On 08.07.2014 13:50, Martin Kletzander wrote:
Signed-off-by: Martin Kletzander <mkletzan@redhat.com> --- docs/formatdomain.html.in | 15 ++ docs/schemas/domaincommon.rng | 17 ++ src/conf/numatune_conf.c | 183 +++++++++++++++++++-- .../qemuxml2argv-numatune-memnode-no-memory.xml | 30 ++++ .../qemuxml2argv-numatune-memnode-nocpu.xml | 25 +++ .../qemuxml2argv-numatune-memnode.xml | 31 ++++ .../qemuxml2argv-numatune-memnodes-problematic.xml | 31 ++++ tests/qemuxml2argvtest.c | 2 + .../qemuxml2xmlout-numatune-memnode.xml | 33 ++++ tests/qemuxml2xmltest.c | 2 + 10 files changed, 356 insertions(+), 13 deletions(-) create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-numatune-memnode-no-memory.xml create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-numatune-memnode-nocpu.xml create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-numatune-memnode.xml create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-numatune-memnodes-problematic.xml create mode 100644 tests/qemuxml2xmloutdata/qemuxml2xmlout-numatune-memnode.xml
diff --git a/docs/formatdomain.html.in b/docs/formatdomain.html.in index ad87b7c..d845c1b 100644 --- a/docs/formatdomain.html.in +++ b/docs/formatdomain.html.in @@ -709,6 +709,8 @@ ... <numatune> <memory mode="strict" nodeset="1-4,^3"/> + <memnode cellid="0" mode="strict" nodeset="1"/> + <memnode cellid="2" mode="preferred" nodeset="2"/> </numatune> ... </domain> @@ -745,6 +747,19 @@
<span class='since'>Since 0.9.3</span> </dd> + <dt><code>memnode</code></dt> + <dd> + Optional <code>memnode</code> elements can specify memory allocation + policies per each guest NUMA node. For those nodes having no + corresponding <code>memnode</code> element, the default from + element <code>memory</code> will be used. Attribute <code>cellid</code> + addresses guest NUMA node for which the settings are applied. + Attributes <code>mode</code> and <code>nodeset</code> have the same + meaning and syntax as in <code>memory</code> element. + + This setting is not compatible with automatic placement. + <span class='since'>QEMU Since 1.2.6</span>
1.2.8 actually
+ </dd> </dl>
+static int +virDomainNumatuneNodeParseXML(virDomainDefPtr def, + xmlXPathContextPtr ctxt) +{ + int n = 0;; + int ret = -1; + size_t i = 0; + xmlNodePtr *nodes = NULL; + + if ((n = virXPathNodeSet("./numatune/memnode", ctxt, &nodes)) < 0) { + virReportError(VIR_ERR_INTERNAL_ERROR, "%s", + _("Cannot extract memnode nodes")); + goto cleanup; + } + + if (!n) + return 0; + + if (def->numatune && def->numatune->memory.specified && + def->numatune->memory.placement == VIR_DOMAIN_NUMATUNE_PLACEMENT_AUTO) { + virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s", + _("Per-node binding is not compatible with " + "automatic NUMA placement.")); + goto cleanup; + } + + if (!def->cpu || !def->cpu->ncells) { + virReportError(VIR_ERR_XML_ERROR, "%s", + _("Element 'memnode' is invalid without " + "any guest NUMA cells")); + goto cleanup; + } + + if (!def->numatune && VIR_ALLOC(def->numatune) < 0) + goto cleanup;
Here you allow def->numatune to be allocated already.
+ + if (VIR_ALLOC_N(def->numatune->mem_nodes, def->cpu->ncells) < 0) + goto cleanup;
Which means, this can exists too. VIR_REALLOC_N() is safer IMO.
+ + def->numatune->nmem_nodes = def->cpu->ncells; + + for (i = 0; i < n; i++) { + const char *tmp = NULL;
No. s/const//.
+ int mode = 0; + unsigned int cellid = 0; + virDomainNumatuneNodePtr mem_node = NULL; + xmlNodePtr cur_node = nodes[i]; + + tmp = virXMLPropString(cur_node, "cellid"); + if (!tmp) { + virReportError(VIR_ERR_XML_ERROR, "%s", + _("Missing required cellid attribute " + "in memnode element")); + goto cleanup; + } + if (virStrToLong_uip(tmp, NULL, 10, &cellid) < 0) { + virReportError(VIR_ERR_XML_ERROR, "%s", + _("Invalid cellid attribute in memnode element"));
Moreover, @tmp is leaked here. And it would be nice to tell users in the error message @tmp somehow.
+ goto cleanup; + } + VIR_FREE(tmp); + + if (cellid >= def->numatune->nmem_nodes) { + virReportError(VIR_ERR_XML_ERROR, "%s", + _("Argument 'cellid' in memnode element must " + "correspond to existing guest's NUMA cell")); + goto cleanup; + } + + mem_node = &def->numatune->mem_nodes[cellid]; + + if (mem_node->nodeset) { + virReportError(VIR_ERR_XML_ERROR, + _("Multiple memnode elements with cellid %u"), + cellid); + goto cleanup; + } + + tmp = virXMLPropString(cur_node, "mode"); + if (!tmp) { + mem_node->mode = VIR_DOMAIN_NUMATUNE_MEM_STRICT; + } else { + if ((mode = virDomainNumatuneMemModeTypeFromString(tmp)) < 0) { + virReportError(VIR_ERR_XML_ERROR, "%s", + _("Invalid mode attribute in memnode element")); + goto cleanup; + } + VIR_FREE(tmp); + mem_node->mode = mode; + } + + tmp = virXMLPropString(cur_node, "nodeset"); + if (!tmp) { + virReportError(VIR_ERR_XML_ERROR, "%s", + _("Missing required nodeset attribute " + "in memnode element")); + goto cleanup; + } + if (virBitmapParse(tmp, 0, &mem_node->nodeset, + VIR_DOMAIN_CPUMASK_LEN) < 0) + goto cleanup; + VIR_FREE(tmp); + } + + ret = 0; + cleanup: + VIR_FREE(nodes); + return ret; +} +
diff --git a/tests/qemuxml2xmloutdata/qemuxml2xmlout-numatune-memnode.xml b/tests/qemuxml2xmloutdata/qemuxml2xmlout-numatune-memnode.xml new file mode 100644 index 0000000..82b5f61 --- /dev/null +++ b/tests/qemuxml2xmloutdata/qemuxml2xmlout-numatune-memnode.xml
Are you sure this is the correct content?
@@ -0,0 +1,33 @@ +<domain type='qemu'> + <name>QEMUGuest</name> + <uuid>9f4b6512-e73a-4a25-93e8-5307802821ce</uuid> + <memory unit='KiB'>24682468</memory>
In the qemuxml2argvdata/qemuxml2argv-numatune-memnode.xml you have 65MB of RAM while here ~23.5GB.
+ <currentMemory unit='KiB'>24682468</currentMemory> + <vcpu placement='static'>32</vcpu>
and only 2 vcpus, while here is 32.
+ <numatune> + <memory mode='strict' nodeset='0-7'/> + <memnode cellid='0' mode='preferred' nodeset='3'/> + <memnode cellid='2' mode='strict' nodeset='1-2,5,7'/> + </numatune> + <os> + <type arch='x86_64' machine='pc'>hvm</type> + <boot dev='hd'/> + </os> + <cpu> + <numa> + <cell id='0' cpus='0' memory='20002'/> + <cell id='1' cpus='1-27,29' memory='660066'/> + <cell id='2' cpus='28-31,^29' memory='24002400'/>
And this is broken too.
+ </numa> + </cpu> + <clock offset='utc'/> + <on_poweroff>destroy</on_poweroff> + <on_reboot>restart</on_reboot> + <on_crash>destroy</on_crash> + <devices> + <emulator>/usr/bin/kvm</emulator> + <controller type='usb' index='0'/> + <controller type='pci' index='0' model='pci-root'/> + <memballoon model='virtio'/> + </devices> +</domain> diff --git a/tests/qemuxml2xmltest.c b/tests/qemuxml2xmltest.c index 3bba565..4beb799 100644 --- a/tests/qemuxml2xmltest.c +++ b/tests/qemuxml2xmltest.c @@ -372,6 +372,8 @@ mymain(void) DO_TEST_DIFFERENT("cpu-numa2");
DO_TEST_DIFFERENT("numatune-auto-prefer"); + DO_TEST_DIFFERENT("numatune-memnode"); + DO_TEST("numatune-memnode-no-memory");
virObjectUnref(driver.caps); virObjectUnref(driver.xmlopt);
I can't ACK this one, until the issue is resolved. If I squash this in, the test passes again: diff --git a/tests/qemuxml2xmloutdata/qemuxml2xmlout-numatune-memnode.xml b/tests/qemuxml2xmloutdata/qemuxml2xmlout-numatune-memnode.xml index 82b5f61..18b00d8 100644 --- a/tests/qemuxml2xmloutdata/qemuxml2xmlout-numatune-memnode.xml +++ b/tests/qemuxml2xmloutdata/qemuxml2xmlout-numatune-memnode.xml @@ -1,13 +1,12 @@ <domain type='qemu'> <name>QEMUGuest</name> <uuid>9f4b6512-e73a-4a25-93e8-5307802821ce</uuid> - <memory unit='KiB'>24682468</memory> - <currentMemory unit='KiB'>24682468</currentMemory> - <vcpu placement='static'>32</vcpu> + <memory unit='KiB'>65536</memory> + <currentMemory unit='KiB'>65536</currentMemory> + <vcpu placement='static'>2</vcpu> <numatune> - <memory mode='strict' nodeset='0-7'/> + <memory mode='strict' nodeset='0-3'/> <memnode cellid='0' mode='preferred' nodeset='3'/> - <memnode cellid='2' mode='strict' nodeset='1-2,5,7'/> </numatune> <os> <type arch='x86_64' machine='pc'>hvm</type> @@ -15,9 +14,8 @@ </os> <cpu> <numa> - <cell id='0' cpus='0' memory='20002'/> - <cell id='1' cpus='1-27,29' memory='660066'/> - <cell id='2' cpus='28-31,^29' memory='24002400'/> + <cell id='0' cpus='0' memory='32768'/> + <cell id='1' cpus='1' memory='32768'/> </numa> </cpu> <clock offset='utc'/> Michal