11:33 a.m.
On 11/20/18 8:56 AM, Wang Huaqiang wrote:
Adding the interface in qemu to report CMT statistic information
through command 'virsh domstats --cpu-total'.
Below is a typical output:
# virsh domstats 1 --cpu-total
Domain: 'ubuntu16.04-base'
...
cpu.cache.monitor.count=2
cpu.cache.monitor.0.name=vcpus_1
cpu.cache.monitor.0.vcpus=1
cpu.cache.monitor.0.bank.count=2
cpu.cache.monitor.0.bank.0.id=0
cpu.cache.monitor.0.bank.0.bytes=4505600
cpu.cache.monitor.0.bank.1.id=1
cpu.cache.monitor.0.bank.1.bytes=5586944
cpu.cache.monitor.1.name=vcpus_4-6
cpu.cache.monitor.1.vcpus=4,5,6
cpu.cache.monitor.1.bank.count=2
cpu.cache.monitor.1.bank.0.id=0
cpu.cache.monitor.1.bank.0.bytes=17571840
cpu.cache.monitor.1.bank.1.id=1
cpu.cache.monitor.1.bank.1.bytes=29106176
Signed-off-by: Wang Huaqiang <huaqiang.wang(a)intel.com>
---
src/libvirt-domain.c | 12 ++++
src/qemu/qemu_driver.c | 160 ++++++++++++++++++++++++++++++++++++++++++++++++-
tools/virsh.pod | 14 +++++
3 files changed, 185 insertions(+), 1 deletion(-)
diff --git a/src/libvirt-domain.c b/src/libvirt-domain.c
index 5b76458..73d602e 100644
--- a/src/libvirt-domain.c
+++ b/src/libvirt-domain.c
@@ -11415,6 +11415,18 @@ virConnectGetDomainCapabilities(virConnectPtr conn,
* "cpu.user" - user cpu time spent in nanoseconds as unsigned long long.
* "cpu.system" - system cpu time spent in nanoseconds as unsigned long
* long.
+ * "cpu.cache.monitor.count" - the number of cache monitors for this
domain
+ * "cpu.cache.monitor.<num>.name" - the name of cache monitor
<num>
+ * "cpu.cache.monitor.<num>.vcpus" - vcpu list of cache monitor
<num>
+ * "cpu.cache.monitor.<num>.bank.count" - the number of cache banks
in
+ * cache monitor <num>
+ * "cpu.cache.monitor.<num>.bank.<index>.id" - host allocated
cache id for
+ * bank <index> in cache
+ * monitor <num>
+ * "cpu.cache.monitor.<num>.bank.<index>.bytes" - the number
of bytes of
+ * last level cache that the
+ * domain is using on cache
+ * bank <index>
*
* VIR_DOMAIN_STATS_BALLOON:
* Return memory balloon device information.
diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c
index 7fb9102..d9e216c 100644
--- a/src/qemu/qemu_driver.c
+++ b/src/qemu/qemu_driver.c
@@ -19929,6 +19929,158 @@ typedef enum {
#define HAVE_JOB(flags) ((flags) & QEMU_DOMAIN_STATS_HAVE_JOB)
+typedef struct _virQEMUResctrlMonData virQEMUResctrlMonData;
+typedef virQEMUResctrlMonData *virQEMUResctrlMonDataPtr;
+struct _virQEMUResctrlMonData {
+ const char *name;
+ char *vcpus;
+ virResctrlMonitorStatsPtr stats;
+ size_t nstats;
+};
+
+
+static int
+qemuDomainGetResctrlMonData(virDomainObjPtr dom,
+ virQEMUResctrlMonDataPtr resdata)
+{
+ virDomainResctrlDefPtr resctrl = NULL;
+ size_t i = 0;
+ size_t j = 0;
+ size_t k = 0;
+
+ for (i = 0; i < dom->def->nresctrls; i++) {
+ resctrl = dom->def->resctrls[i];
+
+ for (j = 0; j < resctrl->nmonitors; j++) {
+ virDomainResctrlMonDefPtr domresmon = NULL;
+ virResctrlMonitorPtr monitor = NULL;
+
+ domresmon = resctrl->monitors[j];
+ monitor = domresmon->instance;
+
+ if (domresmon->tag != VIR_RESCTRL_MONITOR_TYPE_CACHE)
+ continue;
If you want to make this generic, then you could pass this tag from
qemuDomainGetStatsCpuCache as the rest would seemingly be useful for
VIR_RESCTRL_MONITOR_TYPE_MEMBW eventually, just different results.
+
+ /* If virBitmapFormat successfully returns an vcpu string, then
+ * resdata[k].vcpus is assigned with an memory space holding it,
+ * let this newly allocated memory buffer to be freed along with
+ * the free of 'resdata' */
+ if (!(resdata[k].vcpus = virBitmapFormat(domresmon->vcpus)))
+ return -1;
+
+ if (!(resdata[k].name = virResctrlMonitorGetID(monitor))) {
Could this ever be NULL? Perhaps we just assign directly and assume
we're good. Alternatively it's a VIR_STRDUP() w/ the corresponding
VIR_FREE(*->name).
+ virReportError(VIR_ERR_INTERNAL_ERROR,
"%s",
+ _("Could not get monitor ID"));
+ return -1;
+ }
+
+ if (virResctrlMonitorGetCacheOccupancy(monitor,
+ &resdata[k].stats,
+ &resdata[k].nstats) < 0)
+ return -1;
+
+ k++;
+ }
+ }
+
+ return 0;
+}
+
+
+static int
+qemuDomainGetStatsCpuCache(virDomainObjPtr dom,
+ virDomainStatsRecordPtr record,
+ int *maxparams)
+{
+ char param_name[VIR_TYPED_PARAM_FIELD_LENGTH];
+ virQEMUResctrlMonDataPtr resdata = NULL;
+ virDomainResctrlDefPtr resctrl = NULL;
+ virDomainResctrlMonDefPtr domresmon = NULL;
+ unsigned int nresdata = 0;
+ size_t i = 0;
+ size_t j = 0;
+ int ret = -1;
+
+ if (!virDomainObjIsActive(dom))
+ return 0;
+
+ for (i = 0; i < dom->def->nresctrls; i++) {
+ resctrl = dom->def->resctrls[i];
+ for (j = 0; j < resctrl->nmonitors; j++) {
+ domresmon = resctrl->monitors[j];
+ if (domresmon->tag == VIR_RESCTRL_MONITOR_TYPE_CACHE)
+ nresdata++;
+ }
+ }
+
+ if (nresdata == 0)
+ return 0;
+
+ if (VIR_ALLOC_N(resdata, nresdata) < 0)
+ return -1;
Given below - perhaps none of the above really matters if you follow how
virResctrlMonitorGetStats was coded using VIR_APPEND_ELEMENT to append
on each @resdata.
+
+ if (qemuDomainGetResctrlMonData(dom, resdata) < 0)
+ goto cleanup;
+
+ snprintf(param_name, VIR_TYPED_PARAM_FIELD_LENGTH,
+ "cpu.cache.monitor.count");
+ if (virTypedParamsAddUInt(&record->params, &record->nparams,
+ maxparams, param_name, nresdata) < 0)
+ goto cleanup;
+
+ for (i = 0; i < nresdata; i++) {
+ snprintf(param_name, VIR_TYPED_PARAM_FIELD_LENGTH,
+ "cpu.cache.monitor.%zu.name", i);
+ if (virTypedParamsAddString(&record->params,
+ &record->nparams,
+ maxparams,
+ param_name,
+ resdata[i].name) < 0)
+ goto cleanup;
+
+ snprintf(param_name, VIR_TYPED_PARAM_FIELD_LENGTH,
+ "cpu.cache.monitor.%zu.vcpus", i);
+ if (virTypedParamsAddString(&record->params, &record->nparams,
+ maxparams, param_name,
+ resdata[i].vcpus) < 0)
+ goto cleanup;
+
+ snprintf(param_name, VIR_TYPED_PARAM_FIELD_LENGTH,
+ "cpu.cache.monitor.%zu.bank.count", i);
+ if (virTypedParamsAddUInt(&record->params, &record->nparams,
+ maxparams, param_name,
+ resdata[i].nstats) < 0)
+ goto cleanup;
+
+ for (j = 0; j < resdata[i].nstats; j++) {
+ snprintf(param_name, VIR_TYPED_PARAM_FIELD_LENGTH,
+ "cpu.cache.monitor.%zu.bank.%zu.id", i, j);
+ if (virTypedParamsAddUInt(&record->params, &record->nparams,
+ maxparams, param_name,
+ resdata[i].stats[j].id) < 0)
+ goto cleanup;
+
+ snprintf(param_name, VIR_TYPED_PARAM_FIELD_LENGTH,
+ "cpu.cache.monitor.%zu.bank.%zu.bytes", i, j);
+ if (virTypedParamsAddUInt(&record->params, &record->nparams,
+ maxparams, param_name,
+ resdata[i].stats[j].val) < 0)
+ goto cleanup;
+ }
+ }
+
+ ret = 0;
+ cleanup:
+ for (i = 0; i < nresdata; i++) {
+ VIR_FREE(resdata[i].vcpus);
+ VIR_FREE(resdata[i].stats);
+ }
+ VIR_FREE(resdata);
All of this should be replaced by a call to qemuDomainFreeResctrlMonData
which would do the above, but replace the VIR_FREE(resdata[i].stats)
with a call to virResctrlMonitorFreeStats which would essentially:
if (!stats)
return;
for (i = 0; i < nstats; i++)
VIR_FREE(stats[i]);
VIR_FREE(stats);
This being the opposing action of virResctrlMonitorGetStats.
See and test if the attached patch works for you.
John
+
+ return ret;
+}
+
+
static int
qemuDomainGetStatsCpuCgroup(virDomainObjPtr dom,
virDomainStatsRecordPtr record,
@@ -19976,7 +20128,13 @@ qemuDomainGetStatsCpu(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
int *maxparams,
unsigned int privflags ATTRIBUTE_UNUSED)
{
- return qemuDomainGetStatsCpuCgroup(dom, record, maxparams);
+ if (qemuDomainGetStatsCpuCgroup(dom, record, maxparams) < 0)
+ return -1;
+
+ if (qemuDomainGetStatsCpuCache(dom, record, maxparams) < 0)
+ return -1;
+
+ return 0;
}
diff --git a/tools/virsh.pod b/tools/virsh.pod
index 4876656..86a4996 100644
--- a/tools/virsh.pod
+++ b/tools/virsh.pod
@@ -1012,6 +1012,20 @@ I<--cpu-total> returns:
"cpu.time" - total cpu time spent for this domain in nanoseconds
"cpu.user" - user cpu time spent in nanoseconds
"cpu.system" - system cpu time spent in nanoseconds
+ "cpu.cache.monitor.count" - the number of cache monitors for this
+ domain
+ "cpu.cache.monitor.<num>.name" - the name of cache monitor <num>
+ "cpu.cache.monitor.<num>.vcpus" - vcpu list of cache monitor
<num>
+ "cpu.cache.monitor.<num>.bank.count" - the number of cache banks
+ in cache monitor <num>
+ "cpu.cache.monitor.<num>.bank.<index>.id" - host allocated cache
id
+ for bank <index> in
+ cache monitor <num>
+ "cpu.cache.monitor.<num>.bank.<index>.bytes" - the number of
bytes
+ of last level cache
+ that the domain is
+ using on cache bank
+ <index>
I<--balloon> returns: