Re: [PATCH 2/2] cxl/acpi: Use the ACPI CFMWS to create static decoder objects

[Alison Schofield <alison.schofield@intel.com>]

On Tue, Jun 15, 2021 at 03:01:06PM -0700, Dan Williams wrote:
> On Tue, Jun 15, 2021 at 2:09 PM Alison Schofield
> <alison.schofield@intel.com> wrote:
> >
> > Thanks for the review Dan...
> >
> > On Tue, Jun 15, 2021 at 11:48:43AM -0700, Dan Williams wrote:
> > > [ add linu-acpi for variable length array question below ]
> > >
> > >
> > > On Mon, Jun 14, 2021 at 3:57 PM Alison Schofield
> > > <alison.schofield@intel.com> wrote:
> > > >
> > > > The ACPI CXL Early Discovery Table (CEDT) includes a list of CXL memory
> > > > resources in CXL Fixed Memory Window Structures (CFMWS). Retrieve each
> > > > CFMWS in the CEDT and add a cxl_decoder object to the root port (root0)
> > > > for each memory resource.
> > > >
> > > > Signed-off-by: Alison Schofield <alison.schofield@intel.com>
> > > > ---
> > > >  drivers/cxl/acpi.c | 106 +++++++++++++++++++++++++++++++++++++++++++++
> > > >  1 file changed, 106 insertions(+)
> > > >
> > > > diff --git a/drivers/cxl/acpi.c b/drivers/cxl/acpi.c
> > > > index 16f60bc6801f..ac4b3e37e294 100644
> > > > --- a/drivers/cxl/acpi.c
> > > > +++ b/drivers/cxl/acpi.c
> > > > @@ -8,8 +8,112 @@
> > > >  #include <linux/pci.h>
> > > >  #include "cxl.h"
> > > >
> > > > +/* Encode defined in CXL 2.0 8.2.5.12.7 HDM Decoder Control Register */
> > > > +#define CFMWS_INTERLEAVE_WAYS(x)       (1 << (x)->interleave_ways)
> > > > +#define CFMWS_INTERLEAVE_GRANULARITY(x)        ((x)->granularity + 8)
> > > > +
> > > > +/*
> > > > + * CFMWS Restrictions mapped to CXL Decoder Flags
> > > > + * Restrictions defined in CXL 2.0 ECN CEDT CFMWS
> > > > + * Decoder Flags defined in CXL 2.0 8.2.5.12.7 HDM Decoder Control Register
> > > > + */
> > > > +#define CFMWS_TO_DECODE_TYPE2(x) ((x & ACPI_CEDT_CFMWS_RESTRICT_TYPE2) << 2)
> > > > +#define CFMWS_TO_DECODE_TYPE3(x) ((x & ACPI_CEDT_CFMWS_RESTRICT_TYPE3) << 2)
> > > > +#define CFMWS_TO_DECODE_RAM(x)   ((x & ACPI_CEDT_CFMWS_RESTRICT_VOLATILE) >> 2)
> > > > +#define CFMWS_TO_DECODE_PMEM(x)         ((x & ACPI_CEDT_CFMWS_RESTRICT_PMEM) >> 2)
> > > > +#define CFMWS_TO_DECODE_FIXED(x) (x & ACPI_CEDT_CFMWS_RESTRICT_FIXED)
> > > > +
> > > > +#define CFMWS_TO_DECODER_FLAGS(x) (CFMWS_TO_DECODE_TYPE2(x) | \
> > > > +                                  CFMWS_TO_DECODE_TYPE3(x) | \
> > > > +                                  CFMWS_TO_DECODE_RAM(x)   | \
> > > > +                                  CFMWS_TO_DECODE_PMEM(x)  | \
> > > > +                                  CFMWS_TO_DECODE_FIXED(x))
> > >
> > > I don't understand the approach taken above. It seems to assume that
> > > the CXL_DECODER_F_* values are fixed. Those flag values are arbitrary
> > > and mutable. There is no guarantee that today's CXL_DECODER_F_* values
> > > match tomorrow's so I'd rather not have 2 places to check when / if
> > > that happens.
> > >
> >
> > Here's my next take - making the handling resilient.
> > Not so sure on gracefulness. Open for suggestions.
> >
> > -#define CFMWS_TO_DECODE_TYPE2(x) ((x & ACPI_CEDT_CFMWS_RESTRICT_TYPE2) << 2)
> > -#define CFMWS_TO_DECODE_TYPE3(x) ((x & ACPI_CEDT_CFMWS_RESTRICT_TYPE3) << 2)
> > -#define CFMWS_TO_DECODE_RAM(x)   ((x & ACPI_CEDT_CFMWS_RESTRICT_VOLATILE) >> 2)
> > -#define CFMWS_TO_DECODE_PMEM(x)         ((x & ACPI_CEDT_CFMWS_RESTRICT_PMEM) >> 2)
> > -#define CFMWS_TO_DECODE_FIXED(x) (x & ACPI_CEDT_CFMWS_RESTRICT_FIXED)
> > -
> > -#define CFMWS_TO_DECODER_FLAGS(x) (CFMWS_TO_DECODE_TYPE2(x) | \
> > -                                  CFMWS_TO_DECODE_TYPE3(x) | \
> > -                                  CFMWS_TO_DECODE_RAM(x)   | \
> > -                                  CFMWS_TO_DECODE_PMEM(x)  | \
> > -                                  CFMWS_TO_DECODE_FIXED(x))
> > +#define FLAG_TYPE2(x) \
> > +       ((x & ACPI_CEDT_CFMWS_RESTRICT_TYPE2) ? CXL_DECODER_F_TYPE2 : 0)
> > +#define FLAG_TYPE3(x) \
> > +       ((x & ACPI_CEDT_CFMWS_RESTRICT_TYPE3) ? CXL_DECODER_F_TYPE3 : 0)
> > +#define FLAG_RAM(x) \
> > +       ((x & ACPI_CEDT_CFMWS_RESTRICT_VOLATILE) ? CXL_DECODER_F_RAM : 0)
> > +#define FLAG_PMEM(x) \
> > +       ((x & ACPI_CEDT_CFMWS_RESTRICT_PMEM) ? CXL_DECODER_F_PMEM : 0)
> > +#define FLAG_FIXED(x) \
> > +       ((x & ACPI_CEDT_CFMWS_RESTRICT_FIXED) ? CXL_DECODER_F_LOCK : 0)
> > +
> > +#define CFMWS_TO_DECODER_FLAGS(x) (FLAG_TYPE2(x) | FLAG_TYPE3(x) | \
> > +                                  FLAG_RAM(x) | FLAG_PMEM(x)| FLAG_FIXED(x))
> 
> Hmm, why the macros? Just make CFMWS_TO_DECODER_FLAGS a proper function.
> 
> if (cfmws->restrictions & ACPI_CEDT_CFMWS_RESTRICT_TYPE2)
>     flags |= CXL_DECODER_F_TYPE2;
> 
> ...etc
> 
> Unless you foresee where macros were going to be reused somewhere else
> I would just as soon open code them like above.

Open code it is!
Thanks