Re: [PATCH v4 10/14] cxl/region: Add function to find a port's switch decoder by range

[Dan Williams <dan.j.williams@intel.com>]

Robert Richter wrote:
> Factor out code to find the switch decoder of a port for a specific
> address range. Reuse the code to search a root decoder, create the
> function cxl_port_find_switch_decoder() and rework
> match_root_decoder_by_range() to be usable for switch decoders too.
> 
> Signed-off-by: Robert Richter <rrichter@amd.com>
> Reviewed-by: Gregory Price <gourry@gourry.net>
> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
> Reviewed-by: Dave Jiang <dave.jiang@intel.com>
> Tested-by: Gregory Price <gourry@gourry.net>
> ---
>  drivers/cxl/core/region.c | 48 +++++++++++++++++++++++----------------
>  1 file changed, 28 insertions(+), 20 deletions(-)
> 
> diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
> index 70ff4c94fb7a..cf58ee284696 100644
> --- a/drivers/cxl/core/region.c
> +++ b/drivers/cxl/core/region.c
> @@ -3198,33 +3198,48 @@ static int devm_cxl_add_dax_region(struct cxl_region *cxlr)
>  	return rc;
>  }
>  
> -static int match_root_decoder_by_range(struct device *dev,
> -				       const void *data)
> +static int match_decoder_by_range(struct device *dev, const void *data)
>  {
>  	const struct range *r1, *r2 = data;
> -	struct cxl_root_decoder *cxlrd;
> +	struct cxl_decoder *cxld;
>  
> -	if (!is_root_decoder(dev))
> +	if (!is_switch_decoder(dev))
>  		return 0;
>  
> -	cxlrd = to_cxl_root_decoder(dev);
> -	r1 = &cxlrd->cxlsd.cxld.hpa_range;
> +	cxld = to_cxl_decoder(dev);
> +	r1 = &cxld->hpa_range;
>  	return range_contains(r1, r2);
>  }
>  
> +static struct cxl_decoder *
> +cxl_port_find_switch_decoder(struct cxl_port *port, struct range *hpa)
> +{
> +	/*
> +	 * device_find_child() increments the reference count of the
> +	 * the switch decoder's parent port to protect the reference
> +	 * to its child. The port is already a parent of the endpoint
> +	 * decoder's port, at least indirectly in the port hierarchy.
> +	 * Thus, the endpoint already holds a reference for the parent
> +	 * port of the switch decoder. Free the unnecessary reference
> +	 * here.
> +	 */
> +	struct device *cxld_dev __free(put_device) =
> +		device_find_child(&port->dev, hpa, match_decoder_by_range);
> +
> +	return cxld_dev ? to_cxl_decoder(cxld_dev) : NULL;

After seeing this comment block repeated, I am less enthusiastic about
this approach.

I am worried about cases where something wants to potentially hold the
reference outside of the lifetime of the endpoint. I do not think we
necessarily have those today but there is now a mix of "find" helpers
that assume the caller will maintain an endpoint port reference and
subtle bugs looming if that assumption is violated.

Part of the reason for the "put_device early with a comment" approach
was to not need to deal with some of the awkwardness of dropping the
reference in all exit paths.

However, the scope-based cleanup helpers now automate that awkwardness.
So, I would much rather have a DEFINE_FREE(put_cxl_<object>) to pair
with each "find" operation and drop all these repeated "its ok to
violate typical reference expectations" comment blocks.

So, that's a nak for this approach from me.