When testing CEC on the AML-S905X-CC board I noticed that the CEC physical address was not invalidated when the HDMI cable was unplugged. Some more digging showed that meson uses meson_dw_hdmi.c to handle the HPD.
Both dw_hdmi_irq() and dw_hdmi_top_thread_irq() (in meson_dw_hdmi.c) call the dw_hdmi_setup_rx_sense() function. So move the code to invalidate the CEC physical address to that function, so that it is independent of where the HPD interrupt happens.
Tested with both a AML-S905X-CC and a Khadas VIM2 board.
Signed-off-by: Hans Verkuil hverkuil-cisco@xs4all.nl --- Note: an alternative would be to make a new dw-hdmi function such as dw_hdmi_cec_phys_addr_invalidate() that is called from meson_dw_hdmi.c. I decided not to do that since this patch is minimally invasive, but that can obviously be changed if that approach is preferred. --- diff --git a/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c b/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c index c5a854af54f8..e899b31e1432 100644 --- a/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c +++ b/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c @@ -2329,6 +2329,13 @@ void dw_hdmi_setup_rx_sense(struct dw_hdmi *hdmi, bool hpd, bool rx_sense) dw_hdmi_update_power(hdmi); dw_hdmi_update_phy_mask(hdmi); } + if (!hpd && !rx_sense) { + struct cec_notifier *notifier = READ_ONCE(hdmi->cec_notifier); + + if (notifier) + cec_notifier_phys_addr_invalidate(notifier); + } + mutex_unlock(&hdmi->mutex); } EXPORT_SYMBOL_GPL(dw_hdmi_setup_rx_sense); @@ -2369,14 +2376,6 @@ static irqreturn_t dw_hdmi_irq(int irq, void *dev_id) dw_hdmi_setup_rx_sense(hdmi, phy_stat & HDMI_PHY_HPD, phy_stat & HDMI_PHY_RX_SENSE); - - if ((phy_stat & (HDMI_PHY_RX_SENSE | HDMI_PHY_HPD)) == 0) { - struct cec_notifier *notifier; - - notifier = READ_ONCE(hdmi->cec_notifier); - if (notifier) - cec_notifier_phys_addr_invalidate(notifier); - } }
if (intr_stat & HDMI_IH_PHY_STAT0_HPD) {