On 8/27/24 11:58 AM, Maxime Ripard wrote:
> On Mon, Aug 19, 2024 at 01:29:29AM GMT, Cristian Ciocaltea wrote:
>> +static irqreturn_t dw_hdmi_qp_main_hardirq(int irq, void *dev_id)
>> +{
>> + struct dw_hdmi_qp *hdmi = dev_id;
>> + struct dw_hdmi_qp_i2c *i2c = hdmi->i2c;
>> + u32 stat;
>> +
>> + stat = dw_hdmi_qp_read(hdmi, MAINUNIT_1_INT_STATUS);
>> +
>> + i2c->stat = stat & (I2CM_OP_DONE_IRQ | I2CM_READ_REQUEST_IRQ |
>> + I2CM_NACK_RCVD_IRQ);
>> +
>> + if (i2c->stat) {
>> + dw_hdmi_qp_write(hdmi, i2c->stat, MAINUNIT_1_INT_CLEAR);
>> + complete(&i2c->cmp);
>> + }
>> +
>> + if (stat)
>> + return IRQ_HANDLED;
>> +
>> + return IRQ_NONE;
>> +}
>
> If the scrambler is enabled, you need to deal with hotplug. On hotplug,
> the monitor will drop its TMDS ratio and scrambling status, but the
> driver will keep assuming it's been programmed.
>
> If you don't have a way to deal with hotplug yet, then I'd suggest to
> just drop the scrambler setup for now.
Thanks for the heads up!
HPD is partially handled by the RK platform driver, which makes use of
drm_helper_hpd_irq_event(). Since the bridge sets DRM_BRIDGE_OP_DETECT
flag, the dw_hdmi_qp_bridge_detect() callback gets executed, which in turn
verifies the PHY status via ->read_hpd() implemented as
dw_hdmi_qp_rk3588_read_hpd() in the platform driver.
During my testing so far it worked reliably when switching displays with
different capabilities. I don't have a 4K@60Hz display at the moment, but
used the HDMI RX port on the Rock 5B board in a loopback connection to
verify this mode, which triggered the high TMDS clock ratio and scrambling
setup as well.
I just submitted v5 [1] where I reworked a bit the scrambling handling,
which allowed for some code simplification.
[1] https://lore.kernel.org/lkml/20240831-b4-rk3588-bridge-upstream-v5-0-9503bece0136@xxxxxxxxxxxxx/
Regards,
Cristian
