add the new adapter source options
into the if condition that allowed printing the <source> XML fields.
The <adapter type='fc_host'.../> doesn't require other options in order
to be complete.
Ján Tomko [Fri, 29 May 2015 13:03:24 +0000 (15:03 +0200)]
Allocate priv->vioserialaddrs unconditionally
When attempting to hotplug a virtio-serial console to a domain
that had no virtio-serial controllers (not even those that
are added by libvirt when some devices need them) at daemon startup,
report a user-friendly error:
error: Failed to attach device from console.xml
error: internal error: no virtio-serial controllers are available
instead of crashing the daemon:
Process terminating with default action of signal 11 (SIGSEGV): dumping core
Access not within mapped region at address 0x8
at 0x531028F: virDomainVirtioSerialAddrNext (domain_addr.c:916)
by 0x531028F: virDomainVirtioSerialAddrAssign (domain_addr.c:1029)
by 0x1CBF68: qemuDomainAttachChrDevice (qemu_hotplug.c:1565)
by 0x1BCD5E: qemuDomainAttachDeviceLive (qemu_driver.c:7997)
by 0x1BCD5E: qemuDomainAttachDeviceFlags (qemu_driver.c:8743)
Ján Tomko [Fri, 29 May 2015 13:06:14 +0000 (15:06 +0200)]
Properly free the xmlDocPtr when loading pool state
Use xmlFreeDoc instead of plain xmlFree.
4 bytes in 1 blocks are definitely lost in loss record 9 of 1,084
at 0x4C29F80: malloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
by 0x70730D6: xmlStrndup (in /usr/lib64/libxml2.so.2.9.2)
by 0x701E3DC: xmlNewDoc (in /usr/lib64/libxml2.so.2.9.2)
by 0x70C39F8: xmlSAX2StartDocument (in /usr/lib64/libxml2.so.2.9.2)
by 0x7017245: xmlParseDocument (in /usr/lib64/libxml2.so.2.9.2)
by 0x7017606: xmlDoRead (in /usr/lib64/libxml2.so.2.9.2)
by 0x5309DAD: virXMLParseHelper (virxml.c:742)
by 0x5367584: virStoragePoolLoadState (storage_conf.c:1863)
Jim Fehlig [Mon, 27 Apr 2015 18:32:55 +0000 (12:32 -0600)]
libxl: populate build_info vfb in separate function
For HVM domains, vfb info must be populated in the libxl_domain_build_info
struct. Currently this is done in the libxlMakeVfbList function, but IMO
it would be cleaner to populate the build_info vfb in a separate
libxlMakeBuildInfoVfb function. libxlMakeVfbList would then handle only
vfb devices, simiar to the other libxlMake<device>List functions.
A future patch will extend libxlMakeBuildInfoVfb to support SPICE.
The disk pool recalculates the pool allocation, capacity, and available
values each time through processing a newly created disk partition. This
created an issue with the allocation setting since the code used is shared
with the refresh path. Each path calls virStorageBackendDiskReadPartitions
which initializes the pool values and then processes the partition table
from the 'libvirt_parthelper' utility output with the only difference being
create passes a specific volume to be processed while refresh pass a NULL
indicating to process all volumes. That passed volume is check during the
virStorageBackendDiskMakeVol call to see if the current partition described
by the volume key already exists. If it exists, then no adjustments are
made to the allocation and the next entry in the output is checked.
For the create path this resulted in only the most recently created
partition size would be accounted for in the 'allocation' setting. This
patch thus checks whether the incoming volume is NULL before clearing
the pool allocation value.
John Ferlan [Tue, 26 May 2015 13:23:28 +0000 (09:23 -0400)]
storage: Don't adjust pool alloc/avail values for disk backend
Commit id '2ac0e647' for https://bugzilla.redhat.com/show_bug.cgi?id=1206521
was meant to be a generic check for the CreateVol, CreateVolFrom, and
DeleteVol paths to check if the storage backend's changed the pool's view
of allocation or available values.
Unfortunately as it turns out this caused a side effect when the disk backend
created an extended partition there would be no actual storage removed from
the pool, thus the changes would not find any change in allocation or
available and incorrectly update the pool values using the size of the
extended partition. A subsequent refresh of the pool would reset the
values appropriately.
This patch modifies those checks in order to specifically not update the
pool allocation and available for only the disk backend rather than be
generic before and after checks.
Laine Stump [Wed, 27 May 2015 17:59:57 +0000 (13:59 -0400)]
node_device: more informative error log when device isn't found
In a couple of cases, the node device driver (and the test node device
driver which likely copied it) was only logging "Node device not
found" when it couldn't find the requested device. This patch changes
those cases to log the name (and in the case when it's relevant, the
wwnn and wwpn) as well.
virsh: Fix to list online cpus using virsh capabilities
Virsh capabilities will list offline cpus as online when
libvirt is compiled with numactl option disabled. This
fix will list correct set of online cpus.
Ján Tomko [Wed, 27 May 2015 14:33:22 +0000 (16:33 +0200)]
Fix shrinking volumes with the delta flag
This never worked.
In 0.9.10 when this API was introduced, it was intended that
the SHRINK flag combined with DELTA would shrink the volume by
the specified capacity (to avoid passing negative numbers).
See commit 055bbf4.
When the SHRINK flag was finally implemented for the first backend
in 1.2.13 (commit aa9aa6a), it was only implemented for the absolute
values and with the delta flag the volume is always extended,
regardless of the SHRINK flag.
Treat the SHRINK flag as a minus sign when used together with DELTA,
to allow shrinking volumes as was documented in the API since 0.9.10.
Ján Tomko [Wed, 27 May 2015 13:39:07 +0000 (15:39 +0200)]
Simplify allocation check in storageVolResize
Since shrinking a volume below existing allocation is not allowed,
it is not possible for a successful resize with VOL_RESIZE_ALLOCATE
to increase the pool's available value.
Even with the SHRINK flag it is possible to extend the current
allocation or even the capacity. Remove the overflow when
computing delta with this flag and do the check even if the
flag was specified.
Maxim Nestratov [Tue, 26 May 2015 14:57:00 +0000 (17:57 +0300)]
parallels: suppress console output from parallels SDK
It is necessary to have unpolluted screen when connecting to
parallels driver via virsh.
Otherwise a lot of unexpected output one will get on the console.
Lubomir Rintel [Wed, 27 May 2015 17:30:50 +0000 (19:30 +0200)]
interface: don't error out if a bond has no interfaces
It's not a problem at all and causes virt-manager to break down.
Note: netcf 0.2.8 and earlier generates invalid XML for a bond with no
interfaces anyway, so in that case this error in libvirt is never
reached since we fail earlier.
Andrea Bolognani [Tue, 26 May 2015 16:13:40 +0000 (18:13 +0200)]
qemu: Limit rtc-reset-reinjection requirement to x86 only.
The QMP command, like the interrupt reinjection logic it's connected
to, is only implemented in QEMU when TARGET_I386 is defined, so
checking for its availability on any other architecture is pointless.
On the other hand, when we're on x86, we shouldn still make sure that
rtc-reset-reinjection is available and refuse to set the time
otherwise.
The code already exists there, it just modified different flags. I just
noticed this when looking at the code. This patch is better to view
with bigger context or '-W'.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Zhang Bo [Mon, 25 May 2015 06:22:42 +0000 (14:22 +0800)]
util: make it more robust to calculate timeout value
When we change system clock to years ago, a certain CPU may use up 100% cputime.
The reason is that in function virEventPollCalculateTimeout(), we assign the
unsigned long long result to an INT variable,
*timeout = then - now; // timeout is INT, and then/now are long long
if (*timeout < 0)
*timeout = 0;
there's a chance that variable @then minus variable @now may be a very large number
that overflows INT value expression, then *timeout will be negative and be assigned to 0.
Next the 'poll' in function virEventPollRunOnce() will get into an 'endless' while loop there.
thus, the cpu that virEventPollRunOnce() thread runs on will go up to 100%.
Although as we discussed before in https://www.redhat.com/archives/libvir-list/2015-May/msg00400.html
it should be prohibited to set-time while other applications are running, but it does
seems to have no harm to make the codes more robust.
Signed-off-by: Wang Yufei <james.wangyufei@huawei.com> Signed-off-by: Zhang Bo <oscar.zhangbo@huawei.com>
Peter Krempa [Mon, 25 May 2015 15:14:57 +0000 (17:14 +0200)]
qemu: Fix compilation error when enum variable size differs from 'int'
Since commit bcd9a564b631aa virDomainNumatuneGetMode returns the value
via a pointer rather than in the return value. The change triggered
problems with platforms where the compiler decides to use a data type of
size different than integer at the point where we typecast it.
Work around the issue by using an intermediate variable of the correct
type that gets casted back by the default typecasting rules.
Eric Blake [Tue, 26 May 2015 03:33:24 +0000 (21:33 -0600)]
maint: update to latest gnulib
Time to update to new gnulib before a release.
gcc 5.1 introduced a new -Wformat-signedness, and new gnulib now
turns it on by default. However, it is still rather lame at the
moment, because it warns for enums, even though there is no way
to control the signeness of an enum which does not use any members
that are negative or larger than INT_MAX, and even though such an
enum would always print the same for both %d and %u:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66249
In file included from ../../src/util/virarch.c:26:0:
../../src/util/virarch.c: In function 'virArchFromHost':
../../src/util/virarch.c:180:15: error: format '%d' expects argument of type 'int', but argument 9 has type 'unsigned int' [-Werror=format=]
VIR_DEBUG("Mapped %s to %d (%s)",
So this patch turns off the new warning as part of enabling all
other new gcc 5.1 warnings that gnulib now enables.
* .gnulib: Update to latest, in part for gcc 5.1 interaction.
* m4/virt-compile-warnings.m4: Ignore -Wformat-signedness, for now.
John Ferlan [Sat, 23 May 2015 14:19:20 +0000 (10:19 -0400)]
qemu: Add libvirt version check to refresh capabilities algorithm
Rather than an algorithm based solely on libvirtd ctime to refresh the
capabilities add the element of the libvirt build version into the equation.
Since that version wouldn't be there prior to this code being run - don't
fail on reading the capabilities if not found. In this case, the cache
will always be rebuilt when a new libvirt version is installed.
Original commit id 'cbde3589' indicates that the cache file would be
discarded if either the QEMU binary or libvirtd 'ctime' changes; however,
the code only discarded if the QEMU binary time didn't match or if the
new libvirtd ctime was later than what created the cache file.
Since many factors come into play with 'ctime' adjustments (including
perhaps turning back the hands of time), change the logic to also force
a refresh if the ctime of libvirt is different than what's in the cache.
John Ferlan [Mon, 18 May 2015 13:16:53 +0000 (09:16 -0400)]
qemu: Resolve Coverity RESOURCE_LEAK
Recent changes to the -M/--machine processing code in qemuParseCommandLine
caused Coverity to determine there was a possible resource leak with how
the 'list' is managed. Rather than try to add virStringFreeList calls
everywhere - just promote list to the top of the variables and free it
within the error processing code. Also required a couple of other tweaks
in order to avoid double free's.
John Ferlan [Mon, 18 May 2015 12:42:46 +0000 (08:42 -0400)]
conf: Resolve Coverity NEGATIVE_RETURNS
Commit id '73eda710' added virDomainKeyWrapDefParseXML which uses
virXPathNodeSet, but does not handle a -1 return thus causing a possible
loop condition exit problem later when the return value is used.
Change the logic to return the value from virXPathNodeSet if <= 0
storage: conf: Don't set any default <mode> in the XML
The XML parser sets a default <mode> if none is explicitly passed in.
This is then used at pool/vol creation time, and unconditionally reported
in the XML.
The problem with this approach is that it's impossible for other code
to determine if the user explicitly requested a storage mode. There
are some cases where we want to make this distinction, but we currently
can't.
Handle <mode> parsing like we handle <owner>/<group>: if no value is
passed in, set it to -1, and adjust the internal consumers to handle
it.
Cole Robinson [Thu, 21 May 2015 19:29:39 +0000 (15:29 -0400)]
docs: formatstorage: Update <permissions> docs
- Don't redocument the permissions fields for backingstore, just point to
the volume docs.
- Clarify that owner/group are inherited from the parent directory at
volume create/pool build time.
- Clarify that <permissions> fields report runtime values too
Maxim Nestratov [Thu, 21 May 2015 13:49:35 +0000 (16:49 +0300)]
parallels: fix possible crash in case of errors in prlsdkLoadDomain
Cleanup code in prlsdkLoadDomain doesn't take into account the fact
if private domain structure along with freeing function is assigned
or not. In case it is, we shouldn't call it manually because
virDomainObjListRemove calls it and frees pdom.
Also, allocated def structure should be freed only if it's not
assigned to domain. Otherwise it will be called twice: one time by
virDomainObjListRemove and the second by prlsdkLoadDomain itself.
Maxim Nestratov [Thu, 21 May 2015 13:49:34 +0000 (16:49 +0300)]
parallels: move up updating parameter in prlsdkLoadDomain
It is better to get all necessary parameters and check them on newly
created configuration before actually creating a domain with them or
applying them to an existing domain.
Pavel Hrdina [Fri, 22 May 2015 13:56:57 +0000 (15:56 +0200)]
virsh: reject negative values for scaled integer
Some virsh commands have a size parameter, which is handled as scaled
integer. We don't have any *feature* that would allow to use '-1' as
maximum size, so it's safe to reject any negative values for those
commands.
Michal Privoznik [Sat, 23 May 2015 05:48:37 +0000 (07:48 +0200)]
docs: Document new RO repo mirrors
In the upstream discussion on creating a github mirror [1], it turned
out that there are some read-only mirrors of our repository. Lets
advertise them on our downloads page. But do it wisely and discourage
people in sending a pull requests on GitHub.
John Ferlan [Wed, 13 May 2015 15:10:47 +0000 (11:10 -0400)]
network: Resolve Coverity FORWARD_NULL
To silence Coverity just add a 'p &&' in front of the check in
networkFindUnusedBridgeName after the strchr() call. Even though
we know it's not possible to have strchr return NULL since the only
way into the function is if there is a '%' in def->bridge or it's NULL.
John Ferlan [Wed, 13 May 2015 10:24:50 +0000 (06:24 -0400)]
storage: Resolve Coverity FORWARD_NULL
Coverity points out it's possible for one of the virCommand{Output|Error}*
API's to have not allocated 'output' and/or 'error' in which case the
strstr comparison will cause a NULL deref
bhyve/bhyve_monitor.c: In function 'bhyveMonitorIO':
bhyve/bhyve_monitor.c:51:18: error: missing initializer for field 'tv_sec' of 'const struct timespec' [-Werror=missing-field-initializers]
const struct timespec zerowait = {};
Pavel Fedin [Thu, 21 May 2015 16:02:22 +0000 (19:02 +0300)]
Add missing XDR_FLAGS
Fixes build problems on x86_64-cygwin host for aarch64 target:
CC lxc/libvirt_driver_lxc_impl_la-lxc_monitor_protocol.lo
In file included from lxc/lxc_monitor_protocol.c:7:0:
lxc/lxc_monitor_protocol.h:9:21: fatal error: rpc/rpc.h: No such file or directory
CC rpc/libvirt_setuid_rpc_client_la-virnetmessage.lo
In file included from rpc/virnetmessage.h:24:0,
from rpc/virnetmessage.c:26:
rpc/virnetprotocol.h:9:21: fatal error: rpc/rpc.h: No such file or directory
CC lxc/libvirt_lxc-lxc_monitor_protocol.o
In file included from lxc/lxc_monitor_protocol.c:7:0:
lxc/lxc_monitor_protocol.h:9:21: fatal error: rpc/rpc.h: No such file or directory
Laine Stump [Thu, 21 May 2015 17:36:18 +0000 (13:36 -0400)]
util: better error message after failure to initialize firewall backend
If the firewalld backend wasn't available and libvirt decides to try
setting up a "direct" backend, it checks for the presence of iptables,
ip6tables, and ebtables. If they are not found, a message like this is logged:
error : virFirewallValidateBackend:193 : direct firewall backend
requested, but /usr/sbin/ip6tables is not available:
No such file or directory
But then at a later time if an attempt is made to use the virFirewall
API, failure will be indicated with:
error : virFirewallApply:936 : out of memory
This patch changes virFirewallApply to first check if a firewall
backend hadn't been successfully setup, and logs a slightly more
informative message in that case:
error : virFirewallApply:940 : internal error:
Failed to initialize a valid firewall backend
This resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1223876
Laine Stump [Thu, 21 May 2015 15:26:52 +0000 (11:26 -0400)]
interface: allow multiple IPv4 addresses + dhcp on a single interface
As of netcf-0.2.8, netcf supports configuring multipl IPv4 addresses,
as well as simultaneously configuring dhcp and static IPv4 addresses,
on a single interface. This patch updates libvirt's interface.rng to
allow such configurations.
This resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1223688
Laine Stump [Tue, 5 May 2015 22:27:47 +0000 (18:27 -0400)]
netdev: fail when setting up an SRIOV VF if PF is offline
If an SRIOV PF is offline, the kernel won't complain if you set the
mac address and vlan tag for a VF via this PF, and it will even let
you assign the VF to a guest using PCI device assignment or macvtap
passthrough. But in this case (the PF isn't online), the device won't
be usable in the guest.
Silently setting the PF online would solve the connectivity problem,
but as pointed out by Dan Berrange, when an interface is set online
with no associated config, the kernel will by default turn on IPv6
autoconf, which could create unexpected security problems for the
host. For this reason, this patch instead logs an error and fails the
operation.
This resolves: https://bugzilla.redhat.com/show_bug.cgi?id=893738
Originally filed against RHEL6, but present in every version of
libvirt until today.
Maxim Nestratov [Thu, 21 May 2015 17:23:56 +0000 (20:23 +0300)]
node_device: fix libvirt build if WITH_HAL is defined
commit ffc40b63b50ecb changed uniond _virNodeDevCapData into a typedef
named virNodeDevCapData with a struct that contains the union as well
as a type enum. This change necessitated changing every reference to
"caps->type" into "caps->data.type", but the author of that patch
failed to test a build "WITH_HAL". This patch fixes the one place in
the hal backend that needed changing.
Michal Privoznik [Tue, 12 May 2015 16:21:18 +0000 (18:21 +0200)]
sysinfo: Fix reports on ARM
Due to a kernel commit (b4b8f770e), cpuinfo format has changed on
ARMs. Firstly, 'Processor: ...' may not be reported, it's
replaced by 'model name: ...'. Secondly, the "Processor" string
may occur in CPU name, e.g. 'ARMv7 Processor rev 5 (v7l)'.
Therefore, we must firstly look for 'model name' and then for
'Processor' if not found.
Moreover, lines in the cpuinfo file are shuffled, so we better
not manipulate the pointer to start of internal buffer as we may
lost some info.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Michal Privoznik [Mon, 11 May 2015 15:05:21 +0000 (17:05 +0200)]
qemuDomainDetachChrDevice: Fix chardev hot-unplug
Not every chardev is plugged onto virtio-serial bus. However, the
code introduced in 89e991a2aa36b04 assumes that. Incorrectly.
With previous patches we have three options where a chardev can
be plugged: virtio-serial, USB and PCI. This commit fixes the
detach part. However, since we are not auto allocating USB
addresses yet, I'm just marking the place where appropriate code
should go.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Michal Privoznik [Mon, 11 May 2015 13:20:54 +0000 (15:20 +0200)]
qemuDomainAttachChrDevice: Fix chardev hotplug
Not every chardev is plugged onto virtio-serial bus. However, the
code introduced in 89e991a2aa36b04 assumes that. Incorrectly.
With previous patches we have three options where a chardev can
be plugged: virtio-serial, USB and PCI. This commit fixes the
attach part. However, since we are not auto allocating USB
addresses yet, I'm just marking the place where appropriate code
should go.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Implementation is pretty straight-forward. Of course, not all qemus
out there supports the device, so new capability is introduced and
checked prior each use of the device.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Peter Krempa [Thu, 21 May 2015 14:50:19 +0000 (16:50 +0200)]
util: Avoid shadow of 'ulong' in virMemoryMaxValue
Old compilers whine:
src/util/virutil.c: In function 'virMemoryMaxValue':
src/util/virutil.c:2612: error: declaration of 'ulong' shadows a global declaration [-Wshadow]
/usr/include/sys/types.h:151: error: shadowed declaration is here [-Wshadow]
Jiri Denemark [Tue, 5 May 2015 20:07:12 +0000 (22:07 +0200)]
threadpool: Switch to detached threads
Using joinable threads does not help anything, but it can lead to memory
leaks.
When a worker thread exits, it decreases nWorkers or nPrioWorkers and
once both nWorkers and nPrioWorkers are zero (i.e., the last worker is
gone), quit_cond is signaled. When freeing the pool we first tell all
threads to die and then we are waiting for both nWorkers and
nPrioWorkers to become zero. At this point we already know all threads
are gone. So the only reason for calling virThreadJoin of all workers is
to free the memory allocated for joinable threads. If we avoid
allocating this memory, we don't need to take care of freeing it.
Moreover, any memory associated with a worker thread which died before
we asked it to die (e.g., because virCondWait failed in the thread)
would be lost anyway since virThreadPoolFree calls virThreadJoin only
for threads which were running at the time virThreadPoolFree was called.
Jiri Denemark [Thu, 21 May 2015 06:53:40 +0000 (08:53 +0200)]
Add wrappers for virDomainDiskIndexBy*
Sometimes the only thing we need is the pointer to virDomainDiskDef and
having to call virDomainDiskIndexBy* APIs, storing the disk index, and
looking it up in the disks array is ugly. After this patch, we can just
call virDomainDiskBy* and get the pointer in one step.
Erik Skultety [Tue, 5 May 2015 11:24:41 +0000 (13:24 +0200)]
qemu: Log error if domain uses security driver which is not loaded
When starting a domain, if a domain specifies security drivers we do not have
loaded, we fail. However we don't check for this during
reconnect, so any operation relying on security driver functionality would fail.
If someone e.g. starts a domain with selinux driver loaded, then they change
the security driver to 'none' in config, restart the daemon and call dump/save/..,
QEMU will return an error.
As we shouldn't kill the domain, we should at least log an error to let the
user know that domain reconnect wasn't completely clean.
Luyao Huang [Thu, 21 May 2015 05:08:12 +0000 (13:08 +0800)]
conf: Restore the XML parser context in virDomainMemoryDefParseXML
After parsing the memory device XML the function would not restore the
XML parser context causing invalid XPath starting point for the rest of
the elements. This is a regression since 3e4230d2.
The test case addition uses the <idmap> element that is currently unused
by qemu, but parsed after the memory device definition and formatted
always.
Peter Krempa [Tue, 19 May 2015 14:58:24 +0000 (16:58 +0200)]
conf: Catch memory size overflow earlier
virDomainParseMemory parses the size and then rounds up while converting
it to kibibytes. Since the number is limit-checked before the rounding
it's possible to use a number that would be correctly parsed the first
time, but not the second time. For numbers not limited to 32 bit systems
the magic is 9223372036854775807 bytes. That number then can't be parsed
back in kibibytes.
To solve the issue add a second overflow check for the few values that
would cause the problem. Since virDomainParseMemory is used in config
parsing, this avoids vanishing VMs.
Michal Privoznik [Tue, 19 May 2015 09:55:26 +0000 (11:55 +0200)]
virDomainNumatuneGetMode: Report if numatune was defined
So far, we are not reporting if numatune was even defined. The
value of zero is blindly returned (which maps onto
VIR_DOMAIN_NUMATUNE_MEM_STRICT). Unfortunately, we are making
decisions based on this value. Instead, we should not only return
the correct value, but report to the caller if the value is valid
at all.
For better viewing of this patch use '-w'.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Since af2a1f0587d88656f2c14265a63fbc11ecbd924e,
qemuDomainGetNumaParameters() returns invalid value for a running
guest. The problem is that it is getting the information from cgroups,
but the parent cgroup is being left alone since the mentioned commit.
Since the running guest's XML is in sync with cgroups, there is no need
to look into cgroups (unless someone changes the configuration behind
libvirt's back). Returning the info from the definition fixes a bug and
is also a cleanup.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1221047 Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Jim Fehlig [Thu, 23 Apr 2015 23:00:26 +0000 (17:00 -0600)]
xenconfig: fix spice mousemode and copypaste
From xl.cfg950 man page:
spiceagent_mouse=BOOLEAN
Whether SPICE agent is used for client mouse mode. The default is
true (1) (turn on)
spicevdagent=BOOLEAN
Enables spice vdagent. The Spice vdagent is an optional component for
enhancing user experience and performing guest-oriented management
tasks. Its features includes: client mouse mode (no need to grab
mouse by client, no mouse lag), automatic adjustment of screen
resolution, copy and paste (text and image) between client and domU.
It also requires vdagent service installed on domU o.s. to work.
The default is 0.
spice_clipboard_sharing=BOOLEAN
Enables Spice clipboard sharing (copy/paste). It requires spicevdagent
enabled. The default is false (0).
So if spiceagent_mouse is enabled (client mouse mode) or
spice_clipboard_sharing is enabled, spicevdagent must be enabled.
Along with this change, s/spicedvagent/spicevdagent, set
spiceagent_mouse correctly, and add a test for these spice
features.
Jim Fehlig [Thu, 23 Apr 2015 22:00:45 +0000 (16:00 -0600)]
xenconfig: fix spicepasswd handling
The logic related to spicedisable_ticketing and spicepasswd was
inverted. As per man xl.cfg(5), 'spicedisable_ticketing = 1'
means no passwd is required. On the other hand, a passwd is
required if 'spicedisable_ticketing = 0'. Fix the logic and
produce and error if 'spicedisable_ticketing = 0' but spicepasswd
is not provided. Also fix the spice cfg test file.
Jim Fehlig [Thu, 23 Apr 2015 21:42:50 +0000 (15:42 -0600)]
xenconfig: format spice listenAddr when formating ports
Move formating of spice listenAddr to the section of code
where spice ports are formatted. It is more logical to
format address and ports together. Account for the change
in spice cfg test file by moving 'spicehost'.
Laine Stump [Mon, 11 May 2015 17:50:37 +0000 (13:50 -0400)]
node_device: replace duplicated code in hal and udev backends
Both the hal and udev drivers call virPCI*() functions to the the
SRIOV VF/PF info about PCI devices, and the UDEV backend calls
virPCI*() to get IOMMU group info. Since there is now a single
function call in node_device_linux_sysfs.c to do all of this, replace
all that code in the two backends with calls to
nodeDeviceSysfsGetPCIRelatedDevCaps().
Note that this results in the HAL driver (probably) unnecessarily
calling virPCIDevieAddressGetIOMMUGroupNum(), but in the case that the
host doesn't support IOMMU groups, that function turns into a NOP (it
returns -2, which causes the caller to skip the call to
virPCIDeviceAddressGetIOMMUGroupAddresses()). So in the worst case it
is a few extra cycles spent, and in the best case a mythical platform
that supported IOMMU groups but used HAL rather than UDEV would gain
proper reporting of IOMMU group info.
Laine Stump [Mon, 11 May 2015 17:40:21 +0000 (13:40 -0400)]
node_device: update sriov/iommu info before dumpxml of a device
Because reloading a PF driver with a different number of VFs doesn't
result in any sort of event sent from udev to the libvirt node_device
driver, libvirt's cache of that info can be out of date when a request
arrives for the info about a device. To fix this, we refresh that data
at the time of the dumpxml request, similar to what is already done
for netdev link info and SCSI host capabilities.
Since the same is true for iommu group information (for example, some
other device in the same iommu group could have been detached from the
host), we also create a function to update the iommu group info from
sysfs, and a common function that does both. (a later patch will call
this common function from the udev and hal backends).
Laine Stump [Mon, 11 May 2015 17:35:50 +0000 (13:35 -0400)]
node_device: new functions to get sriov/iommu info from sysfs
The udev and hal drivers both already call the same functions as these
new functions added to node_device_linux_sysfs.c, but 1) we need to
call them from node_device_driver.c, and 2) it would be nice to
eliminate the duplicated code from the hal and udev backends.
Laine Stump [Wed, 6 May 2015 20:40:39 +0000 (16:40 -0400)]
node device: prepare node_device_linux_sysfs.c to add more functions
This file contains only a single function, detect_scsi_host_caps(),
which is declared in node_device_driver.h and called from both the hal
and udev backends. Other things common to the hal and udev drivers
can be placed in that file though. As a prelude to adding further
functions, this patch renames the existing function to something
closer in line with other internal libvirt function names
(nodeDeviceSysfsGetSCSIHostCaps()), and puts the declarations into a
separate .h file.
Laine Stump [Fri, 8 May 2015 16:55:00 +0000 (12:55 -0400)]
conf: make virNodeDevCapData an official type
For some reason a union (_virNodeDevCapData) that had only been
declared inside the toplevel struct virNodeDevCapsDef was being used
as an argument to functions all over the place. Since it was only a
union, the "type" attribute wasn't necessarily sent with it. While
this works, it just seems wrong.
This patch creates a toplevel typedef for virNodeDevCapData and
virNodeDevCapDataPtr, making it a struct that has the type attribute
as a member, along with an anonymous union of everything that used to
be in union _virNodeDevCapData. This way we only have to change the
following:
Andrea Bolognani [Fri, 15 May 2015 16:14:40 +0000 (18:14 +0200)]
virsh: Fix dommemstat --period option type.
The option didn't have VSH_OT_INT type even thought it's expected
to be numeric, as shown by the fact that vshCommandOptInt() is later
used to retrieve its value.
Andrea Bolognani [Fri, 15 May 2015 16:14:39 +0000 (18:14 +0200)]
virsh: Improve error message on integer value parsing failure.
Replace more than 30 ad-hoc error messages with a single, generic one
that contains the name of the option being processed and some hints
to help the user understand what could have gone wrong.
Tony Krowiak [Mon, 27 Apr 2015 21:57:30 +0000 (17:57 -0400)]
libvirt: tests: test protected key mgmt ops support
Test the support for enabling/disabling CPACF protected key management
operations for a guest.
Signed-off-by: Tony Krowiak <akrowiak@linux.vnet.ibm.com> Signed-off-by: Viktor Mihajlovski <mihajlov@linux.vnet.ibm.com> Reviewed-by: Boris Fiuczynski <fiuczy@linux.vnet.ibm.com> Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Tony Krowiak <akrowiak@linux.vnet.ibm.com> Signed-off-by: Daniel Hansel <daniel.hansel@linux.vnet.ibm.com> Signed-off-by: Boris Fiuczynski <fiuczy@linux.vnet.ibm.com> Reviewed-by: Boris Fiuczynski <fiuczy@linux.vnet.ibm.com> Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Tony Krowiak <akrowiak@linux.vnet.ibm.com> Signed-off-by: Viktor Mihajlovski <mihajlov@linux.vnet.ibm.com> Signed-off-by: Daniel Hansel <daniel.hansel@linux.vnet.ibm.com> Reviewed-by: Boris Fiuczynski <fiuczy@linux.vnet.ibm.com> Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Jim Fehlig [Fri, 15 May 2015 18:26:30 +0000 (12:26 -0600)]
libxl: provide impl for nodeGetSecurityModel
Currently, the libxl driver does not support any security drivers.
When the qemu driver has no security driver configued,
nodeGetSecurityModel succeeds but returns an empty virSecurityModel
object. Do the same in the libxl driver instead of reporting
this function is not supported by the connection driver:
virNodeGetSecurityModel
Laine Stump [Thu, 30 Apr 2015 20:59:04 +0000 (16:59 -0400)]
qemu: log error when domain has an unsupported IDE controller
We have previously effectively ignored all <controller type='ide'>
elements in a domain definition.
On the i440fx-based machinetypes there is an IDE controller that is
included in the chipset and can't be removed (which is the ide
controller with index='0'>), so it makes sense to ignore that one
controller. However, if an i440fx domain definition has a 2nd
controller, nothing catches this error (unless you also have a disk
attached to it, in which case qemu will complain that you're trying to
use the ide controller named "ide1", which doesn't exist), and if any
other type of domain has even a single controller defined, it will be
incorrectly ignored.
Ignoring a bogus controller definition isn't such a big problem, as
long as an error is logged when any disk is attached to that
non-existent controller. But in the case of q35-based machinetypes,
the hardcoded id ("alias" in libvirt terms) of its builtin SATA
controller is "ide", which happens to be the same id as the builtin
IDE controller on i440fx machinetypes. So libvirt creates a
commandline believing that it is connecting the disk to the builtin
(but actually nonexistent) IDE controller, qemu thinks that libvirt
wanted that disk connected to the builtin SATA controller, and
everybody is happy.
Until you try to connect a 2nd disk to the IDE controller. Then qemu
will complain that you're trying to set unit=1 on a controller that
requires unit=0 (SATA controllers are organized differently than IDE
controllers).
After this patch, if a domain has an IDE controller defined for a
machinetype that has no IDE controllers, libvirt will log an error
about the controller itself as it is building the qemu commandline
(rather than a (possible) error from qemu about disks attached to that
controller). This is done by adding IDE to the list of controller
types that are handled in the loop that creates controller command
strings in qemuBuildCommandline() (previously it would *always* skip
IDE controllers). Then qemuBuildControllerDevStr() is modified to log
an appropriate error in the case of IDE controllers.
In the future, if we add support for extra IDE controllers (piix3-ide
and/or piix4-ide) we can just add it into the IDE case in
qemuBuildControllerDevStr(). For now, nobody seems anxious to add
extra support for an aging and very slow controller, when there are so
many better options available.
projects / libvirt.git / log