Hi Christophe,
christophe.jaillet@xxxxxxxxxx wrote on Wed, 15 Feb 2023 12:08:45 +0100:
> wait_for_completion_timeout() and readl_poll_timeout() don't handle their
> return value the same way.
>
> wait_for_completion_timeout() returns 0 on time out (and >0 in all other
> cases)
> readl_poll_timeout() returns 0 on success and -ETIMEDOUT upon a timeout.
That's a shame, but yeah, excellent catch!
> In order for the error handling path to work in both cases, the logic
> against wait_for_completion_timeout() needs to be inverted.
>
> Fixes: 48e6633a9fa2 ("mtd: nand: mxic-ecc: Add Macronix external ECC engine support")
> Signed-off-by: Christophe JAILLET <christophe.jaillet@xxxxxxxxxx>
> ---
> Compile tested only.
>
> This is really spurious.
> If I'm right, this means that it never worked!
>
> Can any one with the hardware test?
The design I used for development and testing had no interrupt line
available for that IIRC, so I only tested the polling case ('else'
side) and completely overlooked that difference. I might have mentioned
it in the cover letter, if I didn't, it's an oversight.
> ---
> drivers/mtd/nand/ecc-mxic.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/drivers/mtd/nand/ecc-mxic.c b/drivers/mtd/nand/ecc-mxic.c
> index 8afdca731b87..6b487ffe2f2d 100644
> --- a/drivers/mtd/nand/ecc-mxic.c
> +++ b/drivers/mtd/nand/ecc-mxic.c
> @@ -429,6 +429,7 @@ static int mxic_ecc_data_xfer_wait_for_completion(struct mxic_ecc_engine *mxic)
> mxic_ecc_enable_int(mxic);
> ret = wait_for_completion_timeout(&mxic->complete,
> msecs_to_jiffies(1000));
> + ret = ret ? 0 : -ETIMEDOUT;
> mxic_ecc_disable_int(mxic);
> } else {
> ret = readl_poll_timeout(mxic->regs + INTRPT_STS, val,
Thanks,
Miquèl
