Re: [PATCH] cxl: Demote extended linear cache missing emissions

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

Dave Jiang wrote:
> Demote extended linear cache missing emissions to debug level as most
> systems do not have this feature enabled and the complaint about missing
> feature is spammy.
> 
> Suggeted-by: Dan Williams <dan.j.williams@intel.com>
> Signed-off-by: Dave Jiang <dave.jiang@intel.com>
> ---
>  drivers/cxl/acpi.c        | 6 +++---
>  drivers/cxl/core/region.c | 6 +++---
>  2 files changed, 6 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/cxl/acpi.c b/drivers/cxl/acpi.c
> index 712624cba2b6..e56b3d96cdef 100644
> --- a/drivers/cxl/acpi.c
> +++ b/drivers/cxl/acpi.c
> @@ -359,9 +359,9 @@ static int cxl_acpi_set_cache_size(struct cxl_root_decoder *cxlrd)
>  	 */
>  	size = size >> 1;
>  	if (cache_size && size != cache_size) {
> -		dev_warn(&cxld->dev,
> -			 "Extended Linear Cache size %pa != CXL size %pa. No Support!",
> -			 &cache_size, &size);
> +		dev_dbg(&cxld->dev,
> +			"Extended Linear Cache size %pa != CXL size %pa. No Support!",
> +			&cache_size, &size);

The log spam I was seeing was:

"cxl root0: Extended linear cache calculation failed rc:-2"

The concern is whether this mismatch should be reported or result in a
more violent fallback of disabling CXL regions altogether and fallback
to EFI_MEMORY_SP delineated dax devices. I think at any point where the
driver is confused by firmware the CXL subsystem should probably try to
fail safely into vanilla device-dax mode. This needs Smita's work, but
there is really nothing the user can do about a persistent boot warning
unless there is a chance they can get a fix from their BIOS vendor.