One of my previous patches (f2a4e5f176c408) tried to fix crashing
libvirtd on domain detroy. However, we need to copy pattern from
qemuProcessHandleMonitorEOF() instead of decrementing reference
counter. The rationale for this is, if qemu process is dying due
to domain being destroyed, we obtain EOF on both the monitor and
agent sockets. However, if the exit is expected, qemuProcessStop
is called, which cleans both agent and monitor sockets up. We
want qemuAgentClose() to be called iff the EOF is not expected,
so we don't leak an FD and memory. Moreover, there could be race
with qemuProcessHandleMonitorEOF() which could have already
closed the agent socket, in which case we don't want to do
anything.
John Ferlan [Tue, 22 Jan 2013 22:09:26 +0000 (17:09 -0500)]
lxc_process: Avoid passing NULL iface->iname
A followon to commit id: 68dceb635 - if iface->iname is NULL, then
neither virNetDevOpenvswitchRemovePort() nor virNetDevVethDelete()
should be called. Found by Coverity.
Fix nwfilter driver reload/shutdown handling when unprivileged
Although the nwfilter driver skips startup when running in a
session libvirtd, it did not skip reload or shutdown. This
caused errors to be reported when sending SIGHUP to libvirtd,
and caused an abort() in libdbus on shutdown due to trying
to remove a dbus filter that was never added
Eric Blake [Tue, 22 Jan 2013 17:13:01 +0000 (10:13 -0700)]
build: allow virObject to have no parent
When building with static analysis enabled, we turn on attribute
nonnull checking. However, this caused the build to fail with:
../../src/util/virobject.c: In function 'virObjectOnceInit':
../../src/util/virobject.c:55:40: error: null argument where non-null required (argument 1) [-Werror=nonnull]
Creation of the virObject class is the one instance where the
parent class is allowed to be NULL. Making things conditional
will let us keep static analysis checking for all other .c file
callers, without breaking the build on this one exception.
* src/util/virobject.c: Define witness.
* src/util/virobject.h (virClassNew): Use it to force most callers
to pass non-null parameter.
John Ferlan [Tue, 22 Jan 2013 14:15:37 +0000 (09:15 -0500)]
viralloc: Adjust definition of VIR_FREE() for Coverity
The Coverity static analyzer was generating many false positives for the
unary operation inside the VIR_FREE() definition as it was trying to evaluate
the else portion of the "?:" even though the if portion was (1).
Peter Krempa [Tue, 22 Jan 2013 14:27:04 +0000 (15:27 +0100)]
conf: Check if number of vCPUs fits in the storage variable
The count of vCPUs for a domain is extracted as a usingned long variable
but is stored in a unsigned short. If the actual number was too large,
a faulty number was stored.
Peter Krempa [Tue, 22 Jan 2013 11:17:18 +0000 (12:17 +0100)]
virsh-domain: Refactor error paths for cmdCPUStats
This patch fixes the following issues in the cpu-stats virsh command:
1) Renames label failed_params to no_memory to match coding style
2) Uses proper typed parameter cleanup in error paths to avoid leaks
3) Adds a ret variable and simplifies error labels
4) Changes error message to a slightly more descriptive one and gets rid
of the newline at the end:
Before:
$ virsh cpu-stats tr
error: Failed to virDomainGetCPUStats()
error: Requested operation is not valid: domain is not running
After:
$ tools/virsh cpu-stats tr
error: Failed to retrieve CPU statistics for domain 'tr'
error: Requested operation is not valid: domain is not running
John Ferlan [Tue, 22 Jan 2013 14:15:50 +0000 (09:15 -0500)]
tests: Need to initialize 'test' properly on error path
In the error path, the test buffer is free'd, but due to how the free
routine is written the 'test' buffer pointer does not return to the caller
as NULL and then the free'd buffer address is returned to the caller.
John Ferlan [Tue, 22 Jan 2013 14:41:04 +0000 (09:41 -0500)]
storage: Add coverity[dead_error_condition] to avoid error
The local redefinition of PED_PARTITION_PROTECTED results in the error
but is not a problem especially if the built code doesn't have the latest
definitions.
John Ferlan [Tue, 22 Jan 2013 14:41:02 +0000 (09:41 -0500)]
remote: Avoid coverity[leaked_storage] message.
Upon successful return of virNetClientStreamEventAddCallback() the
allocated cbdata field will be freed by virNetClientStreamEventRemoveCallback()
as cbOpaque using the free function remoteStreamCallbackFree().
John Ferlan [Tue, 22 Jan 2013 14:41:01 +0000 (09:41 -0500)]
qemu: Add coverity[negative_returns] tag
This avoids "Event negative_returns: A negative constant "-1" is passed as
an argument to a parameter that cannot be negative.". The called function
uses -1 to determine whether it needs to traverse all the hostdevs.
John Ferlan [Tue, 22 Jan 2013 14:40:58 +0000 (09:40 -0500)]
nodeinfo: Add coverity[dead_error_begin] and [returned_null] tags
The use of switch statements inside a bounded for loop resulted in some
false positives regarding the "default:" label which cannot be reached
since each of the other case statements use the possible for loop values.
A [dead_error_begin] was added before the default label.
Commit id ebdbe25a adjusted the algorithm and the caller guarantees that
the 'params' will have a '_' in the name being searched. Add the [returned_null]
tag to the two instances.
John Ferlan [Tue, 22 Jan 2013 14:40:57 +0000 (09:40 -0500)]
lxc: Add coverity[dead_error_begin] tag in switch stmts
The use of switch statements inside a bounded for loop resulted in some
false positives regarding the "default:" label which cannot be reached
since each of the other case statements use the possible for loop values.
John Ferlan [Fri, 18 Jan 2013 14:34:13 +0000 (09:34 -0500)]
selinux: Resolve resource leak using the default disk label
Commit id a994ef2d1 changed the mechanism to store/update the default
security label from using disk->seclabels[0] to allocating one on the
fly. That change allocated the label, but never saved it. This patch
will save the label. The new virDomainDiskDefAddSecurityLabelDef() is
a copy of the virDomainDefAddSecurityLabelDef().
John Ferlan [Fri, 18 Jan 2013 14:29:05 +0000 (09:29 -0500)]
rpc: Remove unnecessary calls
The code is not reachable as of commit id: bb85f229. Removed
virKeepAliveStop() and virObjectUnref() because 'ka' cannot be
anything but NULL at the cleanup label.
Michal Privoznik [Thu, 17 Jan 2013 07:35:53 +0000 (08:35 +0100)]
safe{read,write}: Document usage with nonblocking FD
Currently, whenever somebody calls saferead() on nonblocking FD
(safewrite() is totally interchangeable for purpose of this message)
he might get wrong return value. For instance, in the first iteration
some data is read. The number of bytes read is stored into local
variable 'nread'. However, in next iterations we can get -1 from
read() with errno == EAGAIN, in which case the -1 is returned despite
fact some data has already been read. So the caller gets confused.
Bare read() should be used for nonblocking FD.
Eric Blake [Wed, 9 Jan 2013 20:50:03 +0000 (13:50 -0700)]
spec: indent %if to make it easier to see conditions
Nested conditionals are hard to read if they are not indented.
We can't add arbitrary whitespace to everything in spec files,
but we CAN add spaces before %if and %define. Use this trick,
plus a fancy sed script that rewrites a spec file into a C
file, so we can use cppi to keep our spec file nice.
|// # RHEL-5 builds are client-only for s390, ppc
|#if a // 0%{?rhel} == 5
|# if a // %{ix86} x86_64 ia64
|# define client_only 1
|# endif
|#endif
and errors from 'make syntax-check' look like:
spec_indentation
cppi: mingw-libvirt.spec.in: line 130: not properly indented
maint.mk: incorrect preprocessor indentation
* libvirt.spec.in: Add some indentation to make it easier to follow
various conditionals.
* mingw-libvirt-spec.in: Likewise.
* cfg.mk (sc_spec_indentation): New syntax check to enforce it.
Peter Krempa [Thu, 17 Jan 2013 13:07:10 +0000 (14:07 +0100)]
qemu: Reject attempts to create snapshots with names containig '/'
The snapshot name is used to create path to the definition save file.
When the name contains slashes the creation of the file fails. Reject
such names.
Peter Krempa [Thu, 17 Jan 2013 13:04:01 +0000 (14:04 +0100)]
qemu: Don't return success if creation of snapshot save file fails
When the snapshot definition can't be saved, the
qemuDomainSnapshotCreate function succeeded without filling some of the
fields in the internal definition.
This patch removes the snapshot and returns failure if the XML file
cannot be written.
When running virDomainDestroy, we need to make sure that no other
background thread cleans up the domain while we're doing our work.
This can happen if we release the domain object while in the
middle of work, because the monitor might detect EOF in this window.
For this reason we have a 'beingDestroyed' flag to stop the monitor
from doing its normal cleanup. Unfortunately this flag was only
being used to protect qemuDomainBeginJob, and not qemuProcessKill
This left open a race condition where either libvirtd could crash,
or alternatively report bogus error messages about the domain already
having been destroyed to the caller
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Claudio Bley [Thu, 17 Jan 2013 15:00:10 +0000 (16:00 +0100)]
docs: Assign classes to documentation elements
In CSS the following class names are available:
* keyword (keywords like "typedef", "struct")
* type (types like "int", "void*")
* comment (comments after members of enums or structs)
* directive (preprocessor directives, #define)
* undisclosed (text saying that the API is not public)
Additionally, kill all of the left-over "programlisting" class
assignments. There are no CSS rules for them.
Jiri Denemark [Tue, 15 Jan 2013 13:51:45 +0000 (14:51 +0100)]
Add virTypedParams* APIs
Working with virTypedParameters in clients written in C is ugly and
requires all clients to duplicate the same code. This set of APIs makes
this code for manipulating with virTypedParameters integral part of
libvirt so that all clients may benefit from it.
Eric Blake [Thu, 17 Jan 2013 23:29:38 +0000 (16:29 -0700)]
build: fix build on BSD
A build on FreeBSD failed with:
util/virportallocator.c:108: error: storage size of 'addr' isn't known
util/virportallocator.c:123: error: 'INADDR_ANY' undeclared (first use in this function)
It turns out that while POSIX allows sockaddr_in to leak in through
<arpa/inet.h> (the way Linux does it), it is not mandatory, and
conforming applications are required to get it through <netinet/in.h>.
* src/util/virportallocator.c: Include header for struct
sockaddr_in.
* tests/virportallocatortest.c: Likewise.
John Ferlan [Thu, 17 Jan 2013 19:17:15 +0000 (14:17 -0500)]
network: Remove dead code getting, but not using ipdef
The fetch of 'ipdef' in networkRefreshDhcpDaemon() when the loop to fill
in ipv4def fails to find an ipv4 address with dhcp defined. The filled in
ipdef value was not used. Code was made unnecessary with commit it 2d5cd1.
qemu: Double mutex unlock in qemuDomainModifyDeviceFlags
The driver mutex was unlocked in qemuDomainModifyDeviceFlags before
entering qemuDomainObjBeginJobWithDriver where it will be unlocked once
more leaving it in an undefined state. The result was that two
threads were simultaneously looking up the domain hash table during
multiple parallel device attach/detach operations.
Luckily this triggered a virHashIterationError.
Signed-off-by: Viktor Mihajlovski <mihajlov@linux.vnet.ibm.com>
Reinier Schoof [Thu, 17 Jan 2013 11:24:04 +0000 (12:24 +0100)]
fixed xt_physdev warning when defining ip(6)tables rules
When starting a VM, /var/log/messages was spammed with the following message:
xt_physdev: using --physdev-out in the OUTPUT, FORWARD and POSTROUTING chains for non-bridged traffic is not supported anymore.
With each extra VM I start, the messages get amplified
exponentially. This results in longer starting times every new VM,
relative the the previously started VM. When I ran a test with
starting 100 equal VM's, the first VM started in about 2 seconds, the
100th VM took 48 seconds to start. I'm running a vanilla 3.7.1 kernel,
but I have the same issue on VM hosts with kernel 3.2.28 or 3.2.0,
running libvirt 0.9.12 and 0.9.8 respectively.
Looking into the warning, it seemed that iptables need an extra argument,
--physdev-is-bridged, in commands like:
iptables -A libvirt-out -m physdev --physdev-is-bridged --physdev-out vnet99 -g FP-vnet99
With that, the warnings in /var/log/messages are gone and running the
test again proved the 100th VM started in 3.8 seconds.
The symptom was that attempts to modify a network device using
virDomainUpdateDeviceFlags() would fail if the original device had a
<boot> element (e.g. "<boot order='1'/>"), even if the updated device
had the same <boot> element. Instead, the following error would be logged:
cannot modify network device boot index setting
It's true that it's not possible to change boot order (internally
known as bootIndex) of a live device; qemuDomainChangeNet checks for
that, but the problem was that the information it was checking was
incorrect.
Explanation:
When a complete domain is parsed, a global (to the domain) "bootMap"
is passed down to the parse for each device; the bootMap is used to
make sure that devices don't have conflicting settings for their boot
orders.
When a single device is parsed by itself (as in the case of
virDomainUpdateDeviceFlags), there is no global bootMap that would be
appropriate to send, so NULL is sent instead. However, although the
lowest level function that parses just the boot order *does* simply
skip the sanity check in that case, the next higher level
"virDomainDeviceInfoParseXML" function refuses to call down to the
lower "virDomainDeviceBootParseXML" if bootMap is NULL. So, the boot
order is never set in the "new" device object, and when it is compared
to the original (which does have a boot order), they don't match.
The fix is to patch virDomainDeviceInfoParseXML to not care about
bootMap, and just always call virDomainDeviceInfoBootParseXML whenever
there is a <boot> element. When we are only parsing a single device,
we don't care whether or not any specified boot order is consistent
with the rest of the domain; we will always do this check later (in
the current case, we do it by verifying that the net bootIndex exactly
matches the old bootIndex).
Laine Stump [Wed, 16 Jan 2013 17:46:01 +0000 (12:46 -0500)]
network: use bandwidth from portgroup when appropriate
The bandwidth plug and unplug functions were assuming that an
interface's bandwidth setting was always specified directly in the
domain's <interface> definition, but that's not necessarily true - it
could have been obtained from a <portgroup> definition in the network
definition. This patch fixes those functions to use
virDomainNetGetActualBandwidth(), which gets the bandwidth pointer
from iface->data.network.actual if it exists, otherwise returns
iface->bandwidth.
John Ferlan [Tue, 15 Jan 2013 18:35:34 +0000 (13:35 -0500)]
network: Resolve some issues around vlan copying
Remove extraneous check for 'netdef' when dereferencing for vlan.nTags.
Prior code would already check if netdef was NULL.
Coverity complained about a path where the 'vlan' was potentially valid,
but a prior checks may not have allocated 'iface->data.network.actual',
so like other paths it needs to be allocated on the fly.
Move the copying of vlan up earlier in networkAllocateActualDevice, so
that actual.type gets properly set.
Since the first assignment to vlan is redundant except in the case of
jumping immediately to validate from the start of the function,
eliminate its initial setting at the top of the function in favor of
calling the helper function virDomainNetGetActualVlan() (which doesn't
depend on the local vlan pointer being initialized) down at validate:
Avoid integer wrap on remotePortMax in QEMU driver
The QEMU driver default max port is 65535, but it then increments
this by 1 to 65536. This maps to 0 in an unsigned short :-( This
was apparently done so that for() loops could use "< max" instead
of "<= max". Remove this insanity and just make the loop do the
right thing.
Claudio Bley [Wed, 9 Jan 2013 13:03:50 +0000 (14:03 +0100)]
Move comments after enum members
The api builder always associates comments to the last member it read,
not to the current member even if there was a comment for the previous
member and a comma was already seen.
This has the effect that the comment for the previous member gets
overwritten and the current member has no comment at all.