Re: [PATCH] spi: fix zero-length scatterlist entries for vmalloc buffers

[Mark Brown]

> Fix the entry count to match the way the loop consumes the buffer, and skip
> zero length transfers entirely.

> +	if (!len) {
> +		/*
> +		 * Zero-length transfers need no DMA mapping.  Leave an empty
> +		 * table behind so spi_unmap_buf_attrs() sees orig_nents == 0.
> +		 */
> +		memset(sgt, 0, sizeof(*sgt));
> +		return 0;
> +	}

This isn't needed - the rest of the patch changes the call to
sg_alloc_table() so it will fail, and returning success will confuse
things downstream.  The KUnit tests should also be failing with this,
they rely on a zero length transfer to force -EINVAL and get into error
handling.

-- 
Mark Brown <broonie@xxxxxxxxxx>