Re: [PATCH RFC v3 0/8] Add support for DisplayPort link training information report

From: Kory Maincent

Date: Mon Nov 16 2026 - 10:14:15 EST




On 11/16/26 14:10, Jani Nikula wrote:
On Mon, 12 Oct 2026, Kory Maincent <kory.maincent@xxxxxxxxxxx> wrote:
DisplayPort link training negotiates the physical-layer parameters needed
for a reliable connection: lane count, link rate, and optionally Display
Stream Compression (DSC). Currently, each driver exposes this state in
its own way, often through driver-specific debugfs entries, with no
standard interface for userspace diagnostic and monitoring tools.

This series introduces generic helpers, for exposing DP link capabilities
and state as standard sysfs entries.

So how did we end up with sysfs here?

In the first RFC it was in DRM properties. I presented it in the display hackfest in Nice last years. People suggested to use sysfs instead but if the need appears later in the development move back to DRM properties. I first though DRM properties seems better as it is the user interface used by distros but I had no strong argument about the change to sysfs, so I moved on to it.

The aim of such development is to guide users to select the most suitable
DisplayPort connector for their needs. For example, if you have a USB-C
hub with lesser capabilities than your computer’s native DisplayPort
connector (such as HBR2 versus HBR3 support), the system could recommend
connecting high-resolution displays directly to the computer’s port
instead of through the hub to ensure optimal performance.

I do wonder how all this is supposed to work in more complex real world
scenarios, though.

I believe the user primarily cares about *modes*. Which display modes
can be used with each display/connector pair.

With the low level link information, are you proposing to compute the
required bandwidth for modes in userspace? I think it's going to be a
lot of work to compute the bandwidth requirements, taking into account
DP SST vs. MST vs. UHBR and DSC and bit depth and everything, and you'll
still miss any driver/GPU specific contraints. Or are you just proposing
to match the physical characteristics only?

This is a requirement from Google, yes they want to compute the bandwidth for modes in the userspace side.
@Manasi could describe in detail how would you implement it?

Given all the connectors and displays and their modes on a system, you
could estimate a) for each display, what modes could be used on each
connector, and b) for each connector, what modes could be used on a
display, if connected. From the user POV, they could be shown why some
display mode is not available on the current connector, and crucially,
on which connector it could be available.

In general, I'm wondering if you're really after higher level
information, something like "connector max available bandwidth
with/without DSC", that would more directly map to the user needs as I
see them than low level link details.

Do you have the userspace somewhere? That's needed for merging UABI
anyway [1].

[1] https://docs.kernel.org/gpu/drm-uapi.html#open-source-userspace-requirements

I wanted to add it to IGT GPU tools, but I was hopping for community review on this RFC before the IGT development.

The new drm_dp_source_set_caps and drm_dp_sink_set_caps helper expose
source and sink training capabilities to userspace via sysfs attributes
under dp_link. Finally drm_dp_set_cur_link_params helper is provided to
report current link information at runtime.

Three drivers are updated as reference implementations: i915 (direct
connector path), ANX7625 and MediaTek (via the bridge connector framework
using a new DRM_BRIDGE_OP_DP flag).

Here is a sysfs example:

cur_link_info_dsc sink_link_caps_nlanes
cur_link_info_link_rate source_link_caps_dsc
cur_link_info_nlanes source_link_caps_link_rates
sink_link_caps_dsc source_link_caps_nlanes
sink_link_caps_link_rates

1620000 2160000 2700000 3240000 4320000 5400000

1620000 2700000 5400000 10000000 20000000

5400000

This series is rebased on top of drm-misc and may conflict with the
current state of the drm i915 tree. Which tree should pick this up,
given that it mixes core development with driver changes?

For development, please just base your work on top of drm-tip, and the
tree(s) will be sorted out later.

Oh, ok, thanks you.

--
Köry Maincent, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com