[PATCH V2 12/13] i3c: mipi-i3c-hci-pci: Add LTR support for Intel controllers

[Frank Li]

On Mon, Nov 17, 2025 at 09:07:05PM +0200, Adrian Hunter wrote:
> On 17/11/2025 20:56, Frank Li wrote:
> > On Mon, Nov 17, 2025 at 08:09:41PM +0200, Adrian Hunter wrote:
> >> On 17/11/2025 19:54, Frank Li wrote:
> >>> On Mon, Nov 17, 2025 at 07:47:05PM +0200, Adrian Hunter wrote:
> >>>> On 17/11/2025 17:59, Frank Li wrote:
> >>>>> On Mon, Nov 17, 2025 at 01:36:36PM +0200, Adrian Hunter wrote:
> >>>>>> On 14/11/2025 22:09, Frank Li wrote:
> >>>>>>> On Fri, Nov 14, 2025 at 08:13:55PM +0200, Adrian Hunter wrote:
> >>>>>>>> Add support for Latency Tolerance Reporting (LTR) for Intel controllers.
> >>>>>>>>
> >>>>>>>> Implement PM ->set_latency_tolerance() callback to set LTR register values.
> >>>>>>>> Also expose LTR register values via debugfs.
> >>>>>>>>
> >>>>>>>> Signed-off-by: Adrian Hunter <adrian.hunter at intel.com>
> >>>>>>>> ---
> >>>>>>>>
> >>>>>>>>
> >>>>>>>> Changes in V2:
> >>>>>>>>
> >>>>>>>> 	Make use of FIELD_PREP()
> >>>>>>>> 	Improve handling of values that exceed the maximum
> >>>>>>>>
> >>>>>>>>
> >>>>>>>>  .../master/mipi-i3c-hci/mipi-i3c-hci-pci.c    | 125 ++++++++++++++++++
> >>>>>>>>  1 file changed, 125 insertions(+)
> >>>>>>>>
> >>>>>>> ...
> >>>>>>>> +static void intel_ltr_expose(struct device *dev)
> >>>>>>>> +{
> >>>>>>>> +	dev->power.set_latency_tolerance = intel_ltr_set;
> >>>>>>>> +	dev_pm_qos_expose_latency_tolerance(dev);
> >>>>>>>> +}
> >>>>>>>> +
> >>>>>>>> +static void intel_ltr_hide(struct device *dev)
> >>>>>>>> +{
> >>>>>>>> +	dev_pm_qos_hide_latency_tolerance(dev);
> >>>>>>>> +	dev->power.set_latency_tolerance = NULL;
> >>>>>>>> +}
> >>>>>>>> +
> >>>>>>>> +static struct dentry *intel_actualize_debugfs_root(bool add)
> >>>>>>>> +{
> >>>>>>>> +	static struct dentry *debugfs_root;
> >>>>>>>> +	static DEFINE_MUTEX(lock);
> >>>>>>>> +	static int ref_cnt;
> >>>>>>>> +
> >>>>>>>> +	guard(mutex)(&lock);
> >>>>>>>> +
> >>>>>>>> +	ref_cnt += add ? 1 : -1;
> >>>>>>>> +
> >>>>>>>> +	if (ref_cnt) {
> >>>>>>>> +		if (IS_ERR_OR_NULL(debugfs_root))
> >>>>>>>> +			debugfs_root = debugfs_create_dir("intel_i3c", NULL);
> >>>>>>>
> >>>>>>> I think it'd better to create common "i3c" debugfs root. So your device
> >>>>>>> use debugfs_create_dir(dev_name(&hci->pci->dev), ...) under this common
> >>>>>>> root.
> >>>>>>
> >>>>>> The debugfs files are being created for the PCI device to contain PCI
> >>>>>> LTR information.  The i3c module has nothing to do with that.
> >>>>>>
> >>>>>> Also, currently, there is no module dependency between this PCI driver
> >>>>>> and i3c, but a dependency would be necessary to coordinate the creation
> >>>>>> and deletion of an 'i3c' debugfs directory.
> >>>>>>
> >>>>>> The extra complexity can't be justified.
> >>>>>
> >>>>> Don't post new versin before we close open discussion. I see your post
> >>>>> patch before this message.
> >>>>
> >>>> Sorry, but it is late in the release cycle, and it saves time when we
> >>>> are in different time zones.
> >>>>
> >>>>>
> >>>>> So, I think not necessary to create to root, which cause complex, (need
> >>>>> lock, and refer counter).
> >>>>
> >>>> So you are OK with V3?
> >>>
> >>> In v3, still create "intel_i3c" root dir, how about direct create dev
> >>> under root
> >>>
> >>> 	debugfs_create_dir(dev_name(&hci->pci->dev), NULL);
> >>>
> >>> to reduce unneccesary complex, (like mutex lock and ref counter).
> >>
> >> Could do.  I added intel_i3c to match intel_lpss which is also
> >> PCI devices with LTR e.g.
> >
> > You can use similar method to create root at module_init.
> >
> > static int __init intel_lpss_init(void)
> > {
> >         intel_lpss_debugfs = debugfs_create_dir("intel_lpss", NULL);
> >         return 0;
> > }
> > module_init(intel_lpss_init);
>
> mipi-i3c-hci-pci is not an Intel driver.  It can be used for any
> PCI device that supports MIPI I3C.
>
> At module init, the driver does not know whether there are any
> Intel PCI devices, so it does not know whether to create "intel_i3c".
> Nor does it know whether other (non-Intel) devices need debugfs.
>
> It would be preferable to use either the current approach, or as
> you already suggested debugfs_create_dir(dev_name(&hci->pci->dev), NULL)

Okay use Debugfs_create_dir(dev_name(&hci->pci->dev), NULL).

Frank

>
>
> --
> linux-i3c mailing list
> linux-i3c at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-i3c