2020-08-07 21:31:11

by Jonathan Adams

Subject: [RFC PATCH 4/7] core/metricfs: expose softirq information through metricfs

Add metricfs support for displaying percpu softirq counters. The
top directory is /sys/kernel/debug/metricfs/softirq. Then there
is a subdirectory for each softirq type. For example:

cat /sys/kernel/debug/metricfs/softirq/NET_RX/values

Signed-off-by: Jonathan Adams <[email protected]>

---

[email protected]: rebased to 5.8-pre6
This is work originally done by another engineer at
google, who would rather not have their name associated with this
patchset. They're okay with me sending it under my name.
---
kernel/softirq.c | 45 +++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 45 insertions(+)

diff --git a/kernel/softirq.c b/kernel/softirq.c
index c4201b7f42b1..1ae3a540b789 100644
--- a/kernel/softirq.c
+++ b/kernel/softirq.c
@@ -25,6 +25,8 @@
#include <linux/smpboot.h>
#include <linux/tick.h>
#include <linux/irq.h>
+#include <linux/jump_label.h>
+#include <linux/metricfs.h>

#define CREATE_TRACE_POINTS
#include <trace/events/irq.h>
@@ -738,3 +740,46 @@ unsigned int __weak arch_dynirq_lower_bound(unsigned int from)
{
return from;
}
+
+#ifdef CONFIG_METRICFS
+
+#define METRICFS_ITEM(name) \
+static void \
+metricfs_##name(struct metric_emitter *e, int cpu) \
+{ \
+ int64_t v = kstat_softirqs_cpu(name##_SOFTIRQ, cpu); \
+ METRIC_EMIT_PERCPU_INT(e, cpu, v); \
+} \
+METRIC_EXPORT_PERCPU_COUNTER(name, #name " softirq", metricfs_##name)
+
+METRICFS_ITEM(HI);
+METRICFS_ITEM(TIMER);
+METRICFS_ITEM(NET_TX);
+METRICFS_ITEM(NET_RX);
+METRICFS_ITEM(BLOCK);
+METRICFS_ITEM(IRQ_POLL);
+METRICFS_ITEM(TASKLET);
+METRICFS_ITEM(SCHED);
+METRICFS_ITEM(HRTIMER);
+METRICFS_ITEM(RCU);
+
+static int __init init_softirq_metricfs(void)
+{
+ struct metricfs_subsys *subsys;
+
+ subsys = metricfs_create_subsys("softirq", NULL);
+ metric_init_HI(subsys);
+ metric_init_TIMER(subsys);
+ metric_init_NET_TX(subsys);
+ metric_init_NET_RX(subsys);
+ metric_init_BLOCK(subsys);
+ metric_init_IRQ_POLL(subsys);
+ metric_init_TASKLET(subsys);
+ metric_init_SCHED(subsys);
+ metric_init_RCU(subsys);
+
+ return 0;
+}
+module_init(init_softirq_metricfs);
+
+#endif
--
2.28.0.236.gb10cc79966-goog

2020-08-08 05:47:56

by Greg Kroah-Hartman

Subject: Re: [RFC PATCH 4/7] core/metricfs: expose softirq information through metricfs

On Fri, Aug 07, 2020 at 02:29:13PM -0700, Jonathan Adams wrote:
> Add metricfs support for displaying percpu softirq counters. The
> top directory is /sys/kernel/debug/metricfs/softirq. Then there
> is a subdirectory for each softirq type. For example:
>
> cat /sys/kernel/debug/metricfs/softirq/NET_RX/values
>
> Signed-off-by: Jonathan Adams <[email protected]>
>
> ---
>
> [email protected]: rebased to 5.8-pre6
> This is work originally done by another engineer at
> google, who would rather not have their name associated with this
> patchset. They're okay with me sending it under my name.
> ---
> kernel/softirq.c | 45 +++++++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 45 insertions(+)
>
> diff --git a/kernel/softirq.c b/kernel/softirq.c
> index c4201b7f42b1..1ae3a540b789 100644
> --- a/kernel/softirq.c
> +++ b/kernel/softirq.c
> @@ -25,6 +25,8 @@
> #include <linux/smpboot.h>
> #include <linux/tick.h>
> #include <linux/irq.h>
> +#include <linux/jump_label.h>
> +#include <linux/metricfs.h>
>
> #define CREATE_TRACE_POINTS
> #include <trace/events/irq.h>
> @@ -738,3 +740,46 @@ unsigned int __weak arch_dynirq_lower_bound(unsigned int from)
> {
> return from;
> }
> +
> +#ifdef CONFIG_METRICFS
> +
> +#define METRICFS_ITEM(name) \
> +static void \
> +metricfs_##name(struct metric_emitter *e, int cpu) \
> +{ \
> + int64_t v = kstat_softirqs_cpu(name##_SOFTIRQ, cpu); \
> + METRIC_EMIT_PERCPU_INT(e, cpu, v); \
> +} \
> +METRIC_EXPORT_PERCPU_COUNTER(name, #name " softirq", metricfs_##name)
> +
> +METRICFS_ITEM(HI);
> +METRICFS_ITEM(TIMER);
> +METRICFS_ITEM(NET_TX);
> +METRICFS_ITEM(NET_RX);
> +METRICFS_ITEM(BLOCK);
> +METRICFS_ITEM(IRQ_POLL);
> +METRICFS_ITEM(TASKLET);
> +METRICFS_ITEM(SCHED);
> +METRICFS_ITEM(HRTIMER);
> +METRICFS_ITEM(RCU);
> +
> +static int __init init_softirq_metricfs(void)
> +{
> + struct metricfs_subsys *subsys;
> +
> + subsys = metricfs_create_subsys("softirq", NULL);
> + metric_init_HI(subsys);
> + metric_init_TIMER(subsys);
> + metric_init_NET_TX(subsys);
> + metric_init_NET_RX(subsys);
> + metric_init_BLOCK(subsys);
> + metric_init_IRQ_POLL(subsys);
> + metric_init_TASKLET(subsys);
> + metric_init_SCHED(subsys);
> + metric_init_RCU(subsys);
> +
> + return 0;
> +}
> +module_init(init_softirq_metricfs);

I like the "simple" ways these look, and think you will be better off
just adding this type of api to debugfs. That way people can use them
anywhere they currently use debugfs.

But note, we already have simple ways of exporting single variable data
in debugfs, so why do we need yet-another-macro for them?

thanks,

greg k-h