ACK: [SRU][R][PATCH 0/1] i915 WARN_ON call trace during CB/WB on MTL/ARL platforms

[Yufeng Gao]

On 8/4/26 10:20, AceLan Kao wrote:
> From: "Chia-Lin Kao (AceLan)" <acelan.kao at canonical.com>
>
> BugLink: https://bugs.launchpad.net/bugs/2144537
>
> [Impact]
> On Dell systems (CID: 202602-38397) with Intel XeLPDP/Arrow Lake-S display
> (8086:7d67), the i915 driver produces a kernel WARNING during cold boot and
> warm boot when no monitor is connected. The BIOS can leave TypeC HPD live
> status bits set on an XeLPDP TC port indicating a connected device, while
> TCSS (TypeC Subsystem) power has not been enabled. During driver probe,
> this causes drm_WARN_ON to fire in xelpdp_tc_phy_get_hw_state() even
> though the driver correctly handles this state afterwards.
> The warning triggers failures in automated cold-boot/warm-boot stress test
> plans (cold-boot-loop-test, warm-boot-loop-test).
> Error log:
> i915 0000:00:02.0: [drm] drm_WARN_ON((tc->mode == TC_PORT_DP_ALT || tc->mode == TC_PORT_LEGACY) && !xelpdp_tc_phy_tcss_power_is_enabled(tc))
> WARNING: CPU: 7 PID: 591 at drivers/gpu/drm/i915/display/intel_tc.c:1177 xelpdp_tc_phy_get_hw_state+0x145/0x150 [i915]
> Call Trace:
>   intel_tc_port_init_mode+0x73/0x270 [i915]
>   intel_tc_port_init+0x1e0/0x2a0 [i915]
>   intel_ddi_init+0xa9a/0x1220 [i915]
>   intel_setup_outputs+0x1f4/0xc20 [i915]
>   intel_display_driver_probe_nogem+0x16f/0x270 [i915]
>   i915_driver_probe+0x251/0x670 [i915]
>
> [Fix]
> Convert the drm_WARN_ON in xelpdp_tc_phy_get_hw_state() to a drm_dbg_kms()
> message. The stale HPD bits left by BIOS are a known handoff condition that
> the driver already handles correctly via intel_tc_port_update_mode(); the
> WARN_ON is a false positive that need not be a warning.
> This is consistent with the approach taken for the analogous AUX power check
> in commit 5830231e6547 ("drm/i915/icl+/tc: Convert AUX powered WARN to a
> debug message"), which is already in the tree.
>
> https://lore.kernel.org/lkml/20260316093001.689992-1-acelan.kao@canonical.com/T/#u
>
> [Test Plan]
> Boot the system without a monitor connected and check dmesg:
> $ sudo dmesg | grep -i "warn_on.*xelpdp\|WARNING.*intel_tc"
> Without the patch: WARNING at intel_tc.c:1177 xelpdp_tc_phy_get_hw_state
> is present in dmesg on affected hardware.
> With the patch: No WARNING or drm_WARN_ON in dmesg related to
> xelpdp_tc_phy_get_hw_state.
> To stress test, run multiple cold/warm boot cycles:
> $ sudo reboot # repeat 10+ times, check dmesg each boot
>
> [Where problems could occur]
> It may affect Intel XeLPDP TC port state detection in the i915 driver on
> MTL/ARL platforms.
> The WARN_ON being converted was asserting that TCSS power must be enabled
> when the port mode is DP-alt or Legacy. By downgrading it to a debug
> message, a genuinely unexpected hardware state (not just a BIOS handoff
> artifact) could go unnoticed and silently result in incorrect TC port
> mode initialisation. This could manifest as a TC/USB4 port failing to
> enumerate connected devices or DisplayPort-alt mode not working on
> affected TC ports after boot.
> The risk is limited to XeLPDP platforms (MTL, ARL) and only triggers
> during the specific BIOS handoff scenario where HPD bits are stale.
>
> Chia-Lin Kao (AceLan) (1):
>    UBUNTU: SAUCE: drm/i915/xelpdp/tc: Convert TCSS power check WARN to a
>      debug message
>
>   drivers/gpu/drm/i915/display/intel_tc.c | 7 ++++---
>   1 file changed, 4 insertions(+), 3 deletions(-)
>
Acked-by: Yufeng Gao <yufeng.gao at canonical.com>