On Tue, 17 Nov 2020 at 14:15, Wolfram Sang
<wsa+renesas@xxxxxxxxxxxxxxxxxxxx> wrote:
>
> From: Masaharu Hayakawa <masaharu.hayakawa.ry@xxxxxxxxxxx>
>
> If some errors are detected at the same time as the access end
> interrupt, the access end interrupt was not cleared. Especially with
> DMA, because then the access end interrupt was never enabled and, thus,
> never cleared. Clear the interrupt register always when a command error
> occurs.
>
> Signed-off-by: Masaharu Hayakawa <masaharu.hayakawa.ry@xxxxxxxxxxx>
> [saito: rebase to v5.4]
> Signed-off-by: Takeshi Saito <takeshi.saito.xv@xxxxxxxxxxx>
> [wsa: rebase and extension of the commit message]
> Signed-off-by: Wolfram Sang <wsa+renesas@xxxxxxxxxxxxxxxxxxxx>
Applied for next, thanks!
Kind regards
Uffe
> ---
>
> Because the test case is hard to reproduce (lots of radio noise and
> temperature changes), I trust the BSP team here. But the reasoning makes
> a lot of sense to me. I verified that there are no regressions with the
> test cases I usually do. The patch is based on mmc/next as of today,
> i.e. with all previous series included.
>
> drivers/mmc/host/tmio_mmc_core.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/mmc/host/tmio_mmc_core.c b/drivers/mmc/host/tmio_mmc_core.c
> index cb4149fd12e0..7f4a28125010 100644
> --- a/drivers/mmc/host/tmio_mmc_core.c
> +++ b/drivers/mmc/host/tmio_mmc_core.c
> @@ -796,8 +796,10 @@ static void tmio_mmc_finish_request(struct tmio_mmc_host *host)
>
> spin_unlock_irqrestore(&host->lock, flags);
>
> - if (mrq->cmd->error || (mrq->data && mrq->data->error))
> + if (mrq->cmd->error || (mrq->data && mrq->data->error)) {
> + tmio_mmc_ack_mmc_irqs(host, TMIO_MASK_IRQ); /* Clear all */
> tmio_mmc_abort_dma(host);
> + }
>
> /* Error means retune, but executed command was still successful */
> if (host->check_retune && host->check_retune(host))
> --
> 2.28.0
>
