Re: [PATCH 2/2] mtd: nand: fsl-ifc: fixup SRAM init for newer ctrl versions

[Kurt Kanzenbach]

Hi Miquel,

On Wed, Aug 08, 2018 at 11:48:32AM +0200, Miquel Raynal wrote:
> Hi Kurt,
>
> Subject prefix should be "mtd: rawnand: fsl_ifc:".

okay, noted.

>
> Kurt Kanzenbach <kurt@xxxxxxxxxxxxx> wrote on Mon,  6 Aug 2018 11:21:37
> +0200:
>
> > Newer versions of the IFC controller use a different method of initializing the
> > internal SRAM: Instead of reading from flash, a bit in the NAND configuration
> > register has to be set in order to trigger the self-initializing process.
> >
> > Signed-off-by: Kurt Kanzenbach <kurt@xxxxxxxxxxxxx>
> > ---
> >  drivers/mtd/nand/raw/fsl_ifc_nand.c | 18 ++++++++++++++++++
> >  include/linux/fsl_ifc.h             |  2 ++
> >  2 files changed, 20 insertions(+)
> >
> > diff --git a/drivers/mtd/nand/raw/fsl_ifc_nand.c b/drivers/mtd/nand/raw/fsl_ifc_nand.c
> > index e4f5792dc589..384d5e12b05c 100644
> > --- a/drivers/mtd/nand/raw/fsl_ifc_nand.c
> > +++ b/drivers/mtd/nand/raw/fsl_ifc_nand.c
> > @@ -30,6 +30,7 @@
> >  #include <linux/mtd/partitions.h>
> >  #include <linux/mtd/nand_ecc.h>
> >  #include <linux/fsl_ifc.h>
> > +#include <linux/iopoll.h>
> >
> >  #define ERR_BYTE		0xFF /* Value returned for read
> >  					bytes when read failed	*/
> > @@ -769,6 +770,23 @@ static int fsl_ifc_sram_init(struct fsl_ifc_mtd *priv)
> >  	uint32_t csor = 0, csor_8k = 0, csor_ext = 0;
> >  	uint32_t cs = priv->bank;
> >
> > +	if (ctrl->version > FSL_IFC_VERSION_1_1_0) {
>
> This is redundant and fsl_ifc_sram_init() is called only if
> "ctrl->version > FSL_FC_VERSION_1_1_0".

No, it's not. It's called when ctrl->version >=
FSL_IFC_VERSION_1_1_0. Therefore, this check is needed.

>
> So this means this function has never worked?

It did work for e.g. IFC controller in version 1.1.0.

However, it worked for the newer versions by accident, because U-Boot
already initialized the SRAM correctly. If you boot without NAND
initialization in U-Boot, then you'll hit the issue.

>
> If this is the case, there should be at least a Fixes: tag.
>
> Maybe it would be cleaner to always call fsl_ifc_sram_init() from the
> probe(), and just exit with a "return 0" here if the version is old?
> (I'll let you choose the way you prefer).

Sounds like a good idea. Otherwise we have to check the version twice.

>
> > +		u32 ncfgr, status;
> > +		int ret;
> > +
> > +		/* Trigger auto initialization */
> > +		ncfgr = ifc_in32(&ifc_runtime->ifc_nand.ncfgr);
> > +		ifc_out32(ncfgr | IFC_NAND_NCFGR_SRAM_INIT_EN, &ifc_runtime->ifc_nand.ncfgr);
> > +
> > +		/* Wait until done */
> > +		ret = readx_poll_timeout(ifc_in32, &ifc_runtime->ifc_nand.ncfgr,
> > +					 status, !(status & IFC_NAND_NCFGR_SRAM_INIT_EN),
> > +					 10, 1000);
>
> Nit: I always prefer when delays/timeouts are defined (and may be
> reused).

Me too. I've missed that there is already a timeout constant
IFC_TIMEOUT_MSECS (500). As it's huge, I'll add a second one.

>
> > +		if (ret)
> > +			dev_err(priv->dev, "Failed to initialize SRAM!\n");
>
> Space

okay.

Thanks,
Kurt

>
> > +		return ret;
> > +	}
> > +
> >  	/* Save CSOR and CSOR_ext */
> >  	csor = ifc_in32(&ifc_global->csor_cs[cs].csor);
> >  	csor_ext = ifc_in32(&ifc_global->csor_cs[cs].csor_ext);
> > diff --git a/include/linux/fsl_ifc.h b/include/linux/fsl_ifc.h
> > index 3fdfede2f0f3..5f343b796ad9 100644
> > --- a/include/linux/fsl_ifc.h
> > +++ b/include/linux/fsl_ifc.h
> > @@ -274,6 +274,8 @@
> >   */
> >  /* Auto Boot Mode */
> >  #define IFC_NAND_NCFGR_BOOT		0x80000000
> > +/* SRAM Initialization */
> > +#define IFC_NAND_NCFGR_SRAM_INIT_EN	0x20000000
> >  /* Addressing Mode-ROW0+n/COL0 */
> >  #define IFC_NAND_NCFGR_ADDR_MODE_RC0	0x00000000
> >  /* Addressing Mode-ROW0+n/COL0+n */
>
>
> Thanks,
> Miquèl