Re: [RFC] PCD: Extended SKU support 1 - inheritance

["Kinney, Michael D" <michael.d.kinney@intel.com>]

Hi Tim,

How would this relationship information be used in a module?
Do you have an example use case where this relationship
information could be used from the selected SKU?  Maybe
there is a different method to solve the same problem?

Please also review the 2nd SKU related RFC titled:

	[RFC 0/2] PCD: Extended SKU support 2 - sub SKU

This 2nd RFC builds on the first SKU RFC.  The SKU inheritance
relationships are not as straight forward when you add the 
concept of generating all possible combinations of SKUs from
multiple subsystems.

Thanks,

Mike

> -----Original Message-----
> From: edk2-devel [mailto:edk2-devel-bounces@lists.01.org] On Behalf Of Tim Lewis
> Sent: Wednesday, April 26, 2017 12:59 PM
> To: Kinney, Michael D <michael.d.kinney@intel.com>; Zeng, Star
> <star.zeng@intel.com>; edk2-devel@lists.01.org
> Cc: Gao, Liming <liming.gao@intel.com>
> Subject: Re: [edk2] [RFC] PCD: Extended SKU support 1 - inheritance
> 
> Mike --
> 
> So your use case for SetSku() is also what we've used.
> 
> Generating a simple table that describes the default relationships between SKUs
> is helpful:
> 
> DEFAULT
> A DEFAULT
> B A DEFAULT
> 
> This could be a #define in the form of a C UINT32 structure initializer:
> 
> #define SKUID_DEFAULT_ORDER 0,1,0,2,1,0
> 
> I think that this should be a part of this RFC.
> 
> -------------
> Having said that, we've used multi-SKU for a long time and there are common error
> cases that could be aided by some tool support.
> 
> 1) When an image is designed to support SKUs A, B, C (out of a possible set of
> SKU A-H), but the SKU detected is: G. This is an error condition that is
> difficult to detect. SetSku() can't do it, because there is no guarantee (today)
> that the PCD driver knows all possible SKUs, only those that have PCDs associated
> with them.
> 
> 2) When C code is designed to run for SKUs A, B, D, H, the image was built for
> A,B,C and the SKU detected during POST was D. By having a macro that returns
> whether board X is supported, code which is not used for the current build can be
> eliminated from the build. For example. #define SKU_NAME_SUPPORTED(sku-
> identifier) can return FALSE if the SKU associated with sku-identifier is not in
> the build.  Also useful for ASL and VFR.
> 
> 3) We have also found it useful to extend the build.exe command-line to allow
> multiple instances of -x to correctly populate the SKUID_IDENTIFIER.
> 
> These might be the subject for additional RFCs.
> 
> Tim
> 
> -----Original Message-----
> From: Kinney, Michael D [mailto:michael.d.kinney@intel.com]
> Sent: Wednesday, April 26, 2017 12:04 PM
> To: Tim Lewis <tim.lewis@insyde.com>; Zeng, Star <star.zeng@intel.com>; edk2-
> devel@lists.01.org; Kinney, Michael D <michael.d.kinney@intel.com>
> Cc: Gao, Liming <liming.gao@intel.com>
> Subject: RE: [RFC] PCD: Extended SKU support 1 - inheritance
> 
> Hi Tim,
> 
> If multiple SKUs are present in a FW image, in most cases A platform specific
> PEIM detects what SKU is present and calls SetSku() for the detected SKU.  The
> detection mechanism is typically very HW specific and may involve checking
> straps, GPIOs, or reading ID values from registers.
> 
> Before the SKU is set, PCD services access values from the DEFAULT SKU (SKU ID
> 0).  After SetSku() is called, the PCD services access values for that specific
> SKU.
> 
> What would be the runtime use case for being able to lookup the relationships
> between SKUs if in a single boot only the DEFAULT SKU and the one SKU passed to
> SetSku() are involved?
> 
> What additions do you suggest to this RFC or to another RFC that builds on top of
> this RFC?
> 
> Thanks,
> 
> Mike
> 
> > -----Original Message-----
> > From: edk2-devel [mailto:edk2-devel-bounces@lists.01.org] On Behalf Of
> > Tim Lewis
> > Sent: Wednesday, April 26, 2017 11:46 AM
> > To: Kinney, Michael D <michael.d.kinney@intel.com>; Zeng, Star
> > <star.zeng@intel.com>; edk2-devel@lists.01.org
> > Cc: Gao, Liming <liming.gao@intel.com>
> > Subject: Re: [edk2] [RFC] PCD: Extended SKU support 1 - inheritance
> >
> > Mike --
> >
> > I am saying that creating a relationship between SKUs that cannot be
> > determined at runtime is a confusing thing to add to the DSC.
> >
> > Tim
> >
> > -----Original Message-----
> > From: Kinney, Michael D [mailto:michael.d.kinney@intel.com]
> > Sent: Wednesday, April 26, 2017 11:40 AM
> > To: Tim Lewis <tim.lewis@insyde.com>; Zeng, Star
> > <star.zeng@intel.com>; edk2- devel@lists.01.org; Kinney, Michael D
> > <michael.d.kinney@intel.com>
> > Cc: Gao, Liming <liming.gao@intel.com>
> > Subject: RE: [RFC] PCD: Extended SKU support 1 - inheritance
> >
> > Tim,
> >
> > This RFC does not change the DEFAULT behavior. If a PCD value override
> > is not provided in a SKU, then the DEFAULT value is used. If this is
> > not how you interpret the RFC, then please provide some clarifications
> > for the next revision of this RFC.
> >
> > This RFC is about reducing the number of PCD statements in a DSC file
> > if there are multiple SKUs and if there is a relationship between SKUs
> > that can be used to support that reduction.  It is an optional new
> > feature.  You do not have to use it.
> >
> > I consider the optimization of the PCD database from a size and
> > performance perspective a separate topic.  The implementation of this
> > RFC does not require any changes to the PCD database format or the way
> > PCD value lookup algorithm in the PCD drivers.
> >
> > We do recognize that if a large number of SKUs with close
> > relationships are present in a single FW build, there may be more
> > optimal ways to store the data and more efficient ways to access the
> > data.  But we would like to consider any work in this area as a
> > separate work item.  Please feel free to add a Bugzilla feature
> > request if you have specific ideas on how to reduce size and/or improve
> performance.
> >
> > Given that this RFC is only about a more efficient DSC file implementation.
> > Can you please describe your concerns with SKUs as they are currently
> > documented in the EDK II specifications? If you have additional
> > feature requests for the EDK II SKU behavior, then I look forward to seeing
> your detailed proposals.
> >
> > Thanks,
> >
> > Mike
> >
> > > -----Original Message-----
> > > From: edk2-devel [mailto:edk2-devel-bounces@lists.01.org] On Behalf
> > > Of Tim Lewis
> > > Sent: Wednesday, April 26, 2017 11:24 AM
> > > To: Kinney, Michael D <michael.d.kinney@intel.com>; Zeng, Star
> > > <star.zeng@intel.com>; edk2-devel@lists.01.org
> > > Cc: Gao, Liming <liming.gao@intel.com>
> > > Subject: Re: [edk2] [RFC] PCD: Extended SKU support 1 - inheritance
> > >
> > > I understand the purpose. Today there is an assumption that IF not
> > > SKU x, then use SKU DEFAULT. This assumption is used (by us) for
> > > many purposes other than the PCD database.
> > >
> > > I'm confused by your statement that "This RFC has no impact to the
> > > PCD database or behavior of PCD services." Currently, as I recall,
> > > there is an optimization that says: only the differences between SKU
> > > A/SKU B and SKU DEFAULT must exist in the PCD database (SKU A and
> > > SKU B are sparsely encoded). Are you saying that this proposed
> > > change will not, in fact, optimize the PCD database for SKU B (that
> > > it will have the full set of values that differ from SKU DEFAULT?) If so, it
> is a pity.
> > >
> > > My other concern is that, we use the defaulting assumption for many
> > > other
> > items.
> > > For example: logos. Leaving it as a build-time construct only
> > > prevents other components from taking advantage of this knowledge.
> > >
> > > Tim
> > >
> > > On a separate, but related topic, the term "SKU" has a broad
> > > connotation, and we use it for many other items board-specific at
> > > build-time
> > and run-time (ie.
> > > Logos). Leaving the defaulting relationship as a build-time
> > > construct prevents usage at runtime. This is similar run-time
> > > problem to the case when (a) there are
> > > 10 SKUs listed in [SkuIds] but (b) 3 are listed in SKUID_IDENTIFIERS
> > > for this particular build. How, at runtime, does C code pick up that
> > > SkuA is supported in the current build?
> > >
> > >
> > >
> > >
> > >
> > > -----Original Message-----
> > > From: Kinney, Michael D [mailto:michael.d.kinney@intel.com]
> > > Sent: Wednesday, April 26, 2017 11:05 AM
> > > To: Tim Lewis <tim.lewis@insyde.com>; Zeng, Star
> > > <star.zeng@intel.com>; edk2- devel@lists.01.org; Kinney, Michael D
> > > <michael.d.kinney@intel.com>
> > > Cc: Gao, Liming <liming.gao@intel.com>
> > > Subject: RE: [RFC] PCD: Extended SKU support 1 - inheritance
> > >
> > > Tim,
> > >
> > > Can you please provide more details on your specific concerns along
> > > with some use cases?
> > >
> > > This RFC does not change the SKU feature in EDK II.  This RFC simply
> > > provides a more efficient way for a developer to provide PCD values
> > > for different SKUs with fewer PCD statements in a DSC file.
> > >
> > > Today, every SKU is a based on the DEFAULT SKU and if a PCD requires
> > > a different value than the DEFAULT SKU value, then a SKU specific
> > > PCD statement
> > is required.
> > >
> > > With this proposal, a relationship between SKUs can be declared in
> > > the DSC file, so SKUs can inherit PCD values from other SKUs.  This
> > > inheritance is limited to the scope of the DSC file.  This RFC has
> > > no impact to the PCD database or behavior of PCD services.
> > >
> > > The example Star provides at the end shows the equivalent set of PCD
> > > settings using the current EDK II syntax and with this RFC's
> > > backwards compatible syntax extensions.
> > >
> > > One way to think about this RFC is that a DSC file using the syntax
> > > in this RFC can be converted to a DSC file without the RFC syntax.
> > > In fact, that is what the implementation of this RFC would do to
> > > build the set of
> > PCD values for each SKU.
> > >
> > > Thanks,
> > >
> > > Mike
> > >
> > >
> > >
> > > > -----Original Message-----
> > > > From: edk2-devel [mailto:edk2-devel-bounces@lists.01.org] On
> > > > Behalf Of Tim Lewis
> > > > Sent: Wednesday, April 26, 2017 9:21 AM
> > > > To: Zeng, Star <star.zeng@intel.com>; edk2-devel@lists.01.org
> > > > Cc: Kinney, Michael D <michael.d.kinney@intel.com>; Gao, Liming
> > > > <liming.gao@intel.com>
> > > > Subject: Re: [edk2] [RFC] PCD: Extended SKU support 1 -
> > > > inheritance
> > > >
> > > > Star --
> > > >
> > > > This assumes that the SKU ID is only used for PCDs, which is not
> > > > the case. The SKU ID may be used by other components, and other
> > > > components may use the
> > > > 0|DEFAULT rule as well.
> > > >
> > > > 1) There is no way to read this defaulting rule at runtime. The
> > > > information is buried in the PCD database, but not available externally.
> > > > 2) There is no way to read the contents of SKUID_IDENTIFIER when
> > > > multiple SKUs are specified. While more than one SKU can be
> > > > specified (separated by spaces), there is no way to pass multiple
> > > > values into build.exe today, nor is there any way to identify
> > > > which SKUs (out of the total list of SKUs) is supported on one build at
> runtime.
> > > >
> > > > Tim
> > > >
> > > >
> > > > -----Original Message-----
> > > > From: edk2-devel [mailto:edk2-devel-bounces@lists.01.org] On
> > > > Behalf Of Zeng, Star
> > > > Sent: Tuesday, April 25, 2017 5:40 AM
> > > > To: edk2-devel@lists.01.org
> > > > Cc: Kinney, Michael D <michael.d.kinney@intel.com>; Zeng, Star
> > > > <star.zeng@intel.com>; Gao, Liming <liming.gao@intel.com>
> > > > Subject: [edk2] [RFC] PCD: Extended SKU support 1 - inheritance
> > > >
> > > > - Requirement
> > > > Simplify the PCDs configuring for multiple SKUs in DSC.
> > > >
> > > >
> > > > - Current limitation
> > > > Non-DEFAULT SKU could only derive from DEFAULT SKU, but could not
> > > > derive from another non-DEFAULT SKU.
> > > > For example below, SkuA and SkuB could only derive from DEFAULT,
> > > > but SkuB could not derive from SkuA.
> > > >
> > > > [SkuIds]
> > > >   0 | DEFAULT
> > > >   1 | SkuA
> > > >   2 | SkuB
> > > >
> > > >
> > > > - Proposal: One non-DEFAULT SKU could be a derivative of another
> > > > non-DEFAULT
> > > SKU.
> > > > This proposal only extends DSC [SkuIds] section syntax and the
> > > > extension is optional.
> > > > This proposal keeps the backward compatibility with current SKU usage.
> > > > BaseTools update is needed to support the syntax extension, and no
> > > > any change in PCD database and driver is required.
> > > >
> > > > DSC syntax:
> > > >   [SkuIds]
> > > >     SkuValue|SkuName[|ParentSkuName]
> > > >       SkuValue: integer, 0 is reserved for DEFAULT SKU.
> > > >       SkuName: string
> > > >       ParentSkuName: string, optional, it is new introduced in
> > > > this proposal and defines which SKU the PCD value will derive from
> > > > for this SKU. The PCD value will derive from DEFAULT SKU for this
> > > > SKU if the
> > > ParentSkuName is absent.
> > > >
> > > >
> > > > - Example: SkuB is a derivative of SkuA, but not a derivative of DEFAULT.
> > > >
> > > >   [SkuIds]
> > > >     0 | DEFAULT
> > > >     1 | SkuA
> > > >     2 | SkuB | SkuA
> > > >
> > > >   [PcdsDynamicDefault.Common.DEFAULT]
> > > >     gXXXPkgTokenSpaceGuid.PcdXXXSignature|"DEFAULT"
> > > >     gXXXPkgTokenSpaceGuid.PcdXXXConfig1|FALSE
> > > >     gXXXPkgTokenSpaceGuid.PcdXXXConfig2|FALSE
> > > >     gXXXPkgTokenSpaceGuid.PcdXXXConfig3|FALSE
> > > >
> > > >   [PcdsDynamicDefault.Common.SkuA]
> > > >     gXXXPkgTokenSpaceGuid.PcdXXXSignature|"SkuA"
> > > >     gXXXPkgTokenSpaceGuid.PcdXXXConfig1|TRUE
> > > >     gXXXPkgTokenSpaceGuid.PcdXXXConfig2|TRUE
> > > >     # No need statement for PcdXXXConfig3 whose value will derive
> > > > from DEFAULT SKU and be FLASE.
> > > >
> > > >   [PcdsDynamicDefault.Common.SkuB]
> > > >     gXXXPkgTokenSpaceGuid.PcdXXXSignature|" SkuB"
> > > >     # No need statement for PcdXXXConfig1 and PcdXXXConfig2 whose
> > > > values will derive from SkuA SKU and be TRUE.
> > > >     gXXXPkgTokenSpaceGuid.PcdXXXConfig3|TRUE
> > > > _______________________________________________
> > > > edk2-devel mailing list
> > > > edk2-devel@lists.01.org
> > > > https://lists.01.org/mailman/listinfo/edk2-devel
> > > > _______________________________________________
> > > > edk2-devel mailing list
> > > > edk2-devel@lists.01.org
> > > > https://lists.01.org/mailman/listinfo/edk2-devel
> > > _______________________________________________
> > > edk2-devel mailing list
> > > edk2-devel@lists.01.org
> > > https://lists.01.org/mailman/listinfo/edk2-devel
> > _______________________________________________
> > edk2-devel mailing list
> > edk2-devel@lists.01.org
> > https://lists.01.org/mailman/listinfo/edk2-devel
> _______________________________________________
> edk2-devel mailing list
> edk2-devel@lists.01.org
> https://lists.01.org/mailman/listinfo/edk2-devel