[PATCH] kernel-shark: Handle traces with sched_wakeup but not sched_waking

[PATCH] kernel-shark: Handle traces with sched_wakeup but not sched_waking

[John Keeping]

If sched_wakeup or sched_wakeup_new is avaiable but sched_waking is not,
then define_wakeup_event() returns true even though waking_event_ptr is
null.

Change find_wakeup_event() to avoid overwriting the output parameter on
error so that the define_wakeup_event() returns true iff
*waking_event_ptr is non-null.

Signed-off-by: John Keeping <john@metanate.com>
---
 src/libkshark-tepdata.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/src/libkshark-tepdata.c b/src/libkshark-tepdata.c
index 865ca82..9740ed9 100644
--- a/src/libkshark-tepdata.c
+++ b/src/libkshark-tepdata.c
@@ -1868,9 +1868,14 @@ int kshark_tep_find_top_stream(struct kshark_context *kshark_ctx,
 static bool find_wakeup_event(struct tep_handle *tep, const char *wakeup_name,
 			      struct tep_event **waking_event_ptr)
 {
-	*waking_event_ptr = tep_find_event_by_name(tep, "sched", wakeup_name);
+	struct tep_event *event;
+
+	event = tep_find_event_by_name(tep, "sched", wakeup_name);
+
+	if (event)
+		*waking_event_ptr = event;
 
-	return (*waking_event_ptr)? true : false;
+	return !!event;
 }
 
 /**
-- 
2.33.0

Re: [PATCH] kernel-shark: Handle traces with sched_wakeup but not sched_waking

[Yordan Karadzhov]

On 23.09.21 г. 17:06, John Keeping wrote:
> If sched_wakeup or sched_wakeup_new is avaiable but sched_waking is not,
> then define_wakeup_event() returns true even though waking_event_ptr is
> null.
> 
> Change find_wakeup_event() to avoid overwriting the output parameter on
> error so that the define_wakeup_event() returns true iff
> *waking_event_ptr is non-null.
> 
> Signed-off-by: John Keeping <john@metanate.com>

Thanks a lot for helping us to improve KernelShark!

It seems that you are interested in analyzing wakeup latency and you are keen of digging into code, so I would like to 
point you to something new that may be interesting for you.

We are trying to develop Python bindings for the tracing libraries, including libkshark. Here is one very simple example 
script that plots the distribution of the latency and generates a description of a KernelShark session directly showing 
the largest latency:
https://github.com/vmware/trace-cruncher/blob/master/examples/sched_wakeup.py

Note that this example is supposed to demonstrate the Python APIs, not to do something that is particularly useful on 
its own. But you should be able to easily adapt it to your own needs. The installation of the Python module is very 
simple. See the README here:
https://github.com/vmware/trace-cruncher

Please let me know if this is something interesting for you. I will be vary happy to receive patches from you for this 
project as well. I am pretty sure you will find bugs ;-)

Once again, thanks a lot for the KernelShark patches!
Cheers,
Yordan


> ---
>   src/libkshark-tepdata.c | 9 +++++++--
>   1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/src/libkshark-tepdata.c b/src/libkshark-tepdata.c
> index 865ca82..9740ed9 100644
> --- a/src/libkshark-tepdata.c
> +++ b/src/libkshark-tepdata.c
> @@ -1868,9 +1868,14 @@ int kshark_tep_find_top_stream(struct kshark_context *kshark_ctx,
>   static bool find_wakeup_event(struct tep_handle *tep, const char *wakeup_name,
>   			      struct tep_event **waking_event_ptr)
>   {
> -	*waking_event_ptr = tep_find_event_by_name(tep, "sched", wakeup_name);
> +	struct tep_event *event;
> +
> +	event = tep_find_event_by_name(tep, "sched", wakeup_name);
> +
> +	if (event)
> +		*waking_event_ptr = event;
>   
> -	return (*waking_event_ptr)? true : false;
> +	return !!event;
>   }
>   
>   /**
>