Re: [PATCH] cxl: Add hpa_to_spa to the root decoder callback operations

[Jonathan Cameron <Jonathan.Cameron@huawei.com>]

On Fri, 11 Jul 2025 11:41:19 -0700
Alison Schofield <alison.schofield@intel.com> wrote:

> On Fri, Jul 11, 2025 at 05:45:30PM +0100, Jonathan Cameron wrote:
> > On Wed, 9 Jul 2025 13:07:46 -0700
> > Dave Jiang <dave.jiang@intel.com> wrote:
> >   
> > > Remove cxl_hpa_to_spa_fn callback from cxl root decoder and add the
> > > callback function to cxl_decoder_ops. Refactor to group all the root
> > > decoder callbacks together.
> > > 
> > > Signed-off-by: Dave Jiang <dave.jiang@intel.com>  
> > 
> > Question inline about mixture of types of CFMWS entry.
> >   
> > > ---
> > >  drivers/cxl/acpi.c        |  9 +++++++--
> > >  drivers/cxl/core/port.c   |  2 +-
> > >  drivers/cxl/core/region.c | 11 ++++++++---
> > >  drivers/cxl/cxl.h         | 13 ++++++++-----
> > >  4 files changed, 24 insertions(+), 11 deletions(-)
> > > 
> > > diff --git a/drivers/cxl/acpi.c b/drivers/cxl/acpi.c
> > > index 48478bc406ee..9f7e2a49dc5a 100644
> > > --- a/drivers/cxl/acpi.c
> > > +++ b/drivers/cxl/acpi.c
> > > @@ -20,6 +20,10 @@ static const guid_t acpi_cxl_qtg_id_guid =
> > >  	GUID_INIT(0xF365F9A6, 0xA7DE, 0x4071,
> > >  		  0xA6, 0x6A, 0xB4, 0x0C, 0x0B, 0x4F, 0x8E, 0x52);
> > >  
> > > +static u64 cxl_default_hpa_to_spa(struct cxl_root_decoder *cxlrd, u64 hpa)
> > > +{
> > > +	return hpa;
> > > +}
> > >  
> > >  static u64 cxl_xor_hpa_to_spa(struct cxl_root_decoder *cxlrd, u64 hpa)
> > >  {
> > > @@ -342,8 +346,9 @@ static int cxl_acpi_get_extended_linear_cache_size(struct resource *backing_res,
> > >  	return hmat_get_extended_linear_cache_size(backing_res, nid, size);
> > >  }
> > >  
> > > -static const struct cxl_rd_ops acpi_rd_ops = {
> > > +static struct cxl_rd_ops acpi_rd_ops = {
> > >  	.get_extended_linear_cache_size = cxl_acpi_get_extended_linear_cache_size,
> > > +	.hpa_to_spa = cxl_default_hpa_to_spa,
> > >  };
> > >  
> > >  DEFINE_FREE(put_cxlrd, struct cxl_root_decoder *,
> > > @@ -425,7 +430,7 @@ static int __cxl_parse_cfmws(struct acpi_cedt_cfmws *cfmws,
> > >  	cxlrd->qos_class = cfmws->qtg_id;
> > >  
> > >  	if (cfmws->interleave_arithmetic == ACPI_CEDT_CFMWS_ARITHMETIC_XOR)
> > > -		cxlrd->hpa_to_spa = cxl_xor_hpa_to_spa;
> > > +		cxlrd->ops->hpa_to_spa = cxl_xor_hpa_to_spa;  
> > 
> > What happens if we have a mixture for different CMWS entries?
> > Some xor, some not.  
> 
> A separate root decoder is created for each CFMWS entry, so no
> conflict. Is that what you're wondering?

This isn't modifying a copy. It's modifying the global structure
that you make non const above (or I'm lacking caffeine.)


> 
> 
> >   
> > >  
> > >  	rc = cxl_decoder_add(cxld, target_map);
> > >  	if (rc)  
> >   
>