[RFC v2 2/4] powerpc/papr_scm: Add perf interface support

From: Kajol Jain
Date: Tue May 25 2021 - 09:23:40 EST


Patch adds support for performance monitoring of papr-scm
nvdimm devices via perf interface. It adds pmu functions
like add/del/read/event_init for nvdimm_pmu struture.

Patch adds a new parameter 'priv' in pdev_archdata structure to save
nvdimm_pmu device pointer, to handle the unregistering of pmu device.

papr_scm_pmu_register function populates the nvdimm_pmu structure
with events, attribute groups along with event handling functions.
Finally the populated nvdimm_pmu structure is passed to register
the pmu device.
Event handling functions internally uses hcall to get events and
counter data.

Result in power9 machine with 2 nvdimm device:

Ex: List all event by perf list

command:# perf list nmem

nmem0/cchrhcnt/ [Kernel PMU event]
nmem0/cchwhcnt/ [Kernel PMU event]
nmem0/critrscu/ [Kernel PMU event]
nmem0/ctlresct/ [Kernel PMU event]
nmem0/ctlrestm/ [Kernel PMU event]
nmem0/fastwcnt/ [Kernel PMU event]
nmem0/hostlcnt/ [Kernel PMU event]
nmem0/hostldur/ [Kernel PMU event]
nmem0/hostscnt/ [Kernel PMU event]
nmem0/hostsdur/ [Kernel PMU event]
nmem0/medrcnt/ [Kernel PMU event]
nmem0/medrdur/ [Kernel PMU event]
nmem0/medwcnt/ [Kernel PMU event]
nmem0/medwdur/ [Kernel PMU event]
nmem0/memlife/ [Kernel PMU event]
nmem0/noopstat/ [Kernel PMU event]
nmem0/ponsecs/ [Kernel PMU event]
nmem1/cchrhcnt/ [Kernel PMU event]
nmem1/cchwhcnt/ [Kernel PMU event]
nmem1/critrscu/ [Kernel PMU event]
...
nmem1/noopstat/ [Kernel PMU event]
nmem1/ponsecs/ [Kernel PMU event]

Signed-off-by: Kajol Jain <kjain@xxxxxxxxxxxxx>
---
arch/powerpc/include/asm/device.h | 5 +
arch/powerpc/platforms/pseries/papr_scm.c | 358 ++++++++++++++++++++++
2 files changed, 363 insertions(+)

diff --git a/arch/powerpc/include/asm/device.h b/arch/powerpc/include/asm/device.h
index 219559d65864..47ed639f3b8f 100644
--- a/arch/powerpc/include/asm/device.h
+++ b/arch/powerpc/include/asm/device.h
@@ -48,6 +48,11 @@ struct dev_archdata {

struct pdev_archdata {
u64 dma_mask;
+ /*
+ * Pointer to nvdimm_pmu structure, to handle the unregistering
+ * of pmu device
+ */
+ void *priv;
};

#endif /* _ASM_POWERPC_DEVICE_H */
diff --git a/arch/powerpc/platforms/pseries/papr_scm.c b/arch/powerpc/platforms/pseries/papr_scm.c
index ef26fe40efb0..f2d57da98ff4 100644
--- a/arch/powerpc/platforms/pseries/papr_scm.c
+++ b/arch/powerpc/platforms/pseries/papr_scm.c
@@ -18,6 +18,8 @@
#include <asm/plpar_wrappers.h>
#include <asm/papr_pdsm.h>
#include <asm/mce.h>
+#include <linux/perf_event.h>
+#include <linux/ctype.h>

#define BIND_ANY_ADDR (~0ul)

@@ -67,6 +69,8 @@
#define PAPR_SCM_PERF_STATS_EYECATCHER __stringify(SCMSTATS)
#define PAPR_SCM_PERF_STATS_VERSION 0x1

+#define to_nvdimm_pmu(_pmu) container_of(_pmu, struct nvdimm_pmu, pmu)
+
/* Struct holding a single performance metric */
struct papr_scm_perf_stat {
u8 stat_id[8];
@@ -116,6 +120,12 @@ struct papr_scm_priv {

/* length of the stat buffer as expected by phyp */
size_t stat_buffer_len;
+
+ /* array to have event_code and stat_id mappings */
+ char **nvdimm_events_map;
+
+ /* count of supported events */
+ u32 total_events;
};

static int papr_scm_pmem_flush(struct nd_region *nd_region,
@@ -329,6 +339,347 @@ static ssize_t drc_pmem_query_stats(struct papr_scm_priv *p,
return 0;
}

+PMU_FORMAT_ATTR(event, "config:0-4");
+
+static struct attribute *nvdimm_pmu_format_attr[] = {
+ &format_attr_event.attr,
+ NULL,
+};
+
+static struct attribute_group nvdimm_pmu_format_group = {
+ .name = "format",
+ .attrs = nvdimm_pmu_format_attr,
+};
+
+static int papr_scm_pmu_get_value(struct perf_event *event, struct device *dev, u64 *count)
+{
+ struct papr_scm_perf_stat *stat;
+ struct papr_scm_perf_stats *stats;
+ struct papr_scm_priv *p = (struct papr_scm_priv *)dev->driver_data;
+ int rc, size;
+
+ /* Allocate request buffer enough to hold single performance stat */
+ size = sizeof(struct papr_scm_perf_stats) +
+ sizeof(struct papr_scm_perf_stat);
+
+ if (!p || !p->nvdimm_events_map)
+ return -EINVAL;
+
+ stats = kzalloc(size, GFP_KERNEL);
+ if (!stats)
+ return -ENOMEM;
+
+ stat = &stats->scm_statistic[0];
+ memcpy(&stat->stat_id,
+ p->nvdimm_events_map[event->attr.config - 1],
+ sizeof(stat->stat_id));
+ stat->stat_val = 0;
+
+ rc = drc_pmem_query_stats(p, stats, 1);
+ if (rc < 0) {
+ kfree(stats);
+ return rc;
+ }
+
+ *count = be64_to_cpu(stat->stat_val);
+ kfree(stats);
+ return 0;
+}
+
+static int papr_scm_pmu_event_init(struct perf_event *event)
+{
+ struct nvdimm_pmu *nd_pmu = to_nvdimm_pmu(event->pmu);
+ struct papr_scm_priv *p;
+
+ if (!nd_pmu)
+ return -EINVAL;
+
+ /* test the event attr type for PMU enumeration */
+ if (event->attr.type != event->pmu->type)
+ return -ENOENT;
+
+ /* it does not support event sampling mode */
+ if (is_sampling_event(event))
+ return -EOPNOTSUPP;
+
+ /* no branch sampling */
+ if (has_branch_stack(event))
+ return -EOPNOTSUPP;
+
+ p = (struct papr_scm_priv *)nd_pmu->dev->driver_data;
+ if (!p)
+ return -EINVAL;
+
+ /* Invalid eventcode */
+ if (event->attr.config == 0 || event->attr.config > p->total_events)
+ return -EINVAL;
+
+ return 0;
+}
+
+static int papr_scm_pmu_add(struct perf_event *event, int flags)
+{
+ u64 count;
+ int rc;
+ struct nvdimm_pmu *nd_pmu = to_nvdimm_pmu(event->pmu);
+
+ if (!nd_pmu)
+ return -EINVAL;
+
+ if (flags & PERF_EF_START) {
+ rc = papr_scm_pmu_get_value(event, nd_pmu->dev, &count);
+ if (rc)
+ return rc;
+
+ local64_set(&event->hw.prev_count, count);
+ }
+
+ return 0;
+}
+
+static void papr_scm_pmu_read(struct perf_event *event)
+{
+ u64 prev, now;
+ int rc;
+ struct nvdimm_pmu *nd_pmu = to_nvdimm_pmu(event->pmu);
+
+ if (!nd_pmu)
+ return;
+
+ rc = papr_scm_pmu_get_value(event, nd_pmu->dev, &now);
+ if (rc)
+ return;
+
+ prev = local64_xchg(&event->hw.prev_count, now);
+ local64_add(now - prev, &event->count);
+}
+
+static void papr_scm_pmu_del(struct perf_event *event, int flags)
+{
+ papr_scm_pmu_read(event);
+}
+
+static ssize_t device_show_string(struct device *dev, struct device_attribute *attr,
+ char *buf)
+{
+ struct perf_pmu_events_attr *d;
+
+ d = container_of(attr, struct perf_pmu_events_attr, attr);
+
+ return sysfs_emit(buf, "%s\n", (char *)d->event_str);
+}
+
+static char *strtolower(char *updated_name)
+{
+ int i = 0;
+
+ while (updated_name[i]) {
+ if (isupper(updated_name[i]))
+ updated_name[i] = tolower(updated_name[i]);
+ i++;
+ }
+ updated_name[i] = '\0';
+ return strim(updated_name);
+}
+
+/* device_str_attr_create : Populate event "name" and string "str" in attribute */
+static struct attribute *device_str_attr_create_(char *name, char *str)
+{
+ struct perf_pmu_events_attr *attr;
+
+ attr = kzalloc(sizeof(*attr), GFP_KERNEL);
+
+ if (!attr)
+ return NULL;
+
+ sysfs_attr_init(&attr->attr.attr);
+ attr->event_str = str;
+ attr->attr.attr.name = strtolower(name);
+ attr->attr.attr.mode = 0444;
+ attr->attr.show = device_show_string;
+
+ return &attr->attr.attr;
+}
+
+static int papr_scm_pmu_check_events(struct papr_scm_priv *p, struct nvdimm_pmu *nd_pmu)
+{
+ struct papr_scm_perf_stat *stat;
+ struct papr_scm_perf_stats *stats, *single_stats;
+ int index, size, rc, count;
+ u32 available_events;
+ struct attribute **events;
+ char *eventcode, *eventname, *statid;
+ struct attribute_group *nvdimm_pmu_events_group;
+
+ if (!p->stat_buffer_len)
+ return -ENOENT;
+
+ available_events = (p->stat_buffer_len - sizeof(struct papr_scm_perf_stats))
+ / sizeof(struct papr_scm_perf_stat);
+
+ /* Allocate memory for events attribute group */
+ nvdimm_pmu_events_group = kzalloc(sizeof(*nvdimm_pmu_events_group), GFP_KERNEL);
+ if (!nvdimm_pmu_events_group)
+ return -ENOMEM;
+
+ /* Allocate the buffer for phyp where stats are written */
+ stats = kzalloc(p->stat_buffer_len, GFP_KERNEL);
+ if (!stats) {
+ rc = -ENOMEM;
+ goto out_nvdimm_pmu_events_group;
+ }
+
+ /* Allocate memory to nvdimm_event_map */
+ p->nvdimm_events_map = kcalloc(available_events, sizeof(char *), GFP_KERNEL);
+ if (!p->nvdimm_events_map) {
+ rc = -ENOMEM;
+ goto out_stats;
+ }
+
+ /* Called to get list of events supported */
+ rc = drc_pmem_query_stats(p, stats, 0);
+ if (rc)
+ goto out_nvdimm_events_map;
+
+ /* Allocate buffer to hold single performance stat */
+ size = sizeof(struct papr_scm_perf_stats) + sizeof(struct papr_scm_perf_stat);
+
+ single_stats = kzalloc(size, GFP_KERNEL);
+ if (!single_stats) {
+ rc = -ENOMEM;
+ goto out_nvdimm_events_map;
+ }
+
+ events = kzalloc(available_events * sizeof(struct attribute *), GFP_KERNEL);
+ if (!events) {
+ rc = -ENOMEM;
+ goto out_single_stats;
+ }
+
+ for (index = 0, stat = stats->scm_statistic, count = 0;
+ index < available_events; index++, ++stat) {
+
+ single_stats->scm_statistic[0] = *stat;
+ rc = drc_pmem_query_stats(p, single_stats, 1);
+
+ if (rc < 0) {
+ pr_info("Event not supported %s for device %s\n",
+ stat->stat_id, nvdimm_name(p->nvdimm));
+ } else {
+ eventcode = kasprintf(GFP_KERNEL, "event=0x%x", count + 1);
+ eventname = kzalloc(strlen(stat->stat_id) + 1, GFP_KERNEL);
+ statid = kzalloc(strlen(stat->stat_id) + 1, GFP_KERNEL);
+
+ if (!eventname || !statid || !eventcode)
+ goto out;
+
+ strcpy(eventname, stat->stat_id);
+ events[count] = device_str_attr_create_(eventname,
+ eventcode);
+ if (!events[count])
+ goto out;
+
+ strcpy(statid, stat->stat_id);
+ p->nvdimm_events_map[count] = statid;
+ count++;
+ continue;
+out:
+ kfree(eventcode);
+ kfree(eventname);
+ kfree(statid);
+ }
+ }
+
+ if (!count)
+ goto out_events;
+
+ events[count] = NULL;
+ p->nvdimm_events_map[count] = NULL;
+ p->total_events = count;
+
+ nvdimm_pmu_events_group->name = "events";
+ nvdimm_pmu_events_group->attrs = events;
+
+ /* Fill attribute groups for the nvdimm pmu device */
+ nd_pmu->attr_groups[NVDIMM_PMU_FORMAT_ATTR] = &nvdimm_pmu_format_group;
+ nd_pmu->attr_groups[NVDIMM_PMU_EVENT_ATTR] = nvdimm_pmu_events_group;
+ nd_pmu->attr_groups[NVDIMM_PMU_NULL_ATTR] = NULL;
+
+ kfree(single_stats);
+ kfree(stats);
+ return 0;
+
+out_events:
+ kfree(events);
+out_single_stats:
+ kfree(single_stats);
+out_nvdimm_events_map:
+ kfree(p->nvdimm_events_map);
+out_stats:
+ kfree(stats);
+out_nvdimm_pmu_events_group:
+ kfree(nvdimm_pmu_events_group);
+ return rc;
+}
+
+/* Function to free the attr_groups which are dynamically allocated */
+static void nvdimm_pmu_mem_free(struct nvdimm_pmu *nd_pmu)
+{
+ if (nd_pmu) {
+ if (nd_pmu->attr_groups[NVDIMM_PMU_EVENT_ATTR])
+ kfree(nd_pmu->attr_groups[NVDIMM_PMU_EVENT_ATTR]->attrs);
+ kfree(nd_pmu->attr_groups[NVDIMM_PMU_EVENT_ATTR]);
+ }
+}
+
+static void papr_scm_pmu_register(struct papr_scm_priv *p)
+{
+ struct nvdimm_pmu *nd_pmu;
+ int rc;
+
+ nd_pmu = kzalloc(sizeof(*nd_pmu), GFP_KERNEL);
+ if (!nd_pmu) {
+ rc = -ENOMEM;
+ goto pmu_err_print;
+ }
+
+ rc = papr_scm_pmu_check_events(p, nd_pmu);
+ if (rc)
+ goto pmu_check_events_err;
+
+ nd_pmu->name = nvdimm_name(p->nvdimm);
+ nd_pmu->event_init = papr_scm_pmu_event_init;
+ nd_pmu->read = papr_scm_pmu_read;
+ nd_pmu->add = papr_scm_pmu_add;
+ nd_pmu->del = papr_scm_pmu_del;
+
+ rc = register_nvdimm_pmu(nd_pmu, p->pdev);
+ if (rc)
+ goto pmu_register_err;
+
+ /*
+ * Set archdata.priv value to nvdimm_pmu structure, to handle the
+ * unregistering of pmu device.
+ */
+ p->pdev->archdata.priv = nd_pmu;
+ return;
+
+pmu_register_err:
+ nvdimm_pmu_mem_free(nd_pmu);
+ kfree(p->nvdimm_events_map);
+pmu_check_events_err:
+ kfree(nd_pmu);
+pmu_err_print:
+ dev_info(&p->pdev->dev, "nvdimm pmu didn't register rc=%d\n", rc);
+}
+
+static void nvdimm_pmu_uinit(struct nvdimm_pmu *nd_pmu)
+{
+ unregister_nvdimm_pmu(&nd_pmu->pmu);
+ nvdimm_pmu_mem_free(nd_pmu);
+ kfree(nd_pmu);
+}
+
/*
* Issue hcall to retrieve dimm health info and populate papr_scm_priv with the
* health information.
@@ -1177,6 +1528,7 @@ static int papr_scm_probe(struct platform_device *pdev)
goto err2;

platform_set_drvdata(pdev, p);
+ papr_scm_pmu_register(p);

return 0;

@@ -1195,7 +1547,13 @@ static int papr_scm_remove(struct platform_device *pdev)

nvdimm_bus_unregister(p->bus);
drc_pmem_unbind(p);
+
+ if (pdev->archdata.priv)
+ nvdimm_pmu_uinit(pdev->archdata.priv);
+
+ pdev->archdata.priv = NULL;
kfree(p->bus_desc.provider_name);
+ kfree(p->nvdimm_events_map);
kfree(p);

return 0;
--
2.27.0