On 2/24/20 1:06 AM, Maxime Ripard wrote: > The HDMI blocks in the BCM2771 have an i2c controller to retrieve the > EDID. This block is split into two parts, the BSC and the AUTO_I2C, > lying in two separate register areas. > > The AUTO_I2C block has a mailbox-like interface and will take away the > BSC control from the CPU if enabled. However, the BSC is the actually > the same controller than the one supported by the brcmstb driver, and > the AUTO_I2C doesn't really bring any immediate benefit. > > Let's use the BSC then, but let's also tie the AUTO_I2C registers with a > separate compatible so that we can enable AUTO_I2C if needed in the > future. > > The AUTO_I2C is enabled by default at boot though, so we first need to > release the BSC from the AUTO_I2C control. > > Cc: Kamal Dasu <kdasu.kdev@xxxxxxxxx> > Cc: Florian Fainelli <f.fainelli@xxxxxxxxx> > Cc: Wolfram Sang <wsa@xxxxxxxxxxxxx> > Cc: bcm-kernel-feedback-list@xxxxxxxxxxxx > Cc: linux-i2c@xxxxxxxxxxxxxxx > Signed-off-by: Maxime Ripard <maxime@xxxxxxxxxx> [snip] > @@ -705,6 +737,7 @@ static SIMPLE_DEV_PM_OPS(brcmstb_i2c_pm, brcmstb_i2c_suspend, > static const struct of_device_id brcmstb_i2c_of_match[] = { > {.compatible = "brcm,brcmstb-i2c"}, > {.compatible = "brcm,brcmper-i2c"}, > + {.compatible = "brcm,bcm2711-hdmi-i2c"}, You could have added the bcm2711_release_bsc here as a function attached with the of_device_id::data member of the structure and do: if (data && data->init_func) rc = data->init_func(dev); But we can defer that until we have a second compatible string that requires the same approach. Akked-by: Florian Fainelli <f.fainelli@xxxxxxxxx> -- Florian