Re: [PATCH] spi: omap2-mcspi: handle DMA setup failures

[Sahaj Chaudhari]

On Tue, Oct 13, 2026 at 04:40:00PM +0100, Mark Brown wrote:
> Do we need to turn off the FIFO here before doing PIO?

Yes, good catch. omap2_mcspi_set_fifo(spi, t, 1) runs before the DMA
attempt, and omap2_mcspi_txrx_pio() polls per-word TXS/RXS without
expecting FIFO mode. v2 disables the FIFO, with the channel off while
its bits change, before falling back.

I injected DMA setup failures on an AM642 with and without this change,
and the PIO fallback completed with no errors or timeouts in both cases.
The bus had no device attached, so that test does not show the FIFO
problem. The change is based on reading the code and not on a
reproduced failure. I have said so in the v2 commit message.

v2: https://lore.kernel.org/r/20261014045617.2894886-1-sahaj123.sc@xxxxxxxxx