Eric Blake [Tue, 18 Mar 2014 21:41:28 +0000 (15:41 -0600)]
vmware: fix parse of disk source
While writing disk source refactoring, I discovered that conversion
from XML to vmware modified the disk source in place; if the same
code is reached twice, the second call behaves differently because
the first call didn't clean up its mess.
I get the SIGSEGV when starting the domain. The thing is, when
starting a domain, we check for its disk presence. For some reason,
when determining the disk chain, we parse the <seclabel/> (don't ask
me why). However, there's no label attribute in the XML, so we end up
calling virParseOwnershipIds() over NULL string:
[Switching to Thread 0x7ffff10c4700 (LWP 30956)]
__strchr_sse42 () at ../sysdeps/x86_64/multiarch/strchr.S:136
136 ../sysdeps/x86_64/multiarch/strchr.S: No such file or directory.
(gdb) bt
#0 __strchr_sse42 () at ../sysdeps/x86_64/multiarch/strchr.S:136
#1 0x00007ffff749f800 in virParseOwnershipIds (label=0x0, uidPtr=uidPtr@entry=0x7ffff10c2df0, gidPtr=gidPtr@entry=0x7ffff10c2df4) at util/virutil.c:2115
#2 0x00007fffe929f006 in qemuDomainGetImageIds (gid=0x7ffff10c2df4, uid=0x7ffff10c2df0, disk=0x7fffe40cb000, vm=0x7fffe40a6410, cfg=0x7fffe409ae00) at qemu/qemu_domain.c:2385
#3 qemuDomainDetermineDiskChain (driver=driver@entry=0x7fffe40120e0, vm=vm@entry=0x7fffe40a6410, disk=disk@entry=0x7fffe40cb000, force=force@entry=false) at qemu/qemu_domain.c:2414
#4 0x00007fffe929f128 in qemuDomainCheckDiskPresence (driver=driver@entry=0x7fffe40120e0, vm=vm@entry=0x7fffe40a6410, cold_boot=cold_boot@entry=true) at qemu/qemu_domain.c:2250
#5 0x00007fffe92b6fc8 in qemuProcessStart (conn=conn@entry=0x7fffd4000b60, driver=driver@entry=0x7fffe40120e0, vm=vm@entry=0x7fffe40a6410, migrateFrom=migrateFrom@entry=0x0, stdin_fd=stdin_fd@entry=-1, stdin_path=stdin_path@entry=0x0, snapshot=snapshot@entry=0x0,
vmop=vmop@entry=VIR_NETDEV_VPORT_PROFILE_OP_CREATE, flags=flags@entry=1) at qemu/qemu_process.c:3813
#6 0x00007fffe93087e8 in qemuDomainObjStart (conn=0x7fffd4000b60, driver=driver@entry=0x7fffe40120e0, vm=vm@entry=0x7fffe40a6410, flags=flags@entry=0) at qemu/qemu_driver.c:6051
#7 0x00007fffe9308e32 in qemuDomainCreateWithFlags (dom=0x7fffcc000d50, flags=0) at qemu/qemu_driver.c:6105
#8 0x00007ffff753c5cc in virDomainCreate (domain=domain@entry=0x7fffcc000d50) at libvirt.c:8861
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Chegu Vinod [Thu, 6 Feb 2014 23:44:36 +0000 (15:44 -0800)]
libvirt support to force convergence of live guest migration
Busy enterprise workloads hosted on large sized VM's tend to dirty
memory faster than the transfer rate achieved via live guest migration.
Despite some good recent improvements (& using dedicated 10Gig NICs
between hosts) the live migration may NOT converge.
Recently support was added in qemu (version 1.6) to allow a user to
choose if they wish to force convergence of their migration via a
new migration capability : "auto-converge". This feature allows for qemu
to auto-detect lack of convergence and trigger a throttle-down of the
VCPUs.
This patch includes the libvirt support needed to trigger this
feature. (Testing is in progress)
Wang Yufei [Thu, 20 Mar 2014 07:14:01 +0000 (07:14 +0000)]
cgroup: Fix start VMs coincidently failed
When I start multi VMs coincidently and any of the cgroup directories
named machine doesn't exist. There's a chance that VM start failed because
of creating directory failed:
Unable to initialize /machine cgroup: File exists
When the errno returned by mkdir in virCgroupMakeGroup is EEXIST,
we should pass it through and continue to start the VM. Signed-off-by: Wang Yufei <james.wangyufei@huawei.com>
The caller may not want all DBus error conditions to be turned
into libvirt errors, so provide a way for the caller to get
back the full DBusError object. They can then check the errors
and only report those that they consider to be fatal.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Introduce alternate way to encode/decode arrays in DBus messages
Currently the DBus helper APIs require the values for an array
to be passed inline in the variadic argument list. This change
introduces support for passing arrays using a pointer to a plain
C array of the basic type. This is of particular benefit for
decoding messages when you don't know how many array elements
are being received.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The dbus_connection_send_with_reply_and_block method will
automatically call dbus_set_error_from_message for us. We
mistakenly thought we had todo it because of a flaw in the
systemd unit test mock impl. The latter should have directly
set the error object, instead of creating an error message
object.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The virDBusMessageRead method should not have side-effects on
the message parameter passed in, so unref'ing it is wrong.
The caller should unref only when they decided they are done
with it.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Add DBus helper methods for creating reply messages
The test suites often have to create DBus method reply messages
with payloads. Create two helpers for simplifying the process
of creating replies with payloads.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Zhou Yimin [Fri, 21 Mar 2014 06:52:40 +0000 (14:52 +0800)]
virlog: Modify virLogParseDefaultPriority's comment of return value
virLogParseDefaultPriority's successful return value is the same as
virLogSetDefaultPriority's successful return value. So it should be 0
rather than the parsed log level.
Jiri Denemark [Thu, 20 Mar 2014 21:34:00 +0000 (22:34 +0100)]
conf: Introduce virDomainDeviceGetInfo API
The offset of virDomainDeviceInfo structure within a device definition
varies with device type and some types do not contain the info structure
at all. This new API makes it easier to access the info structure from a
generic virDomainDeviceDef structure.
Jiri Denemark [Thu, 20 Mar 2014 12:39:20 +0000 (13:39 +0100)]
Pass action to virDomainDefCompatibleDevice
When checking compatibility of a device with a domain definition, we
should know what we're going to do with the device. Because we may need
to check for different things when we're attaching a new device versus
detaching an existing device.
Jiri Denemark [Thu, 20 Mar 2014 12:04:06 +0000 (13:04 +0100)]
Fix usage of virDomainDefCompatibleDevice
A device needs to be checked for compatibility with the domain
definition it corresponds to. Specifically, for VIR_DOMAIN_AFFECT_CONFIG
case we should check against persistent def rather than active def.
When qemu dies early after connecting to its monitor but before we
actually try to read something from the monitor, we would just fail
domain start with useless message:
"An error occurred, but the cause is unknown"
This is because the real error gets reported in a monitor EOF handler
executing within libvirt's event loop.
The fix is to take any error set in qemuMonitor structure and propagate
it into the thread-local error when qemuMonitorClose is called and no
thread-local error is set.
Eric Blake [Thu, 6 Feb 2014 21:46:52 +0000 (14:46 -0700)]
qemu: allow filtering events by regex
When listening for a subset of monitor events, it can be tedious
to register for each event name in series; nicer is to register
for multiple events in one go. Implement a flag to use regex
interpretation of the event filter.
While at it, prove how much I hate the shift key, by adding a
way to filter for 'shutdown' instead of 'SHUTDOWN'. :)
Eric Blake [Fri, 31 Jan 2014 14:02:25 +0000 (07:02 -0700)]
qemu: enable monitor event filtering by name
Filtering monitor events by name requires tracking the name for
the duration of the filtering. In order to free the name, I
found it easiest to just piggyback on the user's freecb function,
which gets called when the event is deregistered.
For events without a name filter, we have the design of multiple
client registrations sharing a common server registration, because
the server side uses the same callback function and we reject
duplicate use of the same function. But with events in the mix,
we want to be able to allow the same function pointer to be used
with more than one event name. The solution is to tweak the
duplicate detection code to only act when there is no additional
filtering; if name filtering is in use, there is exactly one
client registration per server registration. Yes, this means
that there is no longer a bound on the number of server
registrations possible, so a malicious client could repeatedly
register for the same name event to exhaust server memory. On
the other hand, we already restricted monitor events to require
write access (compared to normal events only needing read access),
and separated it into the intentionally unsupported
libvirt-qemu.so, with documentation that using this function is
for debug purposes only; so it is not a security risk worth
worrying about a client trying to abuse multiple registrations.
* src/conf/domain_event.c (virDomainQemuMonitorEventData): New
struct.
(virDomainQemuMonitorEventFilter)
(virDomainQemuMonitorEventCleanup): New functions.
(virDomainQemuMonitorEventDispatchFunc)
(virDomainQemuMonitorEventStateRegisterID): Use new struct.
* src/conf/object_event.c (virObjectEventCallbackListCount)
(virObjectEventCallbackListAddID)
(virObjectEventCallbackListRemoveID)
(virObjectEventCallbackListMarkDeleteID): Drop duplicate detection
when filtering is in effect.
Eric Blake [Thu, 30 Jan 2014 00:14:44 +0000 (17:14 -0700)]
qemu: enable monitor event reporting
Wire up all the pieces to send arbitrary qemu events to a
client using libvirt-qemu.so. If the extra bookkeeping of
generating event objects even when no one is listening turns
out to be noticeable, we can try to further optimize things
by adding a counter for how many connections are using events,
and only dump events when the counter is non-zero; but for
now, I didn't think it was worth the code complexity.
* src/qemu/qemu_driver.c
(qemuConnectDomainQemuMonitorEventRegister)
(qemuConnectDomainQemuMonitorEventDeregister): New functions.
* src/qemu/qemu_monitor.h (qemuMonitorEmitEvent): New prototype.
(qemuMonitorDomainEventCallback): New typedef.
* src/qemu/qemu_monitor_json.c (qemuMonitorJSONIOProcessEvent):
Report events.
* src/qemu/qemu_monitor.c (qemuMonitorEmitEvent): New function, to
pass events through.
* src/qemu/qemu_process.c (qemuProcessHandleEvent): Likewise.
Eric Blake [Wed, 29 Jan 2014 22:30:44 +0000 (15:30 -0700)]
qemu: wire up RPC for qemu monitor events
These are the first async events in the qemu protocol, so this
patch looks rather big compared to most RPC additions. However,
a large majority of this patch is just mechanical copy-and-paste
from recently-added network events. It didn't help that this
is also the first virConnect rather than virDomain prefix
associated with a qemu-specific API.
* src/remote/qemu_protocol.x (qemu_*_domain_monitor_event_*): New
structs and RPC messages.
* src/rpc/gendispatch.pl: Adjust naming conventions.
* daemon/libvirtd.h (daemonClientPrivate): Track qemu events.
* daemon/remote.c (remoteClientFreeFunc): Likewise.
(remoteRelayDomainQemuMonitorEvent)
(qemuDispatchConnectDomainMonitorEventRegister)
(qemuDispatchConnectDomainMonitorEventDeregister): New functions.
* src/remote/remote_driver.c (qemuEvents): Handle qemu events.
(doRemoteOpen): Register for events.
(remoteNetworkBuildEventLifecycle)
(remoteConnectDomainQemuMonitorEventRegister)
(remoteConnectDomainQemuMonitorEventDeregister): New functions.
* src/qemu_protocol-structs: Regenerate.
Eric Blake [Tue, 31 Dec 2013 13:33:42 +0000 (06:33 -0700)]
qemu: create object for qemu monitor events
Create qemu monitor events as a distinct class to normal domain
events, because they will be filtered differently. For ease of
review, the logic for filtering by event name is saved for a later
patch.
* src/conf/domain_event.c (virDomainQemuMonitorEventClass): New
class.
(virDomainEventsOnceInit): Register it.
(virDomainQemuMonitorEventDispose, virDomainQemuMonitorEventNew)
(virDomainQemuMonitorEventDispatchFunc)
(virDomainQemuMonitorEventStateRegisterID): New functions.
* src/conf/domain_event.h (virDomainQemuMonitorEventNew)
(virDomainQemuMonitorEventStateRegisterID): New prototypes.
* src/libvirt_private.syms (conf/domain_conf.h): Export them.
Eric Blake [Thu, 19 Dec 2013 14:06:35 +0000 (07:06 -0700)]
qemu: new API for tracking arbitrary monitor events
Several times in the past, qemu has implemented a new event,
but libvirt has not yet caught up to reporting that event to
the user applications. While it is possible to track libvirt
logs to see that an unknown event was received and ignored,
it would be nicer to copy what 'virsh qemu-monitor-command'
does, and expose this information to the end developer as
one of our unsupported qemu-specific commands.
If you find yourself needing to use this API for more than
just development purposes, please ask on the libvirt list
for a supported counterpart event to be added in libvirt.so.
While the supported virConnectDomainEventRegisterAny() API
takes an id which determines the signature of the callback,
this version takes a string filter and always uses the same
signature. Furthermore, I chose to expose this as a new API
instead of trying to add a new eventID at the top level, in
part because the generic option lacks event name filtering,
and in part because the normal domain event namespace should
not be polluted by a qemu-only event. I also added a flags
argument; unused for now, but we might decide to use it to
allow a user to request event names by glob or regex instead
of literal match.
This API intentionally requires full write access (while
normal event registration is allowed on read-only clients);
this is in part due to the fact that it should only be used
by debugging situations, and in part because the design of
per-event filtering in later patches ended up allowing for
duplicate registrations that could potentially be abused to
exhaust server memory - requiring write privileges means
that such abuse will not serve as a denial of service attack
against users with higher privileges.
* include/libvirt/libvirt-qemu.h
(virConnectDomainQemuMonitorEventCallback)
(virConnectDomainQemuMonitorEventRegister)
(virConnectDomainQemuMonitorEventDeregister): New prototypes.
* src/libvirt-qemu.c (virConnectDomainQemuMonitorEventRegister)
(virConnectDomainQemuMonitorEventDeregister): New functions.
* src/libvirt_qemu.syms (LIBVIRT_QEMU_1.2.1): Export them.
* src/driver.h (virDrvConnectDomainQemuMonitorEventRegister)
(virDrvConnectDomainQemuMonitorEventDeregister): New callbacks.
Ján Tomko [Thu, 20 Mar 2014 15:35:00 +0000 (16:35 +0100)]
Ignore missing files on pool refresh
If we cannot stat/open a file on pool refresh, returning -1 aborts
the refresh and the pool is undefined.
Only treat missing files as fatal unless VolOpenCheckMode is called
with the VIR_STORAGE_VOL_OPEN_ERROR flag. If this flag is missing
(when it's called from virStorageBackendProbeTarget in
virStorageBackendFileSystemRefresh), only emit a warning and return
-2 to let the caller skip over the file.
Ján Tomko [Thu, 20 Mar 2014 15:42:52 +0000 (16:42 +0100)]
Ignore char devices in storage pools by default
Without this, using /dev/mapper as a directory pool
fails in virStorageBackendUpdateVolTargetInfoFD:
cannot seek to end of file '/dev/mapper/control': Illegal seek
Require K&R styled curly braces around function bodies
Although not explicitly requested, we are using K&R (or Kernel)
indentation for curly braces around functions in HACKING file and most
of the code. Using grep -P, this patch add the syntax-check rule for
it (while skipping all the false positives with foreach constructs).
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Scott Sullivan [Thu, 20 Mar 2014 15:05:14 +0000 (16:05 +0100)]
is_selinux_enabled() returns -1 on error, account for this.
Per the documentation, is_selinux_enabled() returns -1 on error.
Account for this. Previously when -1 was being returned the condition
would still be true. I was noticing this because on my system that has
selinux disabled I was getting this in the libvirt.log every 5
seconds:
error : virIdentityGetSystem:173 : Unable to lookup SELinux process context: Invalid argument
With this patch applied, I no longer get these messages every 5
seconds. I am submitting this in case its deemed useful for inclusion.
Anyone have any comments on this change? This is a patch off current
master.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Wojciech Macek [Thu, 20 Mar 2014 08:39:23 +0000 (09:39 +0100)]
bhyve: host API support
New functionalities:
- connectGetMaxVcpus - on bhyve hardcode this value to 16
- nodeGetFreeMemory - do not use physmem_get on FreeBSD, since
it might get wrong value on systems with
more than 100GB of RAM
- nodeGetCPUMap - wrapper only for mapping function, currently not
supported by FreeBSD
- nodeSet/GetMemoryParameters - wrapper only for future improvements,
currently not supported by FreeBSD
The virSocketAddrMask method did not initialize all fields
in the sockaddr_in6 struct. In paticular the 'sin6_scope_id'
field could contain random garbage, which would in turn
affect the result of any later virSocketAddrFormat calls.
This led to ip6tables rules in the FORWARD chain which
matched on random garbage sin6_scope_id. Fortunately these
were ACCEPT rules, so the impact was merely that desired
traffic was blocked, rather than undesired traffic allowed.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Jim Fehlig [Wed, 26 Feb 2014 19:04:34 +0000 (12:04 -0700)]
libxl: move libxlVmCleanup{,Job} to libxl_domain
Move libxlVmCleanup and libxlVmCleanupJob from libxl_driver to
libxl_domain for use by other libxl modules. For consistency,
rename to libxlDomainCleanup and libxlDomainCleanupJob.
Michal Privoznik [Wed, 19 Mar 2014 17:10:34 +0000 (18:10 +0100)]
virNetClientSetTLSSession: Restore original signal mask
Currently, we use pthread_sigmask(SIG_BLOCK, ...) prior to calling
poll(). This is okay, as we don't want poll() to be interrupted.
However, then - immediately as we fall out from the poll() - we try to
restore the original sigmask - again using SIG_BLOCK. But as the man
page says, SIG_BLOCK adds signals to the signal mask:
SIG_BLOCK
The set of blocked signals is the union of the current set and the set argument.
Therefore, when restoring the original mask, we need to completely
overwrite the one we set earlier and hence we should be using:
SIG_SETMASK
The set of blocked signals is set to the argument set.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Commit 49b59a15 fixed one problem but masks another one related to pointer
freeing.
Avoid putting of the virNWFilterSnoopReq once the thread has been started.
It belongs to the thread and the thread will call virNWFilterSnoopReqPut() on it.
Signed-off-by: Stefan Berger <stefanb@linux.vnet.ibm.com>
While running qemuhotplugtest, it was found that valgrind pointed out
the following memory leak:
==7906== 5 bytes in 1 blocks are definitely lost in loss record 7 of 121
==7906== at 0x4A069EE: malloc (vg_replace_malloc.c:270)
==7906== by 0x3E782A754D: xmlStrndup (in /usr/lib64/libxml2.so.2.7.6)
==7906== by 0x4CDAE03: virDomainDeviceInfoParseXML.isra.32 (domain_conf.c:3685)
==7906== by 0x4CE3BB9: virDomainNetDefParseXML (domain_conf.c:6707)
==7906== by 0x4CFBA08: virDomainDefParseXML (domain_conf.c:12235)
==7906== by 0x4CFBC1E: virDomainDefParseNode (domain_conf.c:13039)
==7906== by 0x4CFBD95: virDomainDefParse (domain_conf.c:12981)
==7906== by 0x41FEB4: testQemuHotplug (qemuhotplugtest.c:66)
==7906== by 0x420F41: virtTestRun (testutils.c:201)
==7906== by 0x41F287: mymain (qemuhotplugtest.c:422)
==7906== by 0x4216BD: virtTestMain (testutils.c:784)
==7906== by 0x3E6CE1ED1C: (below main) (libc-start.c:226)
...and 10 more.
Problem is, since 20745748 we do both, parse <alias/> elements from
XML files and call qemuAssignDeviceAliases(). While generating runtime
info for domain at runtime is just fine in the test, we can parse just
inactive XML and remove all <alias/>-es from the XML files.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Add ability to register callback for virCommand dry run
To allow for fault injection of the virCommand dry run,
add the ability to register a callback. The callback will
be passed the argv, env and stdin buffer and is expected
to return the exit status and optionally fill stdout and
stderr buffers.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Remove pointless return values in nwfilter methods
Many nwfilter methods have an int return value but only ever
return 0 and their callers never check the return value either.
These methods can all be void.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Remove 'int stopOnError' parameters in nwfilter methods
Many nwfilter methods have an 'int stopOnError' parameter but
with 1 exception, the callers always pass '1'. The parameter
can therefore be removed from all except one method. That method
will be changed to 'bool stopOnError'
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
A lot of methods have a 'bool incoming' parameter but then
do (incoming) ? ... : .... The round brackets here add nothing
to the code so can be removed.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Jim Fehlig [Mon, 17 Mar 2014 20:22:44 +0000 (14:22 -0600)]
libxl: fix framebuffer port setting for HVM domains
libxl uses the libxl_vnc_info and libxl_sdl_info fields from the
hvm union in libxl_domain_build_info struct when generating QEMU
args for VNC or SDL. These fields were left unset by the libxl
driver, causing libxl to ignore any user settings. E.g. with
<graphics type='vnc' port='5950'/>
port would be ignored and QEMU would instead be invoked with
-vnc 127.0.0.1:0,to=99
Unlike the libxl_domain_config struct, the libxl_domain_build_info
contains only a single libxl_vnc_info and libxl_sdl_info, so
populate these fields from the first vfb in
libxl_domain_config->vfbs.
Signed-off-by: Jim Fehlig <jfehlig@suse.com> Signed-off-by: David Kiarie <davidkiarie4@gmail.com>
Eric Blake [Tue, 18 Mar 2014 20:10:27 +0000 (14:10 -0600)]
maint: avoid unbalanced {} across vbox #ifdef
Emacs is fairly good about navigating across function and scope
boundaries, provided that the code has balanced {}. The vbox
code, however, violated that premise, by splitting 'if () {'
across several #ifdef branches, but sharing the '} else {...}'
outside of the branches. The extra lines of code is worth my
sanity, in a function that is already a horrendous 1100+ lines
long.
* src/vbox/vbox_tmpl.c (vboxDomainGetXMLDesc) Duplicate code
rather than trying to share else branch across #ifdef.
A earlier commit changed the global log buffer so that it only
records messages that are explicitly requested via the log
filters setting. This removes the performance burden, and
improves the signal/noise ratio for messages in the global
buffer. At the same time though, it is somewhat pointless, since
all the recorded log messages are already going to be sent to an
explicit log output like syslog, stderr or the journal. The
global log buffer is thus just duplicating this data on stderr
upon crash.
The log_buffer_size config parameter is left in the augeas
lens to prevent breakage for users on upgrade. It is however
completely ignored hereafter.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Switch to filtering based on log source name instead of filename
Currently the log filter strings are used in a string comparison
against the source filename each time log message is emitted.
If no log filters at all are set, there's obviously no string
comparison to be done. If any single log filter is set though,
this imposes a compute burden on every logging call even if logs
from the file in question are disabled. This string comparison
must also be done while the logging mutex is held, which has
implications for concurrency when multiple threads are emitting
log messages.
This changes the log filtering to be done based on the virLogSource
object name. The virLogSource struct is extended to contain
'serial' and 'priority' fields. Any time the global log filter
rules are changed a global serial number is incremented. When a
log message is emitted, the serial in the virLogSource instance
is compared with the global serial number. If out of date, then
the 'priority' field in the virLogSource instance is updated based
on the new filter rules. The 'priority' field is checked to see
whether the log message should be sent to the log outputs.
The comparisons of the 'serial' and 'priority' fields are done
with no locks held. So in the common case each logging call has
an overhead of 2 integer comparisons, with no locks held. Only
if the decision is made to forward the message to the log output,
or if the 'serial' value is out of date do locks need to be
acquired.
Technically the comparisons of the 'serial' and 'priority' fields
should be done with locks held, or using atomic operations. Both
of these options have a notable performance impact, however, and
since all writes a protected by a global mutex, it is believed
that worst case behaviour where the fields are read concurrently
with being written would merely result in an mistaken emission
or dropping of the log message in question. This is an acceptable
tradeoff for the performance benefit of avoiding locking.
As a quick benchmark, a demo program that registers 500 file
descriptors with the event loop (eg equiv of 500 QEMU monitor
commands), creates pending read I/O on every FD, and then runs
virEventRunDefaultImpl() took 4.6 seconds to do 51200 iterations.
After this optimization it only takes 3.3 seconds, with the log
APIs no longer being a relevant factor in the running time.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Any source file which calls the logging APIs now needs
to have a VIR_LOG_INIT("source.name") declaration at
the start of the file. This provides a static variable
of the virLogSource type.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Turn virLogSource into a struct instead of an enum
As part of the goal to get away from doing string matching on
filenames when deciding whether to emit a log message, turn
the virLogSource enum into a struct which contains a log
"name". There will eventually be one virLogSource instance
statically declared per source file. To minimise churn in this
commit though, a single global instance is used.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Move dtrace probe macros into separate header file
The dtrace probe macros rely on the logging API. We can't make
the internal.h header include the virlog.h header though since
that'd be a circular include. Instead simply split the dtrace
probes into their own header file, since there's no compelling
reason for them to be in the main internal.h header.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Refactor code that skips logging of error messages
The error reporting code will invoke a callback when any error
is raised and the default callback will print to stderr. The
virRaiseErrorFull method also sends all error messages on to the
logging code, which also prints to stderr by default. To avoid
duplicated data on stderr, the logging code has some logic to
skip emission when no log outputs are configured, which checks
whether the virLogSource == VIR_LOG_FROM_ERROR.
Meanwhile the libvirtd daemon can register another callback which
is used to reduce log message priority from error to a lower level.
When this is used we do want messages to end up on stderr, so the
error code will conditionally use either VIR_LOG_FROM_FILE or
VIR_LOG_FROM_ERROR depending on whether such a callback is provided.
This will all complicate later refactoring. By pushing the checks
for whether a log output is present up a level into the error code,
the special cases can be isolated in one place.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
projects / people / liuw / libxenctrl-split / libvirt.git / log