Antwort: Re: [PATCH 2/3] ARM: i.MX6ul: Add Clock support for i.MX6ull

[Sascha Hauer]

On Tue, May 16, 2017 at 04:46:00PM +0200, Stefan Riedmüller wrote:
> Sascha Hauer <s.hauer at pengutronix.de> wrote on 16/05/2017 07:47:39:
> 
> > Von: Sascha Hauer <s.hauer at pengutronix.de>
> > An: Stefan Riedmueller <s.riedmueller at phytec.de>
> > Kopie: barebox at lists.infradead.org
> > Datum: 16/05/2017 07:47
> > Betreff: Re: [PATCH 2/3] ARM: i.MX6ul: Add Clock support for i.MX6ull
> > 
> > On Thu, May 11, 2017 at 11:09:27AM +0200, Stefan Riedmueller wrote:
> > > From linux-4.10 clock support, only skipped some unnecessary clocks
> > > 
> > > Signed-off-by: Stefan Riedmueller <s.riedmueller at phytec.de>
> > > ---
> > >  drivers/clk/imx/clk-imx6ul.c | 73 +++++++++++++++++++++++++++++++
> > ++++++-------
> > >  1 file changed, 62 insertions(+), 11 deletions(-)
> > > 
> > > diff --git a/drivers/clk/imx/clk-imx6ul.c 
> b/drivers/clk/imx/clk-imx6ul.c
> > > index f28660d..b0a6bb0 100644
> > > --- a/drivers/clk/imx/clk-imx6ul.c
> > > +++ b/drivers/clk/imx/clk-imx6ul.c
> > > @@ -66,10 +66,24 @@ static const char *perclk_sels[] = { "ipg", "osc", 
> };
> > >  static const char *lcdif_sels[] = { "lcdif_podf", "ipp_di0", 
> > "ipp_di1", "ldb_di0", "ldb_di1", };
> > >  static const char *csi_sels[] = { "osc", "pll2_pfd2_396m", 
> > "pll3_120m", "pll3_pfd1_540m", };
> > >  static const char *sim_sels[] = { "sim_podf", "ipp_di0", 
> > "ipp_di1", "ldb_di0", "ldb_di1", };
> > > +/* epdc_pre_sels, epdc_sels, esai_sels only exists on i.MX6ULL */
> > > +static const char *epdc_pre_sels[] = { "pll2_bus", 
> > "pll3_usb_otg", "pll5_video_div", "pll2_pfd0_352m", 
> > "pll2_pfd2_396m", "pll3_pfd2_508m", };
> > > +static const char *esai_sels[] = { "pll4_audio_div", 
> > "pll3_pfd2_508m", "pll5_video_div", "pll3_usb_otg", };
> > > +static const char *epdc_sels[] = { "epdc_podf", "ipp_di0", 
> > "ipp_di1", "ldb_di0", "ldb_di1", };
> > > 
> > >  static struct clk *clks[IMX6UL_CLK_END];
> > >  static struct clk_onecell_data clk_data;
> > > 
> > > +static inline int clk_on_imx6ul(void)
> > > +{
> > > +   return of_machine_is_compatible("fsl,imx6ul");
> > > +}
> > > +
> > > +static inline int clk_on_imx6ull(void)
> > > +{
> > > +   return of_machine_is_compatible("fsl,imx6ull");
> > > +}
> > 
> > Can we use cpu_is_imx6ul(l) instead here? This would allow
> > us to compile away the code if we one of the two SoCs is
> > disabled in the config.
> > 
> > Sascha
> > 
> 
> Hi Sascha, 
> 
> I tried to use cpu_is_mx6ul, but it seems it is not valid yet in the clk 
> initcall. I found that __imx_cpu_type is set in a postcore initcall but 
> would be already evaluated in cpu_is_mx6ul in the clk initcall which is a 
> core initcall. So cpu_is_mx6ul would not evaluate correctly at this point. 
> 
> 
> I also checked this with an i.MX6 DualLight where the cpu_is_mx6dl is 
> implemented in the ccm init and there it evaluates wrong. In the ccm 
> initcall it evaluates to 0 but later after the postcore initcalls it 
> evaluates correctly to 1. So it seems the cpu_is_mx6ul/dl is not working 
> correctly in the clk initcall. Did I get this correct or did I miss 
> anything?

I looked at arch/arm/mach-imx/include/mach/imx6.h and saw that
__imx6_cpu_type() only reads from SoC registers, so assumed this would
work. What I didn't see is that the functions depend on cpu_is_mx6()
which goes down to __imx_cpu_type which you found out already.

Just like you I also found the cpu_is_mx6dl() in the i.MX6 clock driver,
but unlike you I just assumed that it works ;)

For now I think we can apply your original patch. I'll keep the issue on
my mental todo list.

Sascha


-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |