Re: [PATCH 2/3] clk: imx8mq: add hdmi_phy_27m clock as pll's reference clock

[Lucas Stach]

Am Donnerstag, den 07.03.2019, 12:56 +0000 schrieb Anson Huang:
> Hi, Lucas
> 
> Best Regards!
> Anson Huang
> 
> > -----Original Message-----
> > > > From: Lucas Stach [mailto:l.stach@xxxxxxxxxxxxxx]
> > Sent: 2019å3æ7æ 20:06
> > > > > > To: Anson Huang <anson.huang@xxxxxxx>; shawnguo@xxxxxxxxxx;
> > > > s.hauer@xxxxxxxxxxxxxx; kernel@xxxxxxxxxxxxxx; Fabio Estevam
> > > > > > > > <fabio.estevam@xxxxxxx>; robh+dt@xxxxxxxxxx; mark.rutland@xxxxxxx;
> > > > mturquette@xxxxxxxxxxxx; sboyd@xxxxxxxxxx; Abel Vesa
> > > > > > <abel.vesa@xxxxxxx>; agx@xxxxxxxxxxx; linux-arm-
> > > > kernel@xxxxxxxxxxxxxxxxxxx; devicetree@xxxxxxxxxxxxxxx; linux-
> > kernel@xxxxxxxxxxxxxxx; linux-clk@xxxxxxxxxxxxxxx
> > > > Cc: dl-linux-imx <linux-imx@xxxxxxx>
> > Subject: Re: [PATCH 2/3] clk: imx8mq: add hdmi_phy_27m clock as pll's
> > reference clock
> > 
> > Am Donnerstag, den 07.03.2019, 03:41 +0000 schrieb Anson Huang:
> > > There is another 27MHz OSC inside i.MX8MQ's display block and it can
> > > be one of reference clocks of all PLLs, add it into clock tree and
> > > also add it as PLL's reference clock.
> > > 
> > > > > > Signed-off-by: Anson Huang <Anson.Huang@xxxxxxx>
> > > ---
> > > Âdrivers/clk/imx/clk-imx8mq.c | 3 ++-
> > > Â1 file changed, 2 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/drivers/clk/imx/clk-imx8mq.c
> > > b/drivers/clk/imx/clk-imx8mq.c index a9b3888..bb1bf9b 100644
> > > --- a/drivers/clk/imx/clk-imx8mq.c
> > > +++ b/drivers/clk/imx/clk-imx8mq.c
> > > @@ -26,7 +26,7 @@ static u32 share_count_nand;
> > > 
> > > Âstatic struct clk *clks[IMX8MQ_CLK_END];
> > > 
> > > -static const char * const pll_ref_sels[] = { "osc_25m", "osc_27m",
> > > "dummy", "dummy", };
> > > +static const char * const pll_ref_sels[] = { "osc_25m", "osc_27m",
> > > +"osc_hdmi_phy_27m", "dummy", };
> > > Âstatic const char * const arm_pll_bypass_sels[] = {"arm_pll",
> > > "arm_pll_ref_sel", };
> > > Âstatic const char * const gpu_pll_bypass_sels[] = {"gpu_pll",
> > > "gpu_pll_ref_sel", };
> > > Âstatic const char * const vpu_pll_bypass_sels[] = {"vpu_pll",
> > > "vpu_pll_ref_sel", }; @@ -281,6 +281,7 @@ static int
> > 
> > imx8mq_clocks_probe(struct platform_device *pdev)
> > > > > > Â	clks[IMX8MQ_CLK_32K] = of_clk_get_by_name(np, "ckil");
> > > > > > Â	clks[IMX8MQ_CLK_25M] = of_clk_get_by_name(np, "osc_25m");
> > > > > > Â	clks[IMX8MQ_CLK_27M] = of_clk_get_by_name(np, "osc_27m");
> > > > > > +	clks[IMX8MQ_CLK_HDMI_PHY_27M] = of_clk_get_by_name(np,
> > > +"osc_hdmi_phy_27m");
> > 
> > This is not acceptable. This adds a new required clock input, without
> > bothering to add the corresponding binding information or thinking about
> > backwards compatibility. At this point there are existing DTs out there, which
> > don't provide this required clock, which will cause a full boot regression. This
> > can only be an optional clock input at this point.
> 
> What do you think if we don't get such clock from DT? Just register this fixed clock
> in clock driver directly, then there will be no dependency of DT.

I don't think that's a good idea, as there can probably be use-cases
where having the XTAL unpopulated is actually a valid system
configuration, so not having this clock must be valid for the CCM
driver.

Regards,
Lucas