[PATCH] coresight: etm4x: Fix timestamp bit field handling

[Leo Yan]

On Mon, May 19, 2025 at 04:12:32PM +0100, Suzuki Kuruppassery Poulose wrote:

[...]

> > > --- a/drivers/hwtracing/coresight/coresight-etm4x-sysfs.c
> > > +++ b/drivers/hwtracing/coresight/coresight-etm4x-sysfs.c
> > > @@ -2315,23 +2315,10 @@ static ssize_t ts_source_show(struct device *dev,
> > >       int val;
> > >       struct etmv4_drvdata *drvdata = dev_get_drvdata(dev->parent);
> > > -    if (!drvdata->trfcr) {
> > > +    val = FIELD_GET(TRFCR_EL1_TS_MASK, drvdata->trfcr);
> > > +    if (!val)
> 
> I think this might be problematic. TS==0 is a reserved value for
> software, and doesn't imply a TS is not in effect.

Good point.

> Thus I think we should retain the older check as before to ensure
> TRFCR is effective.

If we intend to return TS values directly (include '0' case), then it
makes sense to make change like:

    if (!drvdata->trfcr)
        val = -1;
    else
        val = FIELD_GET(TRFCR_EL1_TS_MASK, drvdata->trfcr);

    return sysfs_emit(buf, "%d\n", val);

I checked again in Linux perf tool, it would be fine if we return 0 for
TS==0 case, as the perf tool now only use virtual time counter (see
cs_etm__has_virtual_ts()).

@James, could you confirm if this fine?

Thanks,
Leo