[PATCHv3 3/3] dynamic_debug: Add a flag for dynamic event tracing

Jason Baron jbaron at akamai.com
Tue Nov 9 13:42:48 PST 2021



On 11/9/21 12:49 PM, Sai Prakash Ranjan wrote:
> On 11/9/2021 11:10 PM, Steven Rostedt wrote:
>> On Tue, 9 Nov 2021 23:00:11 +0530
>> Sai Prakash Ranjan <quic_saipraka at quicinc.com> wrote:
>>
>>> Ah that's a very good idea, descriptor does contain the module, file name.
>>> We can probably even pass the module name,file name as string from the
>>> descriptor itself to event?
>>> Perhaps we can do that for all trace events and not just this trace
>>> event? Just like the trace event name displayed
>>> with trace events, perhaps have file name,module name displayed when
>>> dynamic debug is enabled? Filtering by
>>> filename is pretty useful since most of these usecases in debugging will
>>> be with respect to some driver or subsystems.
>> If we add this for all events, it would require a lot of changes to all
>> users of tracepoints, as it would require adding a new parameter to the
>> callbacks.
>>
>> We could add a flag in the registering that states that the callback is OK
>> for it, and it passes that data as well.
>>
>> Let me look into this for a bit. I may not have something this week, but
>> we should look into this more before adding little hacks that do this one
>> at a time like this patch.
>>
>> -- Steve
> 
> Sure, thanks for the help and review. I can skip this patch adding support for dynamic event tracing
> till we have something concrete as the previous patches doesn't depend on this.
> 
> Thanks,
> Sai


Hi,

Yeah there is a 'parallel' thread about adding the tracing ring buffer as a 'back end' to the dynamic debug stuff over here:
https://lore.kernel.org/lkml/20211105192637.2370737-9-jim.cromie@gmail.com/

The attempt there is more generic but I realize now that it is adding the tracing to an 'instance' which is specific to dynamic debug which is being
created via: trace_array_get_by_name(). I would prefer to just have it print to the 'main' trace buffer such that it's easier to read, although I
guess they could still be consolidated via timestamps. Hmmm...I think there was a previous proposal to just add a single tracepoint (that takes a
string) to the dynamic debug layer that could be called if a dynamic debug site is enabled for trace buffer output. Would that satisfy the ftrace
level filtering requirements that you are looking for?

Thanks,

-Jason



More information about the linux-arm-kernel mailing list