2021-10-07 03:53:58

by William Cohen

Subject: Reason for unused flags argument in trace_raw_output_* function?

Hi,

I have some code that analyzes x86_64 binaries (https://developers.redhat.com/articles/2021/08/09/debugging-function-parameters-
dyninst#liveness_analysis) and list which function parameters are unused. I noticed that trace_raw_output_* functions have an unused flags argument. I assume this is is to make the argument list match up with some other function, but I couldn't see which function(s) the trace_raw_output_* functions are trying to match up with. Which functions are the trace_raw_output_* function's arguments trying to match?


-Will Cohen

2021-10-07 13:25:40

by Steven Rostedt

Subject: Re: Reason for unused flags argument in trace_raw_output_* function?

On Wed, 6 Oct 2021 23:49:22 -0400
William Cohen <[email protected]> wrote:

> Hi,
>
> I have some code that analyzes x86_64 binaries
> (https://developers.redhat.com/articles/2021/08/09/debugging-function-parameters-
> dyninst#liveness_analysis) and list which function parameters are unused.
> I noticed that trace_raw_output_* functions have an unused flags
> argument. I assume this is is to make the argument list match up with
> some other function, but I couldn't see which function(s) the
> trace_raw_output_* functions are trying to match up with. Which
> functions are the trace_raw_output_* function's arguments trying to match?
>

The match of other printing functions is in kernel/trace/trace_output.c

And is called from kernel/trace/trace.c where one place it does pass in
flags:


static enum print_line_t print_trace_fmt(struct trace_iterator *iter)
{
[..]
if (event)
return event->funcs->trace(iter, sym_flags, event);

Now if they are still used, is something we should investigate.

This code has changed a lot over the years when that "flags" parameter was
introduced. It could now just be a relic from the old implementations.

Thanks for pointing this out. If I get time I'll look deeper into it, but
feel free to look at it yourself.

-- Steve

2021-10-07 14:36:16

by William Cohen

Subject: Re: Reason for unused flags argument in trace_raw_output_* function?

On 10/7/21 09:20, Steven Rostedt wrote:
> On Wed, 6 Oct 2021 23:49:22 -0400
> William Cohen <[email protected]> wrote:
>
>> Hi,
>>
>> I have some code that analyzes x86_64 binaries
>> (https://developers.redhat.com/articles/2021/08/09/debugging-function-parameters-
>> dyninst#liveness_analysis) and list which function parameters are unused.
>> I noticed that trace_raw_output_* functions have an unused flags
>> argument. I assume this is is to make the argument list match up with
>> some other function, but I couldn't see which function(s) the
>> trace_raw_output_* functions are trying to match up with. Which
>> functions are the trace_raw_output_* function's arguments trying to match?
>>
>
> The match of other printing functions is in kernel/trace/trace_output.c
>
> And is called from kernel/trace/trace.c where one place it does pass in
> flags:
>
>
> static enum print_line_t print_trace_fmt(struct trace_iterator *iter)
> {
> [..]
> if (event)
> return event->funcs->trace(iter, sym_flags, event);


Hi Steve,

Thanks for pointing to where the trace_raw_output_* functions are called from.

>
> Now if they are still used, is something we should investigate.

For the trace_raw_output_* functions the flags arg all seems to be unused. Are there other trace functions should examining for unused args? With the extensive macro use I could see where some vestigial unused argument gets left behind.

>
> This code has changed a lot over the years when that "flags" parameter was
> introduced. It could now just be a relic from the old implementations.
>
> Thanks for pointing this out. If I get time I'll look deeper into it, but
> feel free to look at it yourself.
>
> -- Steve
>

Thanks again,

-Will