[PATCH v3 01/12] perf script: Make printing flags reliable

[Leo Yan]

Hi Adrian,

On Mon, Mar 03, 2025 at 12:44:33PM +0200, Adrian Hunter wrote:
> On 17/02/25 21:58, Leo Yan wrote:
> > Add a check for the generated string of flags.  Print out the raw number
> > if the string generation fails.
> 
> How does it fail?

In practice, I agreed perf_sample__sprintf_flags() will not fail.  This
bases on a careful calculation for every invoking snprintf().

Please see comment below.

> > In another case, if the string length is longer than the aligned size,
> > allow the completed string to be printed.
> > 
> > Reviewed-by: Ian Rogers <irogers at google.com>
> > Reviewed-by: James Clark <james.clark at linaro.org>
> > Signed-off-by: Leo Yan <leo.yan at arm.com>
> > ---
> >  tools/perf/builtin-script.c   | 10 ++++++++--
> >  tools/perf/util/trace-event.h |  2 ++
> >  2 files changed, 10 insertions(+), 2 deletions(-)
> > 
> > diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c
> > index d797cec4f054..2c4b1fb7dc72 100644
> > --- a/tools/perf/builtin-script.c
> > +++ b/tools/perf/builtin-script.c
> > @@ -1709,9 +1709,15 @@ static int perf_sample__fprintf_bts(struct perf_sample *sample,
> >  static int perf_sample__fprintf_flags(u32 flags, FILE *fp)
> >  {
> >  	char str[SAMPLE_FLAGS_BUF_SIZE];
> > +	int ret;
> > +
> > +	ret = perf_sample__sprintf_flags(flags, str, sizeof(str));
> > +	if (ret < 0)
> 
> AFAICT ret is always >= 0

Since I refactored perf_sample__sprintf_flags() in the sequential
patches, for easier capturing and debugging, here checks the return
value to detect any potential issues.

Later when we review a perf log, a printed raw number for error cases
can remind there must be something wrong for printing flags.

> > +		return fprintf(fp, "  raw flags:0x%-*x ",
> > +			       SAMPLE_FLAGS_STR_ALIGNED_SIZE - 12, flags);
> >  
> > -	perf_sample__sprintf_flags(flags, str, sizeof(str));
> > -	return fprintf(fp, "  %-21s ", str);
> > +	ret = max(ret, SAMPLE_FLAGS_STR_ALIGNED_SIZE);
> > +	return fprintf(fp, "  %-*s ", ret, str);
> 
> -21 means the field width is 21 and left-justified.  It should not
> truncate the string.

No, it does not truncate the string.

It calculates a maximum value between the returned length and 21 (
defined in SAMPLE_FLAGS_STR_ALIGNED_SIZE).  It keeps left-justified and
can printing a complete string if the string length is bigger than 21.

Thanks,
Leo

> 
> >  }
> >  
> >  struct printer_data {
> > diff --git a/tools/perf/util/trace-event.h b/tools/perf/util/trace-event.h
> > index ac9fde2f980c..71e680bc3d4b 100644
> > --- a/tools/perf/util/trace-event.h
> > +++ b/tools/perf/util/trace-event.h
> > @@ -145,6 +145,8 @@ int common_flags(struct scripting_context *context);
> >  int common_lock_depth(struct scripting_context *context);
> >  
> >  #define SAMPLE_FLAGS_BUF_SIZE 64
> > +#define SAMPLE_FLAGS_STR_ALIGNED_SIZE	21
> > +
> >  int perf_sample__sprintf_flags(u32 flags, char *str, size_t sz);
> >  
> >  #if defined(LIBTRACEEVENT_VERSION) &&  LIBTRACEEVENT_VERSION >= MAKE_LIBTRACEEVENT_VERSION(1, 5, 0)
>