[PATCH v3 20/31] coresight: cti: Register PM notifier after data initialization

[Mike Leach]

Hi Leo

On Fri, 3 Oct 2025 at 15:50, Leo Yan <leo.yan at arm.com> wrote:
>
> On Fri, Oct 03, 2025 at 10:38:24AM +0100, Mike Leach wrote:
> > Hi Leo,
> >
> > The CTI PM code was initially set up before the CTI was a helper device.
> >
> > If the core bound CTIs, which are helpers to the core bound ETMs, are
> > disabled as part of the path disable helpers, then I think the CTI
> > specific PM code might be redundant.
>
> > If it is not, then should this not  hook into the generic PM code in
> > coresight-core.c?
>
> Correct, CTI devices will be managed as a part of path in
> coresight-core.c (See patches 26 and 31).
>
> However, we still need cti_pm_setup() - when a CPU bound CTI is enabled
> via CTI's sysfs interface, in this case, no path is enabled but purely
> CTI device is enabled.
>
> A side topic not related to this series, I am curious if we can remove
> the enable/disable knobs in coresight-cti-sysfs.c, then CTI will only
> work as helper. This can significantly simplify the CTI driver.
>

Not all CTIs are on the path, but potentially even non-path CTIs can
introduce signal inputs to the CTM that trigger events in the trace
path.

Mike

> Thanks,
> Leo



-- 
Mike Leach
Principal Engineer, ARM Ltd.
Manchester Design Centre. UK