January 2017 - Devel - libvirt List Archives

[libvirt] [PATCH] THREADS.txt: fix typos
by Chen Hanxiao 25 Jan '17

25 Jan '17

From: Chen Hanxiao <chenhanxiao(a)gmail.com> s/wakup/wakeup Signed-off-by: Chen Hanxiao <chenhanxiao(a)gmail.com> --- daemon/THREADS.txt | 2 +- src/util/vireventpoll.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/daemon/THREADS.txt b/daemon/THREADS.txt index 762ca25..ae4e9ba 100644 --- a/daemon/THREADS.txt +++ b/daemon/THREADS.txt @@ -42,7 +42,7 @@ event loop thread handles I/O from the client socket, and once a complete RPC message has been read off the wire (and optionally decrypted), it will be placed on the 'dx' job queue for the associated client object. The job condition will be signalled and -a worker will wakup and process it. +a worker will wakeup and process it. The worker thread must quickly drop its locks on the server and client to allow the main event loop thread to continue running diff --git a/src/util/vireventpoll.h b/src/util/vireventpoll.h index 8844c9c..e78e273 100644 --- a/src/util/vireventpoll.h +++ b/src/util/vireventpoll.h @@ -124,7 +124,7 @@ int virEventPollToNativeEvents(int events); /** * virEventPollInterrupt: wakeup any thread waiting in poll() * - * return -1 if wakup failed + * return -1 if wakeup failed */ int virEventPollInterrupt(void); -- 2.7.4

2 1

[libvirt] [PATCH 0/4] Xen: Improve support for <timer> configurations
by Jim Fehlig 24 Jan '17

24 Jan '17

This series fixes some issues wrt <timer> configuration in the libxl driver and the xenconfig parser/formatter. See patches 1 and 2 for details on the libxl driver changes, and patch 3 for details on the xenconfig parser/formatter changes. Patch 4 adds some tests to check domXML <-> xl.cfg conversions of <timer> configuration. Jim Fehlig (4): libxl: fix timer configuration libxl: support emulate mode of tsc timer xenconfig: add support for more timers tests: add xlconfig tests for <timer> configurations src/libxl/libxl_conf.c | 49 +++++++++---- src/xenconfig/xen_common.c | 87 +++++++++++++++++++++--- tests/xlconfigdata/test-fullvirt-hpet-timer.cfg | 27 ++++++++ tests/xlconfigdata/test-fullvirt-hpet-timer.xml | 64 +++++++++++++++++ tests/xlconfigdata/test-fullvirt-multi-timer.cfg | 28 ++++++++ tests/xlconfigdata/test-fullvirt-multi-timer.xml | 65 ++++++++++++++++++ tests/xlconfigdata/test-fullvirt-tsc-timer.cfg | 27 ++++++++ tests/xlconfigdata/test-fullvirt-tsc-timer.xml | 64 +++++++++++++++++ tests/xlconfigtest.c | 3 + 9 files changed, 392 insertions(+), 22 deletions(-) create mode 100644 tests/xlconfigdata/test-fullvirt-hpet-timer.cfg create mode 100644 tests/xlconfigdata/test-fullvirt-hpet-timer.xml create mode 100644 tests/xlconfigdata/test-fullvirt-multi-timer.cfg create mode 100644 tests/xlconfigdata/test-fullvirt-multi-timer.xml create mode 100644 tests/xlconfigdata/test-fullvirt-tsc-timer.cfg create mode 100644 tests/xlconfigdata/test-fullvirt-tsc-timer.xml -- 2.9.2

2 11

[libvirt] [PATCH 0/3] support setting MTU of libvirt networks
by Laine Stump 24 Jan '17

24 Jan '17

This has been requested several times in #virt on IRC, and on libvirt-users, and a comment posted to https://bugzilla.redhat.com/1224348 reminded me that there's even an open BZ for it, so I finally sat down and wrote the patches for it. The idea of the patches is that an *empty* bridge device can't have its MTU directly set to anything higher than 1500, but once there are any devices on the bridge, its MTU is limited to the smallest of the MTUs of any attached devices. However, when the very first device is attached to the bridge, the bridge will *increase* its own MTU to match if needed. So the trick to setting a high MTU is to set the MTU of every device to the desired value before it is added (especially the first). libvirt already sets the MTU of every device to the MTU of the bridge, so all we need to do is make it so the first device is set from the network config's "mtu" attribute instead - the bridge will get the larger MTU, and all future devices added will get the higher MTU as well. This doesn't take care of setting the MTU on the guest side of the tap device; for now that must be done manually, but qemu has recently added a commandline parameter to send the desired MTU to the guest, and the virtio-net guest driver at least honors this. An upcoming patch to libvirt will automatically set this qemu commandline option so that no intervention will be required (beyond setting the MTU once in the network config). Laine Stump (3): util: add MTU arg to virNetDevTapCreateInBridgePort() conf: support mtu attribute in a network's <bridge> element network: honor mtu setting when creating network docs/formatnetwork.html.in | 8 +++++++- docs/news.xml | 6 +++++- docs/schemas/network.rng | 6 ++++++ src/bhyve/bhyve_command.c | 2 +- src/conf/network_conf.c | 20 ++++++++++++++++++-- src/conf/network_conf.h | 1 + src/network/bridge_driver.c | 1 + src/qemu/qemu_interface.c | 2 +- src/uml/uml_conf.c | 2 +- src/util/virnetdevtap.c | 20 +++++++++++++++----- src/util/virnetdevtap.h | 1 + tests/bhyvexml2argvmock.c | 1 + tests/networkxml2xmlin/set-mtu.xml | 11 +++++++++++ tests/networkxml2xmlout/set-mtu.xml | 11 +++++++++++ tests/networkxml2xmltest.c | 1 + 15 files changed, 81 insertions(+), 12 deletions(-) create mode 100644 tests/networkxml2xmlin/set-mtu.xml create mode 100644 tests/networkxml2xmlout/set-mtu.xml -- 2.9.3

2 7

[libvirt] [PATCH v2 1/2] util: virhostdev: disallow assigning a pci-bridge to a guest
by Shivaprasad G Bhat 23 Jan '17

23 Jan '17

Non-endpoint devices like pci-bridges cannot be passedthrough to guests. Prevent such attempts. Signed-off-by: Shivaprasad G Bhat <sbhat(a)linux.vnet.ibm.com> --- src/util/virhostdev.c | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/src/util/virhostdev.c b/src/util/virhostdev.c index 0673afb..b23fe1f 100644 --- a/src/util/virhostdev.c +++ b/src/util/virhostdev.c @@ -532,6 +532,16 @@ virHostdevPreparePCIDevices(virHostdevManagerPtr mgr, bool strict_acs_check = !!(flags & VIR_HOSTDEV_STRICT_ACS_CHECK); bool usesVFIO = (virPCIDeviceGetStubDriver(pci) == VIR_PCI_STUB_DRIVER_VFIO); struct virHostdevIsPCINodeDeviceUsedData data = { mgr, dom_name, usesVFIO }; + int hdrType = -1; + + if (virPCIGetHeaderType(pci, &hdrType) < 0) + goto cleanup; + + if (hdrType != VIR_PCI_HEADER_ENDPOINT) { + virReportError(VIR_ERR_INTERNAL_ERROR, "%s", _("PCI bridge devices " + "cannot be assigned to guests")); + goto cleanup; + } if (!usesVFIO && !virPCIDeviceIsAssignable(pci, strict_acs_check)) { virReportError(VIR_ERR_OPERATION_INVALID,

3 4

[libvirt] [PATCH go] Fix segfault in Connect.GetMemoryStats
by Matvey Kruglov 23 Jan '17

23 Jan '17

*** BLURB HERE *** Matvey Kruglov (1): Fix segfault in Connect.GetMemoryStats connect.go | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) -- 2.7.4

2 2

[libvirt] [PATCH v2] util: disallow assigning or resetting a pci-bridge device
by Shivaprasad G Bhat 23 Jan '17

23 Jan '17

It is destructive to attempt reset on a pci-bridge, the host can crash. The bridges won't contain any guest data and neither they can be passed through using vfio/stub. So, no point in allowing a reset on them. Signed-off-by: Shivaprasad G Bhat <sbhat(a)linux.vnet.ibm.com> --- src/util/virhostdev.c | 10 ++++++++++ src/util/virpci.c | 11 +++++++++++ 2 files changed, 21 insertions(+) diff --git a/src/util/virhostdev.c b/src/util/virhostdev.c index 0673afb..16b96f3 100644 --- a/src/util/virhostdev.c +++ b/src/util/virhostdev.c @@ -532,6 +532,16 @@ virHostdevPreparePCIDevices(virHostdevManagerPtr mgr, bool strict_acs_check = !!(flags & VIR_HOSTDEV_STRICT_ACS_CHECK); bool usesVFIO = (virPCIDeviceGetStubDriver(pci) == VIR_PCI_STUB_DRIVER_VFIO); struct virHostdevIsPCINodeDeviceUsedData data = { mgr, dom_name, usesVFIO }; + int hdrType = -1; + + if (virPCIGetHeaderType(pci, &hdrType) < 0) + goto cleanup; + + if (hdrType == VIR_PCI_HEADER_PCI_BRIDGE) { + virReportError(VIR_ERR_INTERNAL_ERROR, "%s", _("PCI bridge devices" + " cannot be assigned to guests")); + goto cleanup; + } if (!usesVFIO && !virPCIDeviceIsAssignable(pci, strict_acs_check)) { virReportError(VIR_ERR_OPERATION_INVALID, diff --git a/src/util/virpci.c b/src/util/virpci.c index 0601f49..f205abf 100644 --- a/src/util/virpci.c +++ b/src/util/virpci.c @@ -933,6 +933,17 @@ virPCIDeviceReset(virPCIDevicePtr dev, char *drvName = NULL; int ret = -1; int fd = -1; + int hdrType = -1; + + if (virPCIGetHeaderType(dev, &hdrType) < 0) + return -1; + + if (hdrType != VIR_PCI_HEADER_ENDPOINT) { + virReportError(VIR_ERR_INTERNAL_ERROR, + _("Invalid attempt to reset non-endpoint PCI device %s." + " Only PCI endpoint devices can be reset"), dev->name); + return -1; + } if (activeDevs && virPCIDeviceListFind(activeDevs, dev)) { virReportError(VIR_ERR_INTERNAL_ERROR,

3 3

[libvirt] [PATCH resend v2 0/2] Allow saving QEMU libvirt state to a pipe
by Chen Hanxiao 22 Jan '17

22 Jan '17

This series introduce flag VIR_DOMAIN_SAVE_DIRECT to enable command 'save' to write to PIPE. Base upon patches from Roy Keene <rkeene(a)knightpoint.com> with some fixes. Change from original patch: 1) Check whether the specified path is a PIPE. 2) Rebase on upstream. 3) Add doc for virsh command v2-resend: rebase on upstream v2: rename VIR_DOMAIN_SAVE_PIPE to VIR_DOMAIN_SAVE_DIRECT remove S_ISFIFO check Chen Hanxiao (2): qemu: Allow saving QEMU libvirt state to a pipe virsh: introduce flage --direct for save command include/libvirt/libvirt-domain.h | 1 + src/qemu/qemu_driver.c | 54 ++++++++++++++++++++++++++-------------- tools/virsh-domain.c | 6 +++++ tools/virsh.pod | 5 +++- 4 files changed, 47 insertions(+), 19 deletions(-) -- 2.7.4

1 3

[libvirt] [PATCH v2] qemu_domain: add timestamp in tainting of guests log
by Chen Hanxiao 21 Jan '17

21 Jan '17

From: Chen Hanxiao <chenhanxiao(a)gmail.com> We lacked of timestamp in tainting of guests log, which bring troubles for finding guest issues: such as whether a guest powerdown caused by qemu-monitor-command or others issues inside guests. If we had timestamp in tainting of guests log, it would be helpful when checking guest's /var/log/messages. Signed-off-by: Chen Hanxiao <chenhanxiao(a)gmail.com> --- v2: update logic flow of virDomainObjTaint check src/qemu/qemu_domain.c | 79 +++++++++++++++++++++++++++----------------------- 1 file changed, 42 insertions(+), 37 deletions(-) diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c index c676042..2bc023c 100644 --- a/src/qemu/qemu_domain.c +++ b/src/qemu/qemu_domain.c @@ -3999,46 +3999,51 @@ void qemuDomainObjTaint(virQEMUDriverPtr driver, { virErrorPtr orig_err = NULL; bool closeLog = false; + char *timestamp = NULL; + char uuidstr[VIR_UUID_STRING_BUFLEN]; - if (virDomainObjTaint(obj, taint)) { - char uuidstr[VIR_UUID_STRING_BUFLEN]; - virUUIDFormat(obj->def->uuid, uuidstr); - - VIR_WARN("Domain id=%d name='%s' uuid=%s is tainted: %s", - obj->def->id, - obj->def->name, - uuidstr, - virDomainTaintTypeToString(taint)); - - /* We don't care about errors logging taint info, so - * preserve original error, and clear any error that - * is raised */ - orig_err = virSaveLastError(); - if (logCtxt == NULL) { - logCtxt = qemuDomainLogContextNew(driver, obj, - QEMU_DOMAIN_LOG_CONTEXT_MODE_ATTACH); - if (!logCtxt) { - if (orig_err) { - virSetError(orig_err); - virFreeError(orig_err); - } - VIR_WARN("Unable to open domainlog"); - return; - } - closeLog = true; - } + if (!virDomainObjTaint(obj, taint)) + return; + + virUUIDFormat(obj->def->uuid, uuidstr); + + VIR_WARN("Domain id=%d name='%s' uuid=%s is tainted: %s", + obj->def->id, + obj->def->name, + uuidstr, + virDomainTaintTypeToString(taint)); - if (qemuDomainLogContextWrite(logCtxt, - "Domain id=%d is tainted: %s\n", - obj->def->id, - virDomainTaintTypeToString(taint)) < 0) - virResetLastError(); - if (closeLog) - qemuDomainLogContextFree(logCtxt); - if (orig_err) { - virSetError(orig_err); - virFreeError(orig_err); + if ((timestamp = virTimeStringNow()) == NULL) + goto cleanup; + + /* We don't care about errors logging taint info, so + * preserve original error, and clear any error that + * is raised */ + orig_err = virSaveLastError(); + if (logCtxt == NULL) { + logCtxt = qemuDomainLogContextNew(driver, obj, + QEMU_DOMAIN_LOG_CONTEXT_MODE_ATTACH); + if (!logCtxt) { + VIR_WARN("Unable to open domainlog"); + goto cleanup; } + closeLog = true; + } + + if (qemuDomainLogContextWrite(logCtxt, + "%s: Domain id=%d is tainted: %s\n", + timestamp, + obj->def->id, + virDomainTaintTypeToString(taint)) < 0) + virResetLastError(); + + cleanup: + VIR_FREE(timestamp); + if (closeLog) + qemuDomainLogContextFree(logCtxt); + if (orig_err) { + virSetError(orig_err); + virFreeError(orig_err); } } -- 2.7.4

2 1

[libvirt] [PATCH 0/9] i386: query-cpu-model-expansion test script
by Eduardo Habkost 20 Jan '17

20 Jan '17

This is a follow-up to the series that implements query-cpu-model-expansion. Before including the test script, the series has some fixes to allow the results of query-cpu-model-expansion to be used in the QEMU command-line. The script probably will work on s390x too, but I couldn't test it yet. This series and its dependencies can be pulled from the branch: https://github.com/ehabkost/qemu-hacks.git work/x86-query-cpu-expansion-test --- Cc: Cornelia Huck <cornelia.huck(a)de.ibm.com> Cc: Christian Borntraeger <borntraeger(a)de.ibm.com> Cc: David Hildenbrand <david(a)redhat.com> Cc: libvir-list(a)redhat.com Cc: Jiri Denemark <jdenemar(a)redhat.com> Cc: "Jason J. Herne" <jjherne(a)linux.vnet.ibm.com> Cc: Markus Armbruster <armbru(a)redhat.com> Cc: Richard Henderson <rth(a)twiddle.net> Cc: Igor Mammedov <imammedo(a)redhat.com> Cc: Eric Blake <eblake(a)redhat.com> Eduardo Habkost (9): target-i386: Move "host" properties to base class target-i386: Allow short strings to be used as vendor ID cpu: Support comma escaping when parsing -cpu qemu.py: Make logging optional qtest.py: Support QTEST_LOG environment variable qtest.py: make logging optional qtest.py: Make 'binary' parameter optional tests: Add rules to non-gtester qtest test cases tests: Test case for query-cpu-model-expansion scripts/qemu.py | 25 ++- scripts/qtest.py | 15 +- qom/cpu.c | 32 ++-- target/i386/cpu.c | 83 ++++----- tests/test-x86-cpuid-compat.c | 19 ++ tests/Makefile.include | 40 ++++- tests/query-cpu-model-test.py | 398 ++++++++++++++++++++++++++++++++++++++++++ 7 files changed, 551 insertions(+), 61 deletions(-) create mode 100755 tests/query-cpu-model-test.py -- 2.11.0.259.g40922b1

4 10

[libvirt] [PATCH python] Protect against user accidentally calling constructors directly
by Daniel P. Berrange 20 Jan '17

20 Jan '17

When using libvirt python you must never call the object constructors directly, as these are expecting to be passed a wrapped C object. For example import libvirt c = libvirt.virConnect("qemu:///system") c.listAllDomains() will mysteriously segfault. With this change the user now gets an slightly more helpful error Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/home/berrange/src/virt/libvirt-python/build/libvirt.py", line 3409, in __init__ raise Exception("Expected a wrapped C Object but got %s" % type(_obj)) Exception: Expected a wrapped C Object but got <type 'str'> Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com> --- generator.py | 2 ++ 1 file changed, 2 insertions(+) diff --git a/generator.py b/generator.py index e9be8b1..6bcf80b 100755 --- a/generator.py +++ b/generator.py @@ -1532,6 +1532,8 @@ def buildWrappers(module): elif classname in [ "virDomainSnapshot" ]: classes.write(" self._dom = dom\n") classes.write(" self._conn = dom.connect()\n") + classes.write(" if type(_obj).__name__ not in [\"PyCapsule\", \"PyCObject\"]:\n") + classes.write(" raise Exception(\"Expected a wrapped C Object but got %s\" % type(_obj))\n") classes.write(" self._o = _obj\n\n") destruct=None if classname in classes_destructors: -- 2.9.3

2 1