[RFC DSD v3 03/04] _DSD Formal Language Ruleset

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



_DSD Formal Language Ruleset
============================
Dated:	2016-08-04
Status:	Final


Contents
--------
1. Purpose
2. Formalisms for Describing Property Sets
    2.1 Notes on Database Structure
    2.2 Formal Language for Property Sets
        2.2.1 Set Definition
        2.2.2 Preamble Definition
        2.2.3 Set Location
        2.2.4 Property Definitions
        2.2.5 Additional Syntactical and Semantic Rules
        2.2.6 A Brief and Incomplete Example
3. References
4. Contributors



1. Purpose
----------
This document specifies the formal language to be used for describing
property sets to be registered in the common database of device
properties [1].  This common database is to be used with the ACPI _DSD
(Device Specific Data) device configuration object along with the Device
Properties UUID, daffd814-6eba-4d8c-8a91-bc9bbf4aa301 [2].

Definitions of the terms "device properties", "property sets", "property
subsets", and the process for requesting they be included in the common
database of device properties are documented independently [3].


2. Formalisms for Describing Property Sets
------------------------------------------

2.1. Notes on Database Structure
--------------------------------
Given the database structure described in [1], the directory tree for
a database might look something like the following:

--- Path ----------------------------------            -- Entries --

+-- root						1
     |
     +-- bus ("acpi"|"pci"|"common")			1..n
          |
          +-- vendor directory ("YYDN")			1..n
             |
	     + Device ID ("YYDN0007"|...)		1..n
	        |
		+ revision ("0"|"1"|...)		1..n
                   |
		   + property attributes file		1..n
		   |
		   + property subset tree		1..n
		      |
		      + property attributes file	1..n
		      |
		      + property subset tree		1..n


This is just a quick overview of the possible internal structure.  There
will be only one root, but there are can be one to n subdirectories or
files of each and every flavor.  In parentheses, we can see what some of
the values might be, so that we might have a path that looks like this:

/acpi/YYDN/YYDN0007/0/instruction-cycles-per-fortnight

Again, this is just an example, provided to be more concrete about the
database and its content, and to show how a formal language can be
geared towards describing that content.


2.2. Formal Language for Property Sets
--------------------------------------
When someone asks to register a property set, what information must they
provide?  Informally, we need enough information to be able to place a
file in the database directory, and then stash a bit more information in
that file to describe a property and its possible values.


2.2.1. Set Definition
---------------------
On a more formal basis, we can use BNF to describe what we need:

    <set-definition> ::= <set-preamble> <set-location> <set-properties>

That is, each property set definition consists of a preamble, some
location information, and property definitions.


2.2.2. Preamble Definition
--------------------------
The preamble provides an identifying name for the set, defines what type
of set we have, lists any other sets this one may be derived from, and
who approved the registration; optionally, we can include who submitted
it, and who reviewed it, solely for historical records:

    <set-preamble> ::= <set-name> <set-type> [ <parent-sets> ]
		       <acks> [ <submitters> ] [ <reviewers> ]

    <set-name> ::= "property-set" ":" <name> <EOL>

    <set-type> ::= "set-type" ":" <type-name> <EOL>
    <type-name> ::= "abstract" | "subset" | "definition"

    <parent-sets> ::= "derived-from" ":" <set-list>
    <set-list> ::= <full-path-to-parent> { <EOL> <full-path-to-parent> } <EOL>

    <acks> ::= <acked-by> | <acked-by> <acks>
    <acked-by> ::= "acked-by" ":" <person> <EOL>

    <person> ::= <first-name> <last-name> "<" <email-address> ">"

    <submitters> ::= <submitted-by> | <submitted-by> <submitters>
    <submitted-by> ::= "submitted-by" ":" <person> <EOL>

    <reviewers> ::= <reviewed-by> | <reviewed-by> <reviewers>
    <reviewed-by> ::= "reviewed-by" ":" <person> <EOL>

A <name> may not include an <EOL>, commas (",") or slashes ("/").

A property set can be "abstract", a "subset", or a "definition."  A
definition has no ambiguity: we know precisely where to put the
information in the database, what the property is, and what type it is.
A subset is a collection of property definitions for a single device.
An abstract set could be located in any number of places in the database
directory tree and is only present in order to provide properties other
definitions can re-use.

The <submitters>, if not provided, will be gleaned from the email the
property set was sent in (the From: header, see [3] for details on the
submission process).  The <reviewers> will be gleaned from emails sent
in response to a submission on the dsd@xxxxxxxxxx mailing list; the line
"Reviewed-by: Name <email@address>" will be the indicator to add such a
line to the property set definition.  The <acks> are similar in that (1)
there must be an email on the mailing list with "Acked-by: Name
<email@address>", but (2) the name used will be checked againt the list
of database maintainers and must be present there before it is
considered valid.

There may be multiple <derived-from> lines; <acks>, <submitters> and
<reviewers> can be multiple lines as well.


2.2.3. Set Location
-------------------
Location information tells us where in the database we need to place the
device properties being defined in this set:

    <set-location> ::= <abstract-set-location> | <full-set-location>
    <abstract-set-location> ::= <vendor> [ <bus> ] [ <device-id> ] <revision>
    <full-set-location> ::= <vendor> <bus> <device-id> <revision>

    <vendor> ::= "vendor" ":" <name> <EOL>

    <bus> ::= "bus" ":" <name> [ "," "shared" ] <EOL>

    <device-id> ::= "device-id" ":" <identifier> <EOL>

    <revision> ::= "revision" ":" <revision-string> <EOL>

    <revision-string> ::= <integral-value> { '.' <integral-value> }
    <integral-value> ::= any C integer string

These can be assembled into a path in the database:

    /bus-name/vendor-name/device-id-identifier/revision-string

Unless this is an abstract property set for other sets to inherit from,
all parts of the path must be provided.  For an abstract set, bus and
vendor are the only required parts.

An <identifier> is the string used by the OS reading a property set to
connect a device to a device driver (e.g., PNP0010, or ACPI0007).


2.2.4. Property Definitions
---------------------------
A stanza must be provided to describe each individual property that is
an element of a property set:

    <set-properties> ::= <property-definition> |
			 <property-definition> <set-properties>

    <property-definition> ::= <property-name> <property-type>
			      [ <property-usage> ]
			      [ <property-values> ]
			      [ <property-description> ]
			      [ <property-example> ]
			      [ <property-dependency> ]

    <property-name> ::= "property" ":" <name> <EOL>

    <property-type> ::= "type" ":" <type-name> <EOL>
    <type-name> ::= "integer" | "string" | "reference" | "package"

    <property-usage> ::= "usage" ":" "required" | "optional" <EOL>

    <property-values> ::= "values" ":" <value-list>
    <value-list> ::= <value-constraint> | <value-constraint> <value-list>
    <value-constraint> ::= <string-value> | <integer-value> |
			   <reference-value> | <subpackage-value>
    <string-value> ::= "token" ":" <string> <EOL>
                       "description" ":" <string> <EOL>
    <integer-value> ::= "integer" ":" <range-list> <EOL>
                        "description" ":" <string> <EOL>
    <range-list> ::= <range> { "," <range> }
    <range> ::= <integer> ".." <integer> | <integer>
    <reference-value> ::= "reference" ":" <ACPI-name> <EOL>
                          "description" ":" <string> <EOL>
    <subpackage-value> ::= "subpackage" ":" <subpackage-definition> <EOL>
                           "description" ":" <string> <EOL>
    <subpackage-definition> ::= "{" <field-type> { "," <field-type> } "}"
    <field-type> ::= "integer" | "reference" | "string" |
		     <subpackage-definition>

    <property-description> ::= "description" ":" <free-form-text>

    <property-example> ::= "example" ":" <free-form-text>

    <free-form-text> ::= <any-text> <EOL> { " " <any-text> <EOL> } <EOL>

    <property-dependency> ::= "requires" ":" <property-name-list>
    <property-name-list> ::= <name> { "," <name> }

A <string> starts with the first non-blank character and continues to
the <EOL>; any white space between the last non-blank and the <EOL> will
be removed.


2.2.5. Additional Syntactical and Semantic Rules
------------------------------------------------
Comments start anywhere in a line of text, using the "#", and continue
to the end of the line ("<EOL>").

Most lines consist of a keyword, followed by a ":", followed by text
providing a value, followed by an <EOL>.  The only exceptions are
descriptions and examples that may have several lines of text, with each
continuation line starting with a " ".  The first line that is either an
empty line or starts with something other than a " " terminates the
block of text.

Lines may generally appear in almost any order, within some constraints:

    -- A <set-name> marks the beginning of a property set and all lines
       after it, until the next <set-name>, will be part of that set.
       So, a <set-type> will always define the type of the prior
       <set-name> in the file.

    -- Similarly, a <property-name> marks the beginning of a property
       definition and all lines after it until the next <property-name>,
       will be part of that definition.  So, a <property-type> will
       always define the type of the prior <property-name> in the file.

    -- Set location information always defines the location of the
       previous <set-name> in the file, and applies identically to all
       property definitions provided in the set.

    -- Property values, if given, must follow the order given in the
       BNF: "values:<EOL>" followed by one or more sets of "token:
       <string><EOL> description: <string><EOL>", or the equivalent
       "integer: <range><EOL> description: <string> <EOL>".  This is the
       only rigorous ordering defined.

In general, white space and blank lines are ignored and irrelevant.  The
text for descriptions and examples is the only time it is relevant, and
then only to indicate the continuation of the text.

If a property has a <property-dependency>, each <property-name> must be
a name being defined in the current property set, or a parent set the
property set is derived from.

If a property does not have a <property-usage> explicitly stated, it is
assumed to be "optional."  However, if <property-usage> indicates the
property is "required," and the property also has a
<property-dependency>, any dependencies also become "required."

If a property set is derived from other sets, the full path to the
parent sets must be provided.

Properties in derived sets cannot override definitions in sets they
have inherited from.  If property set A defines a property "some-name",
and property set B is derived from A, then B cannot also contain a
property "some-name", and will be considered in error.

As noted in [1], revision strings consist of groupings of ASCII digits
separated by a '.', or any numeric value commonly used in the C
language.  Comparison is from left to right, and values are compared
numerically; any group not present is assumed to be zero.  For example,
revision '1.4' > '1.3.1', '0x01040000' > '0x01030100', and '20160421' >
'20160312'.  New revisions must always be greater than any existing
revisions; i.e., you cannot add revision '1.3.1' if '1.4' already
exists.



2.2.6. A Brief and Incomplete Example
-------------------------------------
property-set:	NIC Device Properties
set-type:	definition
bus:		pci
vendor:		YYDN
device-id:	YYDN1001
revision:	0
derived-from:	/pci/YYDN/YYDN1000/25

property:	phy-mode
type:		string
description:
	Defines the PHY mode to be used for this device.
values:
	token:		na
	description:	none available
	token:		mii
	description	media independent interface (MII)
	token:		gmii
	description	gigabit MII
	token:		sgmii
	description	serial gigabit MII
example:
	Package (2) { "phy-mode", "gmii" }

property:	phy-channel
type:		integer
values:
	integer:	0..255
	description:	hex value of the PHY channel number
description:
	If present, defines the PHY channel number (in hex) to be used
	by this device.
example:
	Package (2) { "phy-channel", 3 }

submitted-by:	N. Bonaparte <nb@xxxxxxxxxxx>
acked-by:	A. Wellesley <duke@xxxxxxxxxxxxxxxx>


3. References
-------------
[1] See document entitled "_DSD Property Database Ruleset"

[2]
http://www.uefi.org/sites/default/files/resources/_DSD-device-properties-UUID.pdf

[3] See document entitled "_DSD Property Registration Ruleset"


4. Contributors
---------------
In alphabetical order, by first name:

Al Stone <ahs3@xxxxxxxxxx>
Charles Garcia-Tobin <charles.garcia-tobin@xxxxxxx>
Darren Hart <darren.hart@xxxxxxxxx>
David Woodhouse <david.woodhouse@xxxxxxxxx>
Mark Doran <mark.doran@xxxxxxxxx>
Rafael Wysocki <rafael.j.wysocki@xxxxxxxxx>
Robert Gough <robert.gough@xxxxxxxxx>



-- 
ciao,
al
-----------------------------------
Al Stone
Software Engineer
Red Hat, Inc.
ahs3@xxxxxxxxxx
-----------------------------------
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html



[Index of Archives]     [Linux IBM ACPI]     [Linux Power Management]     [Linux Kernel]     [Linux Laptop]     [Kernel Newbies]     [Share Photos]     [Security]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Samba]     [Video 4 Linux]     [Device Mapper]     [Linux Resources]

  Powered by Linux