Quoting Clint Taylor (2024-10-24 19:31:09-03:00) >C10 phy timeouts occur on xe3lpd if the c10 bus is reset every >transaction. Starting with xe3lpd this is bus reset not necessary > Yeah. Resetting the bus on every transaction was never part of the Bspec and it was added as a workaround for issues observed with MTL, right Mika? I wonder if we also see the issues with Xe2_LPD... If not, maybe we could instead do the resets only for Xe_LPD+. In any case, since the reset was not part of the BSpec and was causing issues for Xe3_LPD: Reviewed-by: Gustavo Sousa <gustavo.sousa@xxxxxxxxx> We might want extend the commit message to provide more context (i.e. resets not required by the BSpec and were added as workaround for MTL). -- Gustavo Sousa >Signed-off-by: Clint Taylor <clinton.a.taylor@xxxxxxxxx> >--- > drivers/gpu/drm/i915/display/intel_cx0_phy.c | 6 ++++-- > 1 file changed, 4 insertions(+), 2 deletions(-) > >diff --git a/drivers/gpu/drm/i915/display/intel_cx0_phy.c b/drivers/gpu/drm/i915/display/intel_cx0_phy.c >index 4d6e1c135bdc..c6e0cbff5201 100644 >--- a/drivers/gpu/drm/i915/display/intel_cx0_phy.c >+++ b/drivers/gpu/drm/i915/display/intel_cx0_phy.c >@@ -224,7 +224,8 @@ static int __intel_cx0_read_once(struct intel_encoder *encoder, > * down and let the message bus to end up > * in a known state > */ >- intel_cx0_bus_reset(encoder, lane); >+ if (DISPLAY_VER(i915) < 30) >+ intel_cx0_bus_reset(encoder, lane); > > return REG_FIELD_GET(XELPDP_PORT_P2M_DATA_MASK, val); > } >@@ -313,7 +314,8 @@ static int __intel_cx0_write_once(struct intel_encoder *encoder, > * down and let the message bus to end up > * in a known state > */ >- intel_cx0_bus_reset(encoder, lane); >+ if (DISPLAY_VER(i915) < 30) >+ intel_cx0_bus_reset(encoder, lane); > > return 0; > } >-- >2.25.1 >
