Re: DP routing from iGPU to discrete Barlow Ridge TB5 on HP ZBook Fury G1i

[Mika Westerberg <mika.westerberg@linux.intel.com>]

On Mon, Apr 20, 2026 at 04:47:25PM -0400, Rodrigo Vivi wrote:
> On Mon, Apr 20, 2026 at 04:04:02PM +0000, Francesco Orro wrote:
> > Hi,
> > 
> > Setup:
> >   HP ZBook Fury G1i 16 inch, BIOS X96 01.03.04
> >   Intel Arrow Lake-HX + Nvidia RTX PRO 1000 Blackwell (hybrid graphics)
> >   Kernel 6.19.13
> > 
> > The laptop exposes Thunderbolt 5 on both the integrated MTL iTBT (right
> > USB-C, domain0, 00:0d.2) and a discrete Intel Barlow Ridge (left USB-C,
> > domain1, 00:06.1 -> 05:00.0).
> > 
> > - Right port: i915 drives a DP tunnel over the iTBT host without issues.
> > - Left port: no i915 connector ever sees the external monitor. The TB
> >   driver allocates a DP tunnel on demand but DPRX reads time out, no DP
> >   source is feeding the Barlow DP_IN adapter.
> > - With nvidia-drm.modeset=1, a DP-MST connector on the Nvidia DRM device
> >   sees the monitor via the left port (EDID readable, modes negotiated).
> >   The path Nvidia -> Barlow DP_IN is functional.
> > 
> > Current hypothesis: on this board the discrete Barlow Ridge DP_IN is
> > either hardwired to a Nvidia output or behind a firmware-programmed mux
> > that BIOS latches to Nvidia at POST and is not exposed to the OS. i915
> > would therefore have no path to drive the left ports, by design.
> 
> yes, that is usually the case.
> 
> > 
> > Tests done that did not surface an OS-visible lever:
> > - i2cdetect across all SMBus/i801 busses: no DP mux/retimer at plausible
> >   addresses (0x60-0x7F).
> > - DSDT/SSDT dump (iasl -d): no Device/Method referring to DP routing,
> >   DPMUX, retimer programming, or dock display source selection.
> > - HP WMI + /sys/devices/virtual/firmware-attributes/hp-bioscfg attribute
> >   enumeration: nothing switching dock display source. "Graphics"
> >   attribute only offers Hybrid/Discrete.
> > - PCI config dump of Barlow Ridge: standard USB4 NHI, no vendor-specific
> >   capability hinting at source selection.
> > - acpi.no_usb4_osc kernel param (locally patched) to opt out of OSC USB4
> >   native support: no change in i915 visibility of the dock.
> > 
> > Is there any known way to route an Intel DDI output to a PCIe-attached
> > Barlow Ridge DP_IN adapter, or is this strictly a design where the dGPU
> > is the only valid DP source for the discrete TB controller on this class
> > of HP workstations?
> 
> I believe this is physically connected or static mux without any way to
> re-route this. But I might be wrong.
> 

This is my understanding too based on the systems I've seen.