October 2015 - Devel - libvirt List Archives

[libvirt] [PATCH] tests: Fix virnetdaemontest to be skipped in non-yajl environment
by Prerna 27 Oct '15

27 Oct '15

When libvirt is compiled without yajl-devel, virnetdaemontest fails. This causes 'make check' to fail with nondescript errors, such as following: [snip]............................... PASS: read-bufsiz PASS: read-non-seekable PASS: start TEST: virsh-uriprecedence .... 4 OK PASS: virsh-uriprecedence PASS: vcpupin /home/prerna/trees/libvirt/tests/virsh-all: skipping test: This test is very expensive, so it is disabled by default. To run it anyway, rerun: make check VIR_TEST_EXPENSIVE=1 SKIP: virsh-all PASS: virsh-optparse PASS: virsh-schedinfo PASS: virsh-synopsis PASS: virsh-undefine ======================================= 1 of 108 tests failed (7 tests were not run) Please report to libvir-list(a)redhat.com ======================================= make[2]: *** [check-TESTS] Error 1 make[2]: Leaving directory `/home/prerna/build/libvirt/tests' make[1]: *** [check-am] Error 2 make[1]: Leaving directory `/home/prerna/build/libvirt/tests' make: *** [check-recursive] Error 1 ..................................... This patch skips virnetdaemontest in absence of yajl-devel, so that make check gracefully runs to completion. Signed-off-by: Prerna Saxena <saxenap.ltc(a)gmail.com> --- tests/virnetdaemontest.c | 5 +++++ 1 files changed, 5 insertions(+), 0 deletions(-) diff --git a/tests/virnetdaemontest.c b/tests/virnetdaemontest.c index 24cbd54..17d1f7f 100644 --- a/tests/virnetdaemontest.c +++ b/tests/virnetdaemontest.c @@ -290,6 +290,11 @@ mymain(void) { int ret = 0; +#if !WITH_YAJL + fputs("libvirt not compiled with yajl, skipping this test\n", stderr); + return EXIT_AM_SKIP; +#endif + if (virInitialize() < 0 || virEventRegisterDefaultImpl() < 0) { virDispatchError(NULL); -- 1.7.1

2 3

[libvirt] [PATCH 0/4] qemu: hostdev: Unify naming
by Andrea Bolognani 26 Oct '15

26 Oct '15

I was working on the QEMU hostdev code and the inconsistent naming used kept making things just that little harder for me to follow, so I unified it. Cheers. Andrea Bolognani (4): qemu: hostdev: Unify naming for qemuPrepareHost*Devices() qemu: hostdev: Unify naming for qemuDomainReAttachHost*Devices() qemu: hostdev: Unify naming for qemuUpdateActiveHost*Devices() qemu: hostdev: Introduce qemuUpdateActiveHostDevices() src/qemu/qemu_hostdev.c | 74 ++++++++++++++++++++++++++++++------------------- src/qemu/qemu_hostdev.h | 36 ++++++++++++------------ src/qemu/qemu_hotplug.c | 13 ++++----- src/qemu/qemu_process.c | 8 +----- 4 files changed, 72 insertions(+), 59 deletions(-) -- 2.4.3

3 8

[libvirt] [PATCH v2 0/2] tests: Add nodeinfo test data utility scripts
by Andrea Bolognani 23 Oct '15

23 Oct '15

Changes from v1: * use a better approach to copy host data (thanks Martin) * acknowledge the fact that the scripts require bash * other small cleanups and improvements Andrea Bolognani (2): tests: Add script to display nodeinfo test data tests: Add script to copy nodeinfo test data from host tests/nodeinfodata/copy-from-host.sh | 113 +++++++++++++++++++++++++++++++++++ tests/nodeinfodata/display.sh | 113 +++++++++++++++++++++++++++++++++++ 2 files changed, 226 insertions(+) create mode 100755 tests/nodeinfodata/copy-from-host.sh create mode 100755 tests/nodeinfodata/display.sh -- 2.4.3

3 7

[libvirt] outreachy 2015
by mesode akwe 22 Oct '15

22 Oct '15

hello! i am mesode from cameroon and am interested in participating in the libvirt-designer enhancement project. where do i go from here? thanks.

1 0

[libvirt] [RFC] Simplifying usage of {Enter, Exit}Monitor and {Begin/End}Job
by Jiri Denemark 22 Oct '15

22 Oct '15

Hi all. Looking at the qemu driver code, we have a lot of code with the following pattern: if (doSomething() < 0) goto cleanup; if (qemuDomainObjBeginJob() < 0) goto cleanup; if (doSomethingElse() < 0) goto endjob; qemuDomainObjEnterMonitor(); if (qemuMonitorSomething() < 0) { qemuDomainObjExitMonitor(); goto endjob; } if (qemuMonitorSomethingElse() < 0) { qemuDomainObjExitMonitor(); goto endjob; } if (qemuDomainObjExitMonitor() < 0) goto endjob; ... ret = 0; endjob: qemuDomainObjEndJob(); cleanup: ... return ret; Sometimes qemuDomainObjExitMonitor is in its own label instead of being explicitly called on every single error path. Sometimes qemuDomainObjEndJob is called explicitly followed by goto cleanup. In either case, it's pretty easy to get this wrong. It's too easy to jump to a wrong label (especially when moving code around) or forget to call the appropriate exit function before jumping to a label. So what if we make the code more robust by changing who needs to track whether we are in a monitor or have a job. Now it's the caller that tracks it while I think we could teach the job/monitor APIs themselves to track the state: bool inJob = false; bool inMonitor = false; if (doSomething() < 0) goto cleanup; if (qemuDomainObjBeginJob(..., &inJob) < 0) goto cleanup; if (doSomethingElse() < 0) goto cleanup; qemuDomainObjEnterMonitor(..., &inMonitor); if (qemuMonitorSomething() < 0) goto cleanup; if (qemuMonitorSomethingElse() < 0) goto cleanup; if (qemuDomainObjExitMonitor(..., &inMonitor) < 0) goto cleanup; ... ret = 0; cleanup: if (qemuDomainObjExitMonitor(..., &inMonitor) < 0) ret = -1; qemuDomainObjEndJob(..., &inJob); ... return ret; It's not a lot shorter or longer but it saves us from jumping to different labels and it makes sure we always exit the monitor and end the job. So is it worth the effort or do you thing it's even worse than before or do you have any other ideas? Thanks, Jirka

3 4

[libvirt] [PATCHv2] virsh: fix no error when pass a count <= 0 to setvcpus
by Luyao Huang 22 Oct '15

22 Oct '15

https://bugzilla.redhat.com/show_bug.cgi?id=1248277 When count <= 0, the client exit without set an error. Signed-off-by: Luyao Huang <lhuang(a)redhat.com> --- v2: - use vshCommandOptUInt to forbid negative number, and check if count is zero. (thanks Peter and Andrea) tools/virsh-domain.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/tools/virsh-domain.c b/tools/virsh-domain.c index 4191548..3f032f4 100644 --- a/tools/virsh-domain.c +++ b/tools/virsh-domain.c @@ -6873,7 +6873,7 @@ static bool cmdSetvcpus(vshControl *ctl, const vshCmd *cmd) { virDomainPtr dom; - int count = 0; + unsigned int count = 0; bool ret = false; bool maximum = vshCommandOptBool(cmd, "maximum"); bool config = vshCommandOptBool(cmd, "config"); @@ -6900,9 +6900,14 @@ cmdSetvcpus(vshControl *ctl, const vshCmd *cmd) if (!(dom = virshCommandOptDomain(ctl, cmd, NULL))) return false; - if (vshCommandOptInt(ctl, cmd, "count", &count) < 0 || count <= 0) + if (vshCommandOptUInt(ctl, cmd, "count", &count) < 0) goto cleanup; + if (count == 0) { + vshError(ctl, _("Can't set 0 processors for a VM")); + goto cleanup; + } + /* none of the options were specified */ if (!current && flags == 0) { if (virDomainSetVcpus(dom, count) != 0) -- 1.8.3.1

3 2

[libvirt] [PATCH] tests: Remove unused nodeinfo test data
by Andrea Bolognani 22 Oct '15

22 Oct '15

A bunch of files that we don't currently parse, and are very unlikely to ever start parsing, made their way into the nodeinfo test data. Get rid of them. --- tests/nodeinfodata/linux-f21-mustang/cpu/modalias | 1 - tests/nodeinfodata/linux-raspberrypi/cpu/cpuidle/current_driver | 1 - tests/nodeinfodata/linux-raspberrypi/cpu/cpuidle/current_governor_ro | 1 - tests/nodeinfodata/linux-rhelsa-3.19.0-mustang/cpu/modalias | 1 - tests/nodeinfodata/linux-test8/cpu/cpuidle/current_driver | 1 - tests/nodeinfodata/linux-test8/cpu/cpuidle/current_governor_ro | 1 - tests/nodeinfodata/linux-test8/cpu/sched_mc_power_savings | 1 - tests/nodeinfodata/linux-test8/cpu/sched_smt_power_savings | 1 - 8 files changed, 8 deletions(-) delete mode 100644 tests/nodeinfodata/linux-f21-mustang/cpu/modalias delete mode 100644 tests/nodeinfodata/linux-raspberrypi/cpu/cpuidle/current_driver delete mode 100644 tests/nodeinfodata/linux-raspberrypi/cpu/cpuidle/current_governor_ro delete mode 100644 tests/nodeinfodata/linux-rhelsa-3.19.0-mustang/cpu/modalias delete mode 100644 tests/nodeinfodata/linux-test8/cpu/cpuidle/current_driver delete mode 100644 tests/nodeinfodata/linux-test8/cpu/cpuidle/current_governor_ro delete mode 100644 tests/nodeinfodata/linux-test8/cpu/sched_mc_power_savings delete mode 100644 tests/nodeinfodata/linux-test8/cpu/sched_smt_power_savings diff --git a/tests/nodeinfodata/linux-f21-mustang/cpu/modalias b/tests/nodeinfodata/linux-f21-mustang/cpu/modalias deleted file mode 100644 index 18bb0b6..0000000 --- a/tests/nodeinfodata/linux-f21-mustang/cpu/modalias +++ /dev/null @@ -1 +0,0 @@ -cpu:type:aarch64:feature:,0000,0001,0002 diff --git a/tests/nodeinfodata/linux-raspberrypi/cpu/cpuidle/current_driver b/tests/nodeinfodata/linux-raspberrypi/cpu/cpuidle/current_driver deleted file mode 100644 index 621e94f..0000000 --- a/tests/nodeinfodata/linux-raspberrypi/cpu/cpuidle/current_driver +++ /dev/null @@ -1 +0,0 @@ -none diff --git a/tests/nodeinfodata/linux-raspberrypi/cpu/cpuidle/current_governor_ro b/tests/nodeinfodata/linux-raspberrypi/cpu/cpuidle/current_governor_ro deleted file mode 100644 index c35a724..0000000 --- a/tests/nodeinfodata/linux-raspberrypi/cpu/cpuidle/current_governor_ro +++ /dev/null @@ -1 +0,0 @@ -menu diff --git a/tests/nodeinfodata/linux-rhelsa-3.19.0-mustang/cpu/modalias b/tests/nodeinfodata/linux-rhelsa-3.19.0-mustang/cpu/modalias deleted file mode 100644 index 18bb0b6..0000000 --- a/tests/nodeinfodata/linux-rhelsa-3.19.0-mustang/cpu/modalias +++ /dev/null @@ -1 +0,0 @@ -cpu:type:aarch64:feature:,0000,0001,0002 diff --git a/tests/nodeinfodata/linux-test8/cpu/cpuidle/current_driver b/tests/nodeinfodata/linux-test8/cpu/cpuidle/current_driver deleted file mode 100644 index 57d5dd3..0000000 --- a/tests/nodeinfodata/linux-test8/cpu/cpuidle/current_driver +++ /dev/null @@ -1 +0,0 @@ -acpi_idle diff --git a/tests/nodeinfodata/linux-test8/cpu/cpuidle/current_governor_ro b/tests/nodeinfodata/linux-test8/cpu/cpuidle/current_governor_ro deleted file mode 100644 index c35a724..0000000 --- a/tests/nodeinfodata/linux-test8/cpu/cpuidle/current_governor_ro +++ /dev/null @@ -1 +0,0 @@ -menu diff --git a/tests/nodeinfodata/linux-test8/cpu/sched_mc_power_savings b/tests/nodeinfodata/linux-test8/cpu/sched_mc_power_savings deleted file mode 100644 index 573541a..0000000 --- a/tests/nodeinfodata/linux-test8/cpu/sched_mc_power_savings +++ /dev/null @@ -1 +0,0 @@ -0 diff --git a/tests/nodeinfodata/linux-test8/cpu/sched_smt_power_savings b/tests/nodeinfodata/linux-test8/cpu/sched_smt_power_savings deleted file mode 100644 index 573541a..0000000 --- a/tests/nodeinfodata/linux-test8/cpu/sched_smt_power_savings +++ /dev/null @@ -1 +0,0 @@ -0 -- 2.4.3

2 2

[libvirt] [PATCHv2 0/3] improve the error report for virNodeAllocPages and virNodeGetFreePages
by Luyao Huang 21 Oct '15

21 Oct '15

v1: https://www.redhat.com/archives/libvir-list/2015-September/msg01054.html v2: -improve the error message -improve the code struct -add a new patch for rework the exist way to check node and page size Luyao Huang (3): util: split the virNumaGetHugePageInfoPath into separate function util: move the pagesize and node check and error report to one place util: Produce friendlier error message to user src/util/virnuma.c | 113 ++++++++++++++++++++++++++++++++++------------------- 1 file changed, 72 insertions(+), 41 deletions(-) -- 1.8.3.1

3 5

[libvirt] [PATCH] conf: Fix error message to use correct parameter
by John Ferlan 21 Oct '15

21 Oct '15

Fix a cut-n-paste error from commit id '35eecdde' where the previous check for max_sectors seems to have been copied, but the error message parameter not updated to be ioeventfd Signed-off-by: John Ferlan <jferlan(a)redhat.com> --- pushing as trivial src/conf/domain_conf.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c index 02cc8ad..0c559d2 100644 --- a/src/conf/domain_conf.c +++ b/src/conf/domain_conf.c @@ -7985,7 +7985,7 @@ virDomainControllerDefParseXML(xmlNodePtr node, if (ioeventfd && (def->ioeventfd = virTristateSwitchTypeFromString(ioeventfd)) < 0) { virReportError(VIR_ERR_XML_ERROR, - _("Malformed 'ioeventfd' value %s'"), max_sectors); + _("Malformed 'ioeventfd' value %s'"), ioeventfd); goto error; } -- 2.1.0

1 0

[libvirt] gem install ruby-libvirt fails on FreeBSD 10.2
by Rickard von Essen 21 Oct '15

21 Oct '15

Hi, Installing the ruby-libvirt gem fails on FreeBSD 10.2 since it can't locate the lib and include dir. Installing with: gem install ruby-libvirt -- --with-libvirt-include=/usr/local/include/libvirt --with-libvirt-lib=/usr/local/lib/libvirt.so works fine. It would be great if this worked out of the box, e.g. detects that it is building on BSD and applies the above settings as default, or at least it could be provided as a hint if running make fails to guide ruby noobs, like me. Regards, Rickard von Essen

2 2