The function takes two important arguments (among many others): @node
and @page_size. From these two a path under /sys is constructed. The
path is then used to read and write the desired size of huge pages
pool. However, if the path does not exists due to either @node or
@page_size having nonexistent value (e.g. there's no such NUMA node or
no page size like -2), an cryptic error message is produced:
virsh # allocpages --pagesize 2049 --pagecount 8 --cellno -2
error: Failed to open file '/sys/devices/system/node/node-2/hugepages/hugepages-2049kB/nr_hugepages': No such file or directory
Add two more checks to catch this and therefore produce much more
friendlier error messages.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
dnsmasq conf file contents needs to have quotes escaped for it to
work. Because of this, the network-create/start for a network with
quotes in the name fails. The patch escapes strings for the entries
that go into the conf file.
Signed-off-by: Shivaprasad G Bhat <sbhat@linux.vnet.ibm.com>
Maxim Nestratov [Wed, 3 Jun 2015 21:10:00 +0000 (00:10 +0300)]
parallels: process '/' mount point correctly for containers
Since we are going to add block devices as root disks we have
to specify root mount point for boot block devices. But we
shouldn't do this if a filesystem disk with such
target mount point already exists.
Statistics provided through PCS SDK. As we have only async interface in SDK we
need to be subscribed to statistics in order to get it. Trivial solution on
every stat request to subscribe, wait event and then unsubscribe will lead to
significant delays in case of a number of successive requests, as the event
will be delivered on next PCS server notify cycle. On the other hand we don't
want to keep unnesessary subscribtion. So we take an hibrid solution to
subcsribe on first request and then keep a subscription while requests are
active. We populate cache of statistics on subscribtion events and use this
cache to serve libvirts requests.
* Cache details.
Cache is just handle to last arrived event, we call this cache
as if this handle is valid it is used to serve synchronous
statistics requests. We use number of successive events count
to detect that user lost interest to statistics. We reset this
count to 0 on every request. If more than PARALLELS_STATISTICS_DROP_COUNT
successive events arrive we unsubscribe. Special value of -1
of this counter is used to differentiate between subscribed/unsubscribed state
to protect from delayed events.
Values of PARALLELS_STATISTICS_DROP_COUNT and PARALLELS_STATISTICS_TIMEOUT are
just drop-ins, choosen without special consideration.
* Thread safety issues
Use parallelsDomObjFromDomainRef in parallelsDomainBlockStats as
we could wait on domain lock down on stack in prlsdkGetStatsParam
and if we won't keep reference we could get dangling pointer
on return from wait.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@parallels.com>
parallels: return only success from PCS event handler
2 reasons to to this.
1. PCS SDK really don't care of handler return value.
2. It hard to imagine how notifier can handle
subscriber failures. Even if there are some situations
we probably will use some special error codes and
not just throw error codes we get from SDK itself.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@parallels.com>
Use issuer type instead of event type to group
vm related events. This saves us from
explicit enumeration of all vm event types in
prlsdkHandleVmEvent.
Therefore, after 65c61e50 the test if failing for me. Problem is,
we are trying to assign couple of negative values into char
assuming some will overflow and some don't. That can't be the
case if 'char' is unsigned by default. Lets use more explicit types
instead: int8_t and uint8_t where is no ambiguity.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Commit 825df8c3 refactored virProcess{Set,Get}Affinity routines,
however broke BSD implementation because of the incorrect variable
name. Fix build by using a proper variable name.
Peter Krempa [Mon, 8 Jun 2015 07:32:24 +0000 (09:32 +0200)]
util: Properly return error from virGetUserID and virGetGroupID stubs
The stubs for the two functions that are compiled on platforms that
don't have HAVE_GETPWUID_R and friends defined do not return error but
report an error message. The calling code then assumes that the @uid or
@gid arguments were filled, which is not the case in the stubs.
When building without avahi support, we used VIR_DEBUG() to note that to
the user. However, functions that fail because of that (return NULL/-1)
did not set the error message. This was the only file that forgot to do
such thing.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
The virStorageBackendFileSystemIsMounted is called from three source paths
checkPool, startPool, and stopPool. Both start and stop validate the FS
fields before calling *IsMounted; however the check path there is no call.
This could lead the code into returning a true in "isActive" if for some
reason the target path for the pool was mounted. The assumption being
that if it was mounted, then we believe we started/mounted it.
It's also of note that commit id '81165294' added an error message for
the start/mount path regarding that the target is already mounted so
fail the start. That check was adjusted by commit id '13fde7ce' to
only message if actually mounted.
At one time this led to the libvirtd restart autostart code to declare
that the pool was active even though the startPool would inhibit startup
and the stopPool would inhibit shutdown. The autostart path changed as
of commit id '2a31c5f0' as part of the keep storage pools started between
libvirtd restarts.
This patch adds the same check made prior to start/mount and stop/unmount
to ensure we have a valid configuration before attempting to see if the
target is already mounted to declare "isActive" or not. Finding an improper
configuration will now cause an error at checkPool, which should make it
so we can no longer be left in a situation where the pool was started and
we have no way to stop it.
Currently the assumption on the error message is that there are
no source device paths defined when the number of devices check
fails, but in reality the XML could have had none or it could have
had more than the value supported. Adjust the error message accordingly
to make it clearer what the error really is.
Eric Blake [Fri, 5 Jun 2015 03:44:48 +0000 (21:44 -0600)]
remote: fix odd comma operator
Commit 1882c0bd accidentally used ',' instead of ';'; oddly
enough, the result was still syntactically valid (yes, C is
a fun language). But it made me do a double take; it's better
to use idiomatic syntax.
Peter Krempa [Thu, 4 Jun 2015 11:47:02 +0000 (13:47 +0200)]
qemu: Update balloon info only if job is allowed
In qemuDomainUpdateCurrentMemorySize I misplaced the actual update of
the balloon size to a place where it may not be initialized. Move it a
few lines above.
Peter Krempa [Thu, 4 Jun 2015 11:20:12 +0000 (13:20 +0200)]
conf: Fix mistakes in pointer usage in virDomainObjGetDefs
Coverity rightfully determined that in commit 3d021381c71221e563182f03
I made a mistake in the first check if @persDef is not NULL is
dereferencing it rather than checking.
Additionally if the vm is online the code would set @liveDef twice
rather than modifying @persDef. Fix both mistakes.
qemu: Check for qemu capability when calling virDomainGetBlockIoTune()
When getting block device I/O tuning data there is no check for whether
QEMU supports such options and the call fails on
qemuMonitorGetBlockIoThrottle() when getting the particular throttle
data. So try reporting a better error when blkdeviotune is not
supported.
Peter Krempa [Fri, 29 May 2015 12:37:20 +0000 (14:37 +0200)]
conf: Add new helpers to resolve virDomainModificationImpact to domain defs
virDomainLiveConfigHelperMethod that is used for this job now does
modify the flags but still requires the callers to extract the correct
definition objects.
In addition coverity and other static analyzers are usually unhappy as
they don't grasp the fact that @flags are upadted according to the
correct def to be present.
To work this issue around and simplify the calling chain let's add a new
helper that will work only on drivers that always copy the persistent
def to a transient at start of a vm. This will allow to drop a few
arguments. The new function syntax will also fill two definition
pointers rather than modifying the @flags parameter.
Peter Krempa [Thu, 28 May 2015 13:31:09 +0000 (15:31 +0200)]
libxl: Don't remove vcpu pin definition in libxlDomainCleanup
The vCPU pinning definition gets removed when the domain definition is
being freed later. If there is no next configuration it would remove the
configured pinning.
Peter Krempa [Mon, 25 May 2015 12:19:50 +0000 (14:19 +0200)]
qemu: libxl: vcpupin: Don't reset pinning when pinning to all pcpus
In the pre-NUMA ages pinning a vCPU to all pCPUs was eaqual to deleting
the pinning info. Now it does not entirely work that way. Pinning a vCPU
to all pCPUs might be a desired operation. Additionally removal of the
pinning will result into using the default pinning information at the
next boot which might be different from all vcpus.
This patch removes the false assumption that we should remove the
pinning after pinning to all vCPUs and tweaks the documentation for
virsh.
A later patch will implement a new flag for the virDomainPinVcpuFlags
API that will allow to remove the pinning in a sane way.
Peter Krempa [Mon, 25 May 2015 13:43:29 +0000 (15:43 +0200)]
conf: Store cpu count as unsigned int
While we probably won't see machines with more than 65536 cpus for a
while lets store the cpu count as an integer so that we can avoid quite
a lot of overflow checks in our code.
Peter Krempa [Wed, 27 May 2015 15:33:23 +0000 (17:33 +0200)]
qemu: Refactor qemuDomainGetInfo
Since the returned structure uses "unsigned long" for memory sizes add a
few overflow checks to notify the user in case we are not able to
represent given values.
Peter Krempa [Wed, 27 May 2015 13:51:52 +0000 (15:51 +0200)]
qemu: Add helper to update domain balloon size and refactor usage places
When qemu does not support the balloon event the current memory size
needs to be queried. Since there are two places that implement the same
logic, split it out into a function and reuse.
Peter Krempa [Wed, 27 May 2015 13:04:14 +0000 (15:04 +0200)]
qemu: process: Update current balloon state to maximum on vm startup
After libvirt issues the balloon resize command, the current balloon
size needs to be changed to the maximum memory size since the vCPUs were
not started and thus the balloon driver could not return the memory.
Since GetXMLDesc and other APIs return the balloon size without updating
it in case they are not able to obtain the job and the memory balloon
does not support the asynchronous event the sizing might be incorrect.
Ján Tomko [Wed, 13 Aug 2014 11:06:04 +0000 (13:06 +0200)]
Always add 'console' matching the 'serial' device
We have been formatting the first serial device also
as a console device, but only if there were no other consoles.
If there is a <serial> device present in the XML, but no serial
<console>, or if there isn't any <console> at all but the domain
definition hasn't gone through a parse->format->parse round-trip,
the <console> device would not be formatted.
Change the code to always add the stub device for the first
serial device.
check if console/channel PTY is null before attempting to open
Console/channel devices have their pty devices assigned when the emulator is
actually started. If time is spent in guest preparation, someone attempts
to open the console/channel, the libvirt crashes in virChrdevLockFilePath().
The patch attempts to fix the crash by adding a check before attempting to
open.
Signed-off-by: Shivaprasad G Bhat <sbhat@linux.vnet.ibm.com>
Lubomir Rintel [Mon, 1 Jun 2015 18:40:23 +0000 (20:40 +0200)]
virnetdev: fix moving of 802.11 phys
There was a couple of problems with the style fixes applied to the original
patch:
1.) virFileReadAllQuiet comparison was incorrectly parenthesized when moved
into a condition, causing the len to be set to the result of comparison. This,
together with the removed underflow check would underflow the phy buffer.
2.) The logic was broken. Failure to call "ip" would abort the function, thus
the "iw" branch would never be reached.
This aims to fix the issues and work around possible style complains :)
util: Clear output broadcast address before filling it in
Since commit 55ace7c4789c8a7408139460f4b639cee00e5125, the sockettest
fails without VIR_TEST_DEBUG set. The problem is found by test number
42 (co-incidence?), which tests range '192.168.122.1' -
'192.168.122.255' in network '192.168.122.0/24'. That is supposed to
fail because the end address is equal to the broadcast address.
When comparing these two in 'virSocketAddrEqual(end, &broadcast)',
there is a check for sin_addr as well as for sin_port. That port,
however, is different when we do not enable test debugging. With the
testing enabled, the port is 0 (correctly initialized), but without that
it has a random number there. And that's because the structure is not
initialized anywhere.
By zeroing the structure before filling in the info, we make sure we
return only the address and not any information that was not requested.
And the test work once again.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Jiri Denemark [Mon, 1 Jun 2015 13:06:16 +0000 (15:06 +0200)]
virsh: Fix Ctrl-C behavior when watching a job
When watching a job (save, managedsave, dump, migrate) virsh spawns a
thread to call the appropriate API and waits for the result while
watching for interruption signals (SIGINT, Ctrl-C on the terminal).
Whenever such signal is caught, virsh calls virDomainAbortJob, stops
waiting for the job, and returns the result of virDomainAbortJob.
This is wrong because the job might have finished in the meantime or it
might have been cancelled by someone else and virsh would just report
the failure to abort the job. However, we are not interested in the
virDomainAbortJob's result at all, we need to keep waiting for the main
job to finish and report its result instead.
Peter Krempa [Thu, 21 May 2015 16:25:36 +0000 (18:25 +0200)]
util: bitmap: Add virBitmapToDataBuf that does not allocate the buffer
Since some functions can be optimized by reusing the buffers that they
already have instead of allocating and copying new ones, lets split
virBitmapToData to two functions where one only converts the data and
the second one is a wrapper that allocates the buffer if necessary.
Peter Krempa [Thu, 21 May 2015 12:55:18 +0000 (14:55 +0200)]
conf: Refactor emulatorpin handling
Store the emulator pinning cpu mask as a pure virBitmap rather than the
virDomainPinDef since it stores only the bitmap and refactor
qemuDomainPinEmulator to do the same operations in a much saner way.
As a side effect virDomainEmulatorPinAdd and virDomainEmulatorPinDel can
be removed since they don't add any value.
Peter Krempa [Thu, 28 May 2015 14:22:13 +0000 (16:22 +0200)]
qemu: Fix possible crash in qemuProcessSetVcpuAffinities
In case when <vcpu ... cpuset=""> is not specified, the vcpupin array is
not guaranteed to be allocated to def->vcpus. This would cause a crash
for TCG since it does not report thread IDs for vCPUs.
Maxim Nestratov [Tue, 26 May 2015 17:12:00 +0000 (20:12 +0300)]
parallels: set virtType depending on driver name
We remember driver name in a new field 'drivername' within
private parallels connection structure. When a new domain
is defined we use this name to set corresponding virtType.
We set VIR_DOMAIN_VIRT_VZ for 'vz' driver and
VIR_DOMAIN_VIRT_PARALLELS for 'Parallels'.
Maxim Nestratov [Tue, 26 May 2015 17:12:00 +0000 (20:12 +0300)]
parallels: add a new vz connection driver and hypervisor structures
We add this connection driver just as an exact copy with different
name to keep backward compatibility.
Vz stands for Virtuozzo, which is a new name of Parallels Cloud Server.
Maxim Nestratov [Tue, 26 May 2015 17:12:00 +0000 (20:12 +0300)]
parallels: accept vz as a driver uri and name
If 'parallels:///system' uri is specified then connection is made to
'Parallels' driver and domain type will be VIR_DOMAIN_VIRT_PARALLELS.
In case of 'vz:///system' connection is established to 'vz' driver
and domain type will be VIR_DOMAIN_VIRT_VZ.
Luyao Huang [Sun, 31 May 2015 11:29:46 +0000 (19:29 +0800)]
qemu: Do not release device address on successful RNG attach
Commit id '980b265d' neglected to check for a successful status when
deciding whether to release the device address for the RNG attach thus
the address would be released even though the device was added.
Luyao Huang [Sun, 31 May 2015 13:27:25 +0000 (21:27 +0800)]
qemu: Need to return status of RNG device removal
Commit id '862473fa' neglected to return the status from the
qemuDomainRemoveRNGDevice call in qemuDomainRemoveDevice causing
the function to always fail when receiving an RNG device unplug
event. Additionally the domain status/state would not be updated
in the processDeviceDeletedEvent path.
Eric Blake [Tue, 2 Jun 2015 15:17:43 +0000 (09:17 -0600)]
build: silence ar warnings on rawhide
Newer binutils 'ar' has added an option 'D' for deterministic
builds, and at least on rawhide, this option is enabled by default.
But it conflicts with the 'u' optimization where the linker only
modifies libraries based on file timestamps, but can result in
different library ordering based on which files were touched last.
Thus, it results in some noisy compilation, for every CCLD line:
CCLD libvirt_driver_qemu_impl.la
ar: `u' modifier ignored since `D' is the default (see `U')
Upstream automake has decided that defaulting ARFLAGS to 'cru' is
no longer beneficial, and that switching the default to 'cr' will
both silence the noise and not penalize modern build systems.
But rather than wait for newer automake to propagate to all systems
that already have newer binutils, we might as well just use the new
default ourselves, even on older platforms.
* configure.ac: Default AR[_]FLAGS to 'cr', not 'cru'.
Laine Stump [Tue, 26 May 2015 19:44:24 +0000 (15:44 -0400)]
util: report all address range errors in virSocketAddrGetRange()
There are now many more reasons that virSocketAddrGetRange() could
fail, so it is much more informative to report the error there instead
of in the caller. (one of the two callers was previously assuming
success, which is almost surely safe based on the parsing that has
already happened to the config by that time, but it still is nicer to
account for an error "just in case")
Part of fix for: https://bugzilla.redhat.com/show_bug.cgi?id=985653
Laine Stump [Tue, 26 May 2015 19:32:59 +0000 (15:32 -0400)]
network: cleanup range loop in networkDnsmasqConfContents
This loop had automatic variable definitions mixed with code. This
patch moves the definitions to the top of the function and puts
cleanup for them at the bottom. No functional change.
Part of fix for: https://bugzilla.redhat.com/show_bug.cgi?id=985653
Laine Stump [Fri, 22 May 2015 21:32:02 +0000 (17:32 -0400)]
network: validate DHCP ranges are completely within defined network
virSocketAddrGetRange() has been updated to take the network address
and prefix, and now checks that both the start and end of the range
are within that network, thus validating that the entire range of
addresses is in the network. For IPv4, it also checks that ranges to
not start with the "network address" of the subnet, nor end with the
broadcast address of the subnet (this check doesn't apply to IPv6,
since IPv6 doesn't have a broadcast or network address)
Negative tests have been added to the network update and socket tests
to verify that bad ranges properly generate an error.
This resolves: https://bugzilla.redhat.com/show_bug.cgi?id=985653
projects / people / liuw / libxenctrl-split / libvirt.git / log