Re: [PATCH v2 2/4] i2c: imx: Fix race condition in dma read

From: Esben Haabendal
Date: Thu Aug 09 2018 - 07:57:16 EST


Uwe Kleine-KÃnig <u.kleine-koenig@xxxxxxxxxxxxxx> writes:

> On Mon, Jul 09, 2018 at 11:43:02AM +0200, Esben Haabendal wrote:
>> From: Esben Haabendal <eha@xxxxxxxx>
>>
>> This fixes a race condition, where the DMAEN bit ends up being set after
>> I2C slave has transmitted a byte following the dummy read. When that
>> happens, an interrupt is generated instead, and no DMA request is generated
>> to kickstart the DMA read, and a timeout happens after DMA_TIMEOUT (1 sec).
>>
>> Fixed by setting the DMAEN bit before the dummy read.
>>
>> Signed-off-by: Esben Haabendal <eha@xxxxxxxx>
>> ---
>> drivers/i2c/busses/i2c-imx.c | 8 +++++---
>> 1 file changed, 5 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/i2c/busses/i2c-imx.c b/drivers/i2c/busses/i2c-imx.c
>> index 39cfd98c7b23..d86f152176a4 100644
>> --- a/drivers/i2c/busses/i2c-imx.c
>> +++ b/drivers/i2c/busses/i2c-imx.c
>> @@ -668,9 +668,6 @@ static int i2c_imx_dma_read(struct imx_i2c_struct *i2c_imx,
>> struct imx_i2c_dma *dma = i2c_imx->dma;
>> struct device *dev = &i2c_imx->adapter.dev;
>>
>> - temp = imx_i2c_read_reg(i2c_imx, IMX_I2C_I2CR);
>> - temp |= I2CR_DMAEN;
>> - imx_i2c_write_reg(temp, i2c_imx, IMX_I2C_I2CR);
>>
>> dma->chan_using = dma->chan_rx;
>> dma->dma_transfer_dir = DMA_DEV_TO_MEM;
>> @@ -810,6 +807,11 @@ static int i2c_imx_read(struct imx_i2c_struct *i2c_imx, struct i2c_msg *msgs, bo
>> if ((msgs->len - 1) || block_data)
>> temp &= ~I2CR_TXAK;
>> imx_i2c_write_reg(temp, i2c_imx, IMX_I2C_I2CR);
>> + if (i2c_imx->dma && msgs->len >= DMA_THRESHOLD && !block_data) {
>> + temp = imx_i2c_read_reg(i2c_imx, IMX_I2C_I2CR);
>> + temp |= I2CR_DMAEN;
>> + imx_i2c_write_reg(temp, i2c_imx, IMX_I2C_I2CR);
>> + }
>
> Does this need to be a separate write to the I2CR register? Just before
> the if there is temp written to this register, so probably this can be
> changed to just:
>
> if (i2c_imx->dma && msgs->len >= DMA_THRESHOLD && !block_data)
> temp |= I2CR_DMAEN;
>
> when moved before up one line.

Sure, that is clearly better.

> I don't find documentation for the LS processors where this
> register is described though (and the imx family doesn't seem to support
> DMA for i2c).

Yes, unfortunately, I think the LS1021A reference manual requires NDA.

> Other than that this looks reasonable and warrants a
>
> Fixes: ce1a78840ff7 ("i2c: imx: add DMA support for freescale i2c driver")

I will add that.

/Esben