[U-Boot] [PATCH] Fix FSL DDR clock adjust calculation.

[Joakim Tjernlund]

On Wed, 2015-09-02 at 09:34 -0500, York Sun wrote:
> 
> On 09/02/2015 09:31 AM, Joakim Tjernlund wrote:
> > On Wed, 2015-09-02 at 08:35 -0500, York Sun wrote:
> > > 
> > > On 09/02/2015 06:41 AM, Joakim Tjernlund wrote:
> > > > T1040 RM specifies CLK_ADJUST as 5 bits starting at bit pos 9
> > > > in DDR_DDR_SDRAM_CLK_CNTL, update code to match.
> > > > 
> > > > Signed-off-by: Joakim Tjernlund <joakim.tjernlund at transmode.se>
> > > > ---
> > > >  drivers/ddr/fsl/ctrl_regs.c | 2 +-
> > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > > 
> > > > diff --git a/drivers/ddr/fsl/ctrl_regs.c b/drivers/ddr/fsl/ctrl_regs.c
> > > > index 3919257..57077e1 100644
> > > > --- a/drivers/ddr/fsl/ctrl_regs.c
> > > > +++ b/drivers/ddr/fsl/ctrl_regs.c
> > > > @@ -1756,7 +1756,7 @@ static void set_ddr_sdram_clk_cntl(fsl_ddr_cfg_regs_t *ddr,
> > > >  	clk_adjust = popts->clk_adjust;
> > > >  	ddr->ddr_sdram_clk_cntl = (0
> > > >  				   | ((ss_en & 0x1) << 31)
> > > > -				   | ((clk_adjust & 0xF) << 23)
> > > > +				   | ((clk_adjust & 0x1F) << 22)
> > > >  				   );
> > > >  	debug("FSLDDR: clk_cntl = 0x%08x\n", ddr->ddr_sdram_clk_cntl);
> > > >  }
> > > > 
> > > 
> > > NACK. This is logically correct but it needs to consider all DDR controllers.
> > > Older controllers don't have the same bits for this field. So far we don't have
> > > to use this this fine granularity on clock adjustment.
> > 
> > I have no idea what older controllers do with these bits so I leave it to you to find out.
> > 
> > Anyhow, current code is broken as the programmed value is doubled(we see this on our board
> > which cannot boot linux without this fix)
> 
> Jocke,
> 
> Can you refer to Freescale board? You may be using correct clk_adjust value on
Custom board based on T1042RDB

> your board. But when you use this driver, you may have to adjust the value by
> shifting 1 bit. You will lose some values but we haven't found it difficult.
> This value is important but can also tolerate a wide range.

Sure, but how should anyone know that beforehand? The clk_adjust was calculated using
Freescales CW tool and added to table with the write lvl data.

Are the values for the T104x boards already compensated like you suggest?
Then they will (possibly) break when this bug is fixed.

 Jocke