Re: [PATCH v4 29/35] mtd: rawnand: docg4: convert driver to nand_scan()

[Boris Brezillon]

On Fri, 20 Jul 2018 17:15:21 +0200
Miquel Raynal <miquel.raynal@xxxxxxxxxxx> wrote:

> Two helpers have been added to the core to make ECC-related
> configuration between the detection phase and the final NAND scan. Use
> these hooks and convert the driver to just use nand_scan() instead of
> both nand_scan_ident() and nand_scan_tail().
> 
> Signed-off-by: Miquel Raynal <miquel.raynal@xxxxxxxxxxx>
> ---
>  drivers/mtd/nand/raw/docg4.c | 55 ++++++++++++++++++++++++++------------------
>  1 file changed, 32 insertions(+), 23 deletions(-)
> 
> diff --git a/drivers/mtd/nand/raw/docg4.c b/drivers/mtd/nand/raw/docg4.c
> index 4dccdfba6140..fec4353ff4ef 100644
> --- a/drivers/mtd/nand/raw/docg4.c
> +++ b/drivers/mtd/nand/raw/docg4.c
> @@ -1227,10 +1227,9 @@ static void __init init_mtd_structs(struct mtd_info *mtd)
>  	 * required within a nand driver because they are performed by the nand
>  	 * infrastructure code as part of nand_scan().  In this case they need
>  	 * to be initialized here because we skip call to nand_scan_ident() (the
> -	 * first half of nand_scan()).  The call to nand_scan_ident() is skipped
> -	 * because for this device the chip id is not read in the manner of a
> -	 * standard nand device.  Unfortunately, nand_scan_ident() does other
> -	 * things as well, such as call nand_set_defaults().
> +	 * first half of nand_scan()).  The call to nand_scan_ident() could be
> +	 * skipped because for this device the chip id is not read in the manner
> +	 * of a standard nand device.
>  	 */
>  
>  	struct nand_chip *nand = mtd_to_nand(mtd);
> @@ -1315,6 +1314,27 @@ static int __init read_id_reg(struct mtd_info *mtd)
>  
>  static char const *part_probes[] = { "cmdlinepart", "saftlpart", NULL };
>  
> +static int docg4_attach_chip(struct nand_chip *chip)
> +{
> +	struct mtd_info *mtd = nand_to_mtd(chip);
> +	struct docg4_priv *doc = (struct docg4_priv *)(chip + 1);
> +
> +	init_mtd_structs(mtd);
> +
> +	/* Initialize kernel BCH algorithm */
> +	doc->bch = init_bch(DOCG4_M, DOCG4_T, DOCG4_PRIMITIVE_POLY);
> +	if (!doc->bch)
> +		return -EINVAL;
> +

You need a ->detach_chip() hook to free the BCH context, don't you?

> +	reset(mtd);
> +
> +	return read_id_reg(mtd);
> +}
> +
> +static const struct nand_controller_ops docg4_controller_ops = {
> +	.attach_chip = docg4_attach_chip,
> +};
> +
>  static int __init probe_docg4(struct platform_device *pdev)
>  {
>  	struct mtd_info *mtd;
> @@ -1350,26 +1370,16 @@ static int __init probe_docg4(struct platform_device *pdev)
>  	mtd->dev.parent = &pdev->dev;
>  	doc->virtadr = virtadr;
>  	doc->dev = dev;
> -
> -	init_mtd_structs(mtd);
> -
> -	/* initialize kernel bch algorithm */
> -	doc->bch = init_bch(DOCG4_M, DOCG4_T, DOCG4_PRIMITIVE_POLY);
> -	if (doc->bch == NULL) {
> -		retval = -EINVAL;
> -		goto free_nand;
> -	}
> -
>  	platform_set_drvdata(pdev, doc);
>  
> -	reset(mtd);
> -	retval = read_id_reg(mtd);
> -	if (retval == -ENODEV) {
> -		dev_warn(dev, "No diskonchip G4 device found.\n");
> -		goto free_bch;
> -	}
> -
> -	retval = nand_scan_tail(mtd);
> +	/*
> +	 * Asking for 0 chips is useless here but it warns the user that the use
> +	 * of the nand_scan() function is a bit abused here because the
> +	 * initialization is actually a bit specific and re-handled again in the
> +	 * ->attach_chip() hook. It will probably leak some memory though.

I don't get the last part. Is there really a memory leak? In that case
we should find a solution to prevent that.

> +	 */
> +	nand->dummy_controller.ops = &docg4_controller_ops;
> +	retval = nand_scan(mtd, 0);
>  	if (retval)
>  		goto free_bch;
>  
> @@ -1389,7 +1399,6 @@ static int __init probe_docg4(struct platform_device *pdev)
>  	nand_cleanup(nand);
>  free_bch:
>  	free_bch(doc->bch);

This should be done in the ->detach_chip() hook.

> -free_nand:
>  	kfree(nand);
>  unmap:
>  	iounmap(virtadr);