On Mon, Oct 19, 2026 at 09:00:41PM +0800, Qianhui Xu wrote: > From: Xu Qianhui <[email protected]> > > dsi_7nm_phy_enable() pre-seeds pll_enable_cnt = 1 before calling > dsi_pll_enable_pll_bias(). That makes the bias helper take its > early-return path, so its PLL_SYSTEM_MUXES = 0xc0 write is skipped on the > first (bonded) bring-up and the register is left at 0x0. Panels wired > for bonded DSI then come up with a blank display. > > Write PLL_SYSTEM_MUXES in dsi_7nm_phy_enable() as well. The write in > dsi_pll_enable_pll_bias() is kept for the warm re-enable path where > pll_enable_cnt is 0. > > Fixes: 44784327815b ("drm/msm/dsi: phy: rework PLL bias enable/disable") > Signed-off-by: Xu Qianhui <[email protected]> > --- > drivers/gpu/drm/msm/dsi/phy/dsi_phy_7nm.c | 9 +++++++++ > 1 file changed, 9 insertions(+) > > diff --git a/drivers/gpu/drm/msm/dsi/phy/dsi_phy_7nm.c > b/drivers/gpu/drm/msm/dsi/phy/dsi_phy_7nm.c > index 1fc3b4f0c..5dd4830a1 100644 > --- a/drivers/gpu/drm/msm/dsi/phy/dsi_phy_7nm.c > +++ b/drivers/gpu/drm/msm/dsi/phy/dsi_phy_7nm.c > @@ -1144,6 +1144,15 @@ static int dsi_7nm_phy_enable(struct msm_dsi_phy *phy, > data = DSI_7nm_PHY_CMN_CTRL_0_DIGTOP_PWRDN_B | > DSI_7nm_PHY_CMN_CTRL_0_PLL_SHUTDOWNB; > writel(data, base + REG_DSI_7nm_PHY_CMN_CTRL_0); > + /* > + * Program the PLL output muxes here as well. The pll_enable_cnt > + * pre-seed above makes dsi_pll_enable_pll_bias() take its early-return > + * path, so on the first (bonded) bring-up PLL_SYSTEM_MUXES would > + * otherwise never be written and the panel would stay dark. The bias > + * function keeps its own write for the warm re-enable path > + * (pll_enable_cnt == 0). > + */
This belongs to the commit message rather than the code. > + writel(0xc0, pll->phy->pll_base + REG_DSI_7nm_PHY_PLL_SYSTEM_MUXES); > spin_unlock_irqrestore(&pll->pll_enable_lock, flags); > > /* Assert PLL core reset */ > -- > 2.53.0 > -- With best wishes Dmitry
