Hi,
On Mon, Jul 22, 2019 at 08:55:27PM +0200, Michał Mirosław wrote:
>
> Driver was not disabling TXRDY interrupt after last TX byte.
> This caused interrupt storm until transfer timeouts for slow
> or broken device on the bus. The patch fixes the interrupt storm
> on my SAMA5D2-based board.
>
> Cc: stable@xxxxxxxxxxxxxxx # 5.2.x
> [v5.2 introduced file split; the patch should apply to i2c-at91.c before the split]
> Fixes: fac368a0404 ("i2c: at91: add new driver")
> Signed-off-by: Michał Mirosław <mirq-linux@xxxxxxxxxxxx>
Acked-by: Ludovic Desroches <ludovic.desroches@xxxxxxxxxxxxx>
Sounds good to me as AT91_TWI_TXRDY is only used to send the next byte.
Raag, you added you in the copy list as it seems close to your issue. I am
really sorry, I didn't have time to go further with your issue. Is this
patch solves your issue?
Thanks for this patch.
Ludovic
> ---
> drivers/i2c/busses/i2c-at91-master.c | 9 +++++----
> 1 file changed, 5 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/i2c/busses/i2c-at91-master.c b/drivers/i2c/busses/i2c-at91-master.c
> index e87232f2e708..a3fcc35ffd3b 100644
> --- a/drivers/i2c/busses/i2c-at91-master.c
> +++ b/drivers/i2c/busses/i2c-at91-master.c
> @@ -122,9 +122,11 @@ static void at91_twi_write_next_byte(struct at91_twi_dev *dev)
> writeb_relaxed(*dev->buf, dev->base + AT91_TWI_THR);
>
> /* send stop when last byte has been written */
> - if (--dev->buf_len == 0)
> + if (--dev->buf_len == 0) {
> if (!dev->use_alt_cmd)
> at91_twi_write(dev, AT91_TWI_CR, AT91_TWI_STOP);
> + at91_twi_write(dev, AT91_TWI_IDR, AT91_TWI_TXRDY);
> + }
>
> dev_dbg(dev->dev, "wrote 0x%x, to go %zu\n", *dev->buf, dev->buf_len);
>
> @@ -542,9 +544,8 @@ static int at91_do_twi_transfer(struct at91_twi_dev *dev)
> } else {
> at91_twi_write_next_byte(dev);
> at91_twi_write(dev, AT91_TWI_IER,
> - AT91_TWI_TXCOMP |
> - AT91_TWI_NACK |
> - AT91_TWI_TXRDY);
> + AT91_TWI_TXCOMP | AT91_TWI_NACK |
> + (dev->buf_len ? AT91_TWI_TXRDY : 0));
> }
> }
>
> --
> 2.20.1
>
