March 2014 - Devel - libvirt List Archives

[libvirt] [libvirt-java] [PATCH 0/2] Ensure JNA callbacks cannot be GCed
by Chris Ellis 26 Mar '14

26 Mar '14

Hi. This is a small patch set to fix a few issues I've discovered while testing Claudio's patch set. The first patch is trivial, it adds the ConnectionCloseListener interface and corresponding enum, which were missing. The second patch ensures that the JNA callbacks cannot be garbage collected whilst still registered with the libvirt C library. Durring testing, I was finding that events would stop working. Wireshark showed the remote daemon sending the event and some time spent tracing through the libvirt calls led me to discover that the JNA callback objects were being GCed. The JNA documentation states that a reference to the Callback object must be held whilst it is in use by the C layer. This patch updates the map of event listeners to also hold a reference to the callback object. Regards, Chris Chris Ellis (2): Add close listener types Ensure JNA callbacks cannot get GCed src/main/java/org/libvirt/Connect.java | 37 ++++++++++++++-------- .../org/libvirt/event/ConnectionCloseListener.java | 9 ++++++ .../org/libvirt/event/ConnectionCloseReason.java | 8 +++++ 3 files changed, 41 insertions(+), 13 deletions(-) create mode 100644 src/main/java/org/libvirt/event/ConnectionCloseListener.java create mode 100644 src/main/java/org/libvirt/event/ConnectionCloseReason.java -- 1.8.4.5

2 5

[libvirt] question for <interface> in ethernet type and multiple queues in KVM (start VM failed)
by Wangrui (K) 26 Mar '14

26 Mar '14

Hi, I use ethernet vif for VM (libvirt 1.1.0 qemu 1.5.1). xml as such <interface type='ethernet'> <script path='' /> <target dev='tap_mq'/> <mac address='52:54:00:19:9e:4e'/> <model type='virtio' /> <driver name='vhost' queues='2'/> </interface> tap_mq is a tap device with multi_queue property which was created on host by ip command, such as follows #ip tuntap add tap_mq mode tap multi_queue I want to use multiple queue feature for this vif so I set queues='2' in xml. When start VM , an error occurs "error: unsupported configuration: Multiqueue network is not supported for: ethernet". This error is reported from function qemuBuildInterfaceCommandLine /* Currently nothing besides TAP devices supports multiqueue. */ if (net->driver.virtio.queues > 0 && !(actualType == VIR_DOMAIN_NET_TYPE_NETWORK || actualType == VIR_DOMAIN_NET_TYPE_BRIDGE)) { virReportError(VIR_ERR_CONFIG_UNSUPPORTED, _("Multiqueue network is not supported for: %s"), virDomainNetTypeToString(actualType)); return -1; } The comment specifies that only TAP device supports multiqueue. But VIR_DOMAIN_NET_TYPE_ETHERNET is also TAP device I think. IMHO, libvirt can build qemu commandline as "-device virtio-net,netdev=net1,mq=on,vectors=5,mac=52:54:00:19:9e:4e" "-netdev type=tap,ifname=tap0,id=net1,vhost=on,vhostforce=on,queues=2". Qemu can open tapfds and vhostfds by itself other than libvirt opens and transfers them to qemu in this case. Or there's other way to use multiqueue feature in ethernet tpye ? Regards, Rui

4 4

[libvirt] [PATCH] maint: use $(SED) instead of sed for syntax-check
by Roman Bogorodskiy 26 Mar '14

26 Mar '14

Some syntax-check rules use GNU sed specific regexps, so allow to override which sed would be used to fix 'syntax-check' for non GNU-userland systems. --- cfg.mk | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) diff --git a/cfg.mk b/cfg.mk index 207dfeb..bd984bd 100644 --- a/cfg.mk +++ b/cfg.mk @@ -614,7 +614,7 @@ sc_libvirt_unmarked_diagnostics: $(_sc_search_regexp) @{ grep -nE '\<$(func_re) *$.*;$$' $$($(VC_LIST_EXCEPT)); \ grep -A1 -nE '\<$(func_re) *\(.*,$$' $$($(VC_LIST_EXCEPT)); } \ - | sed 's/_("\([^\"]\|\\.$\+"//;s/[ ]"%s"//' \ + | $(SED) 's/_("$[^\"]\|\\.$\+"//;s/[ ]"%s"//' \ | grep '[ ]"' && \ { echo '$(ME): found unmarked diagnostic(s)' 1>&2; \ exit 1; } || : @@ -639,7 +639,7 @@ sc_prohibit_newline_at_end_of_diagnostic: sc_prohibit_diagnostic_without_format: @{ grep -nE '\<$(func_re) *$.*;$$' $$($(VC_LIST_EXCEPT)); \ grep -A2 -nE '\<$(func_re) *\(.*,$$' $$($(VC_LIST_EXCEPT)); } \ - | sed -rn -e ':l; /[,"]$$/ {N;b l;}' \ + | $(SED) -rn -e ':l; /[,"]$$/ {N;b l;}' \ -e '/(xenapiSessionErrorHandler|vah_(error|warning))/d' \ -e '/\<$(func_re) *\([^"]*"([^%"]|"\n[^"]*")*"[,)]/p' \ | grep -vE 'VIR_ERROR' && \ @@ -661,7 +661,7 @@ sc_prohibit_useless_translation: # or \n on one side of the split. sc_require_whitespace_in_translation: @grep -n -A1 '"$$' $$($(VC_LIST_EXCEPT)) \ - | sed -ne ':l; /"$$/ {N;b l;}; s/"\n[^"]*"/""/g; s/\\n/ /g' \ + | $(SED) -ne ':l; /"$$/ {N;b l;}; s/"\n[^"]*"/""/g; s/\\n/ /g' \ -e '/_(.*[^\ ]""[^\ ]/p' | grep . && \ { echo '$(ME): missing whitespace at line split' 1>&2; \ exit 1; } || : @@ -681,11 +681,11 @@ sc_preprocessor_indentation: sc_spec_indentation: @if cppi --version >/dev/null 2>&1; then \ for f in $$($(VC_LIST_EXCEPT) | grep '\.spec\.in$$'); do \ - sed -e 's|#|// #|; s|%ifn*\(arch$* |#if a // |' \ + $(SED) -e 's|#|// #|; s|%ifn*$arch$* |#if a // |' \ -e 's/%$else\|endif\|define$/#\1/' \ -e 's/^$ *$\1\1\1#/#\1/' \ -e 's|^$ *[^#/ ]$|// \1|; s|^$ */[^/]$|// \1|' $$f \ - | cppi -a -c 2>&1 | sed "s|standard input|$$f|"; \ + | cppi -a -c 2>&1 | $(SED) "s|standard input|$$f|"; \ done | { if grep . >&2; then false; else :; fi; } \ || { echo '$(ME): incorrect preprocessor indentation' 1>&2; \ exit 1; }; \ @@ -777,7 +777,7 @@ sc_prohibit_cross_inclusion: # elements added to the enum by using a _LAST marker. sc_require_enum_last_marker: @grep -A1 -nE '^[^#]*VIR_ENUM_IMPL *\(' $$($(VC_LIST_EXCEPT)) \ - | sed -ne '/VIR_ENUM_IMPL[^,]*,$$/N' \ + | $(SED) -ne '/VIR_ENUM_IMPL[^,]*,$$/N' \ -e '/VIR_ENUM_IMPL[^,]*,[^,]*[^_,][^L,][^A,][^S,][^T,],/p' \ -e '/VIR_ENUM_IMPL[^,]*,[^,]\{0,4\},/p' \ | grep . && \ @@ -878,7 +878,7 @@ ifeq (0,$(MAKELEVEL)) # b653eda3ac4864de205419d9f41eec267cb89eeb # # Keep this logic in sync with autogen.sh. - _submodule_hash = sed 's/^[ +-]//;s/ .*//' + _submodule_hash = $(SED) 's/^[ +-]//;s/ .*//' _update_required := $(shell \ cd '$(srcdir)'; \ test -d .git || { echo 0; exit; }; \ -- 1.8.4.3

3 4

[libvirt] [PATCH] qemu: remove redundant virQEMUDriverGetConfig
by Tomoki Sekiyama 26 Mar '14

26 Mar '14

qemuDomainSetSchedulerParametersFlags() calls virQEMUDriverGetConfig() twice and makes the reference counter leak. This removes redundant call. --- src/qemu/qemu_driver.c | 1 - 1 file changed, 1 deletion(-) diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index 76f789e..ca971f3 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -9050,7 +9050,6 @@ qemuDomainSetSchedulerParametersFlags(virDomainPtr dom, if (virDomainSetSchedulerParametersFlagsEnsureACL(dom->conn, vm->def, flags) < 0) goto cleanup; - cfg = virQEMUDriverGetConfig(driver); if (!cfg->privileged) { virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s", _("CPU tuning is not available in session mode"));

4 3

[libvirt] [PATCH 0/3] Resolve recent/new Coverity warnings
by John Ferlan 26 Mar '14

26 Mar '14

For virlog.c and commandtest.c - it seems recent changes within the modules have allowed Coverity to dig deeper and find different issues. Not sure why Coverity has that "ability"... Since I was making changes, figured I'd clean up the lxc_controller fd leak as well - these should clean the Jenkins build environment. John Ferlan (3): Coverity: Resolve a CHECKED_RETURN message Coverity: Resolve a FORWARD_NULL Coverity: Resolve a RESOURCE_LEAK src/lxc/lxc_controller.c | 14 ++++++++++---- src/util/virlog.c | 2 +- tests/commandtest.c | 1 + 3 files changed, 12 insertions(+), 5 deletions(-) -- 1.8.5.3

3 8

[libvirt] [libvirt-java] [PATCH] Set Java source level in pom
by Chris Ellis 26 Mar '14

26 Mar '14

Hi Simple patch to set the Java source level within the pom, this makes it easier to import the project into Eclipse. I've also added an ignore for a local pom file. I've added plugins to publish the sources and javadocs into Maven. This allows Maven (and IDE integration) to automatically use the javadocs and source. This should make life easier for users of libvirt-java. Regards, Chris Chris Ellis (1): Update pom to include Java source level .gitignore | 1 + pom.xml.in | 41 +++++++++++++++++++++++++++++++++++++++++ 2 files changed, 42 insertions(+) -- 1.8.4.5

3 6

[libvirt] [PATCHv2 0/2] Fix zero cpu shares handling
by Ján Tomko 26 Mar '14

26 Mar '14

Treat 0 as a valid value for cputune shares. v1: https://www.redhat.com/archives/libvir-list/2014-March/msg00215.html v2: squashed the fixup that was sent as a separate patch to the first commit and fixed two more occurences (one of them pointed out by Martin) Ján Tomko (2): Treat zero cpu shares as a valid value Show the real cpu shares value in live xml src/conf/domain_conf.c | 12 ++++---- src/conf/domain_conf.h | 1 + src/lxc/lxc_cgroup.c | 13 ++++++-- src/lxc/lxc_driver.c | 8 ++++- src/lxc/lxc_native.c | 8 +++-- src/parallels/parallels_driver.c | 1 + src/qemu/qemu_cgroup.c | 14 ++++++--- src/qemu/qemu_command.c | 2 +- src/qemu/qemu_driver.c | 13 ++++++-- src/vmx/vmx.c | 3 +- .../qemuxml2argv-cputune-zero-shares.args | 5 ++++ .../qemuxml2argv-cputune-zero-shares.xml | 35 ++++++++++++++++++++++ tests/qemuxml2argvtest.c | 1 + tests/qemuxml2xmltest.c | 1 + 14 files changed, 97 insertions(+), 20 deletions(-) create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-cputune-zero-shares.args create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-cputune-zero-shares.xml -- 1.8.3.2

2 5

[libvirt] [libivrt][RFC]Send "netdev_del" command after receive the DEVICE_DELETED event
by x00221466 26 Mar '14

26 Mar '14

Hi all: There maybe a bug about detaching virtual net device（such as virtio nic、 RTL8138、E1000）。 When live detaching the net device, there are some problems: (1)If the Guest OS don't support the hot plugging pci device, detach the virtual network device by Libvirt, the "net device" in Qemu will still exist, but "hostnet"(tap) in Qemu will be removed. so the net device in Guest OS will be of no effect. (2)If reject the nic in Guest OS, Qemu will remove the "net device", then Qemu send DEVICE_DELETED to Libvirt, Libvirt receive the event in event-loop thread and release info of the net device in qemuDomainRemoveNetDevice func. but "hostnet" in Qemu still exist. So next live attaching virtual net device will be failed because of "Duplicate ID". #virsh attach-device win2008_st_r2_64 net.xml --live error: Failed to attach device from net.xml error: internal error: unable to execute QEMU command 'netdev_add': Duplicate ID 'hostnet0' for netdev (3)In addition, in qemuDomainDetachNetDevice, detach net device func, "netdev_del" command will be sent after sending "device_del" command at once. So it is violent to remove the tap device before the net device is completely removed. So I think it's more logical that doing the work of sending Qemu command "netdev_del" after receive the DEVICE_DELETED event. It can avoid the conflict of device info between Libvirt side and Qemu side. I create a thread in qemuDomainRemoveDevice,the handle of DEVICE_DELETED event, to execute QEMU command "netdev_del". But I find there is another problem when sending monitor msg in multiply threads. in "qemuMonitorSend", monitor lock maybe unlocked and locked, so another thread has the opportunity to get the monitor lock, and use the same monitor. It will occur error. I wonder if it is workable that sending "netdev_del" command after receive the DEVICE_DELETED event. Thanks. -xie

1 0

[libvirt] [PATCH python 0/4] setup.py: Don't unconditionally run pkg-config
by Cole Robinson 25 Mar '14

25 Mar '14

This series reorganizes setup.py to not unconditionally call pkg-config. This will allow 'setup.py clean' and 'setup.py --help' to work correctly even if pkg-config or libvirt-devel aren't installed. https://bugzilla.redhat.com/show_bug.cgi?id=1074170 Cole Robinson (4): setup.py: Remove unused import setup.py: Move module list building to its own function setup.py: Make have_libvirt_lxc a function setup.py: Allow running --help or clean without pkg-config setup.py | 144 +++++++++++++++++++++++++++++++++++++-------------------------- 1 file changed, 85 insertions(+), 59 deletions(-) -- 1.8.5.3

2 6

[libvirt] [PATCH] bhyve: don't fail on busy tap devices
by Roman Bogorodskiy 25 Mar '14

25 Mar '14

We use virBhyveTapGetRealDeviceName() to map network interface name to a real device path, trying to open possible devices and getting names by ioctl. Make it skip devices that fail to open with EBUSY because they're most likely already used by other VMs. --- src/bhyve/bhyve_command.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/src/bhyve/bhyve_command.c b/src/bhyve/bhyve_command.c index 6d64c27..e569aab 100644 --- a/src/bhyve/bhyve_command.c +++ b/src/bhyve/bhyve_command.c @@ -69,6 +69,8 @@ virBhyveTapGetRealDeviceName(char *name) goto cleanup; } if ((fd = open(devpath, O_RDWR)) < 0) { + if (errno == EBUSY) + continue; virReportSystemError(errno, _("Unable to open '%s'"), devpath); goto cleanup; } -- 1.8.4.2

2 2