Hi Aisheng,
On Thu, Aug 8, 2019 at 8:02 AM Aisheng Dong <aisheng.dong@xxxxxxx> wrote:
> I think the problem of this patch is that it breaks the defer probe of DMA enablement
> which seems like a normal case in kernel. e.g. DMA driver probed later than I2C.
Please note that prior to commit e1ab9a468e3b ("i2c: imx: improve the
error handling in i2c_imx_dma_request()") there was no DMA defer probe
support at all, so the path I sent does not break defer probe. It just
restores the original behaviour.
> I wonder if there's a real customer requirement that wanting the I2C slave devices (e.g. RTC)
> to be probed early but using I2C with DMA and build DMA as module.
>
> If there is that requirement, maybe the correct way to fix this issue to explicitly disable I2C DMA
> in DT rather than simply fallback to polling mode for all the cases including the reasonable defer probe.
We should try to avoid fixing things by changing existing devicetrees.
I will submit a revert patch soon.
