[PATCH] phy: rockchip: inno-hdmi: Change TMDS rate handling to configure() ops
Jonas Karlman
jonas at kwiboo.se
Thu May 7 11:03:39 PDT 2026
Hi Neil,
On 5/6/2026 5:00 PM, Neil Armstrong wrote:
> On 5/3/26 19:29, Jonas Karlman wrote:
>> The commit 10ed34d6eaaf ("phy: Add HDMI configuration options")
>> introduced a way for HDMI PHYs to be configured through the generic
>> phy_configure() function.
>>
>> This driver currently derives the TMDS character rate from the pixel
>> clock and the PHY bus width setting. However, no in-tree consumer of
>> this PHY has ever called phy_set_bus_width() to change the TMDS rate.
>>
>> Change the TMDS character rate handling to depend on the configure() ops
>> before any PHY consumer needs to configure a TMDS character rate that is
>> different from the pixel clock rate.
>>
>> Signed-off-by: Jonas Karlman <jonas at kwiboo.se>
>> ---
>> A near future drm/rockchip: dw_hdmi: series plans to include a call to
>> phy_configure() to configure this HDMI PHYs TMDS character rate.
>> ---
>> drivers/phy/rockchip/phy-rockchip-inno-hdmi.c | 30 ++++++++++---------
>> 1 file changed, 16 insertions(+), 14 deletions(-)
>>
>> diff --git a/drivers/phy/rockchip/phy-rockchip-inno-hdmi.c b/drivers/phy/rockchip/phy-rockchip-inno-hdmi.c
>> index 1483907413fa..7f0563d4d482 100644
>> --- a/drivers/phy/rockchip/phy-rockchip-inno-hdmi.c
>> +++ b/drivers/phy/rockchip/phy-rockchip-inno-hdmi.c
>> @@ -245,6 +245,7 @@ struct inno_hdmi_phy {
>> struct clk *phyclk;
>> unsigned long pixclock;
>> unsigned long tmdsclock;
>> + struct phy_configure_opts_hdmi hdmi_cfg;
>> };
>>
>> struct pre_pll_config {
>> @@ -554,19 +555,10 @@ static inline void inno_update_bits(struct inno_hdmi_phy *inno, u8 reg,
>> static unsigned long inno_hdmi_phy_get_tmdsclk(struct inno_hdmi_phy *inno,
>> unsigned long rate)
>> {
>> - int bus_width = phy_get_bus_width(inno->phy);
>> -
>> - switch (bus_width) {
>> - case 4:
>> - case 5:
>> - case 6:
>> - case 10:
>> - case 12:
>> - case 16:
>> - return (u64)rate * bus_width / 8;
>> - default:
>> - return rate;
>> - }
>> + if (inno->hdmi_cfg.tmds_char_rate)
>> + return inno->hdmi_cfg.tmds_char_rate;
>> +
>> + return rate;
>
> Can't you keep both until dw-hdmi calls the configure op ?
I probably could, just not sure how much use it will be. I could split
the changes in two different patches for a v2 if that is preferred?
The only known user that calls phy_set_bus_width() for this PHY are my
out-of-tree HDMI 2.0 patches for Rockchip RK3228/RK3328, i.e. those
originating from LibreELEC (also carried by other distros), and also the
downstream vendor kernel use a different implementation.
I am currently re-working those HDMI 2.0 patches, current working tree
can be found at [1], where the patch "drm/rockchip: dw_hdmi: configure
PHY in atomic_mode_set" [2] adds a call to phy_configure(). That part
of the multi-series effort should reach mailing list any day now.
[1] https://github.com/Kwiboo/linux-rockchip/commits/next-20260430-rk-hdmi-v2/
[2] https://github.com/Kwiboo/linux-rockchip/commit/555dfa562f40d22a63577c746ab42b0ec1f3ebee
Regards,
Jonas
>
>> }
>>
>> static irqreturn_t inno_hdmi_phy_rk3328_hardirq(int irq, void *dev_id)
>> @@ -602,6 +594,16 @@ static irqreturn_t inno_hdmi_phy_rk3328_irq(int irq, void *dev_id)
>> return IRQ_HANDLED;
>> }
>>
>> +static int inno_hdmi_phy_configure(struct phy *phy,
>> + union phy_configure_opts *opts)
>> +{
>> + struct inno_hdmi_phy *inno = phy_get_drvdata(phy);
>> +
>> + inno->hdmi_cfg = opts->hdmi;
>> +
>> + return 0;
>> +}
>> +
>> static int inno_hdmi_phy_power_on(struct phy *phy)
>> {
>> struct inno_hdmi_phy *inno = phy_get_drvdata(phy);
>> @@ -668,6 +670,7 @@ static int inno_hdmi_phy_power_off(struct phy *phy)
>>
>> static const struct phy_ops inno_hdmi_phy_ops = {
>> .owner = THIS_MODULE,
>> + .configure = inno_hdmi_phy_configure,
>> .power_on = inno_hdmi_phy_power_on,
>> .power_off = inno_hdmi_phy_power_off,
>> };
>> @@ -1392,7 +1395,6 @@ static int inno_hdmi_phy_probe(struct platform_device *pdev)
>> }
>>
>> phy_set_drvdata(inno->phy, inno);
>> - phy_set_bus_width(inno->phy, 8);
>>
>> if (inno->plat_data->ops->init) {
>> ret = inno->plat_data->ops->init(inno);
>
More information about the Linux-rockchip
mailing list