qemu-e2k/docs/qapi-code-gen.txt

1160 lines
45 KiB
Plaintext
Raw Normal View History

= How to use the QAPI code generator =
Copyright IBM Corp. 2011
Copyright (C) 2012-2016 Red Hat, Inc.
This work is licensed under the terms of the GNU GPL, version 2 or
later. See the COPYING file in the top-level directory.
== Introduction ==
QAPI is a native C API within QEMU which provides management-level
functionality to internal and external users. For external
users/processes, this interface is made available by a JSON-based wire
format for the QEMU Monitor Protocol (QMP) for controlling qemu, as
well as the QEMU Guest Agent (QGA) for communicating with the guest.
The remainder of this document uses "Client JSON Protocol" when
referring to the wire contents of a QMP or QGA connection.
To map Client JSON Protocol interfaces to the native C QAPI
implementations, a JSON-based schema is used to define types and
function signatures, and a set of scripts is used to generate types,
signatures, and marshaling/dispatch code. This document will describe
how the schemas, scripts, and resulting code are used.
== QMP/Guest agent schema ==
A QAPI schema file is designed to be loosely based on JSON
(http://www.ietf.org/rfc/rfc7159.txt) with changes for quoting style
and the use of comments; a QAPI schema file is then parsed by a python
code generation program. A valid QAPI schema consists of a series of
top-level expressions, with no commas between them. Where
dictionaries (JSON objects) are used, they are parsed as python
OrderedDicts so that ordering is preserved (for predictable layout of
generated C structs and parameter lists). Ordering doesn't matter
between top-level expressions or the keys within an expression, but
does matter within dictionary values for 'data' and 'returns' members
of a single expression. QAPI schema input is written using 'single
quotes' instead of JSON's "double quotes" (in contrast, Client JSON
Protocol uses no comments, and while input accepts 'single quotes' as
an extension, output is strict JSON using only "double quotes"). As
in JSON, trailing commas are not permitted in arrays or dictionaries.
Input must be ASCII (although QMP supports full Unicode strings, the
QAPI parser does not). At present, there is no place where a QAPI
schema requires the use of JSON numbers or null.
Comments are allowed; anything between an unquoted # and the following
newline is ignored. Although there is not yet a documentation
generator, a form of stylized comments has developed for consistently
documenting details about an expression and when it was added to the
schema. The documentation is delimited between two lines of ##, then
the first line names the expression, an optional overview is provided,
then individual documentation about each member of 'data' is provided,
and finally, a 'Since: x.y.z' tag lists the release that introduced
the expression. Optional members are tagged with the phrase
'#optional', often with their default value; and extensions added
after the expression was first released are also given a '(since
x.y.z)' comment. For example:
##
# @BlockStats:
#
# Statistics of a virtual block device or a block backing device.
#
# @device: #optional If the stats are for a virtual block device, the name
# corresponding to the virtual block device.
#
# @stats: A @BlockDeviceStats for the device.
#
# @parent: #optional This describes the file block device if it has one.
#
# @backing: #optional This describes the backing block device if it has one.
# (Since 2.0)
#
# Since: 0.14.0
##
{ 'struct': 'BlockStats',
'data': {'*device': 'str', 'stats': 'BlockDeviceStats',
'*parent': 'BlockStats',
'*backing': 'BlockStats'} }
The schema sets up a series of types, as well as commands and events
that will use those types. Forward references are allowed: the parser
scans in two passes, where the first pass learns all type names, and
the second validates the schema and generates the code. This allows
the definition of complex structs that can have mutually recursive
types, and allows for indefinite nesting of Client JSON Protocol that
satisfies the schema. A type name should not be defined more than
once. It is permissible for the schema to contain additional types
not used by any commands or events in the Client JSON Protocol, for
the side effect of generated C code used internally.
There are seven top-level expressions recognized by the parser:
'include', 'command', 'struct', 'enum', 'union', 'alternate', and
'event'. There are several groups of types: simple types (a number of
built-in types, such as 'int' and 'str'; as well as enumerations),
complex types (structs and two flavors of unions), and alternate types
(a choice between other types). The 'command' and 'event' expressions
can refer to existing types by name, or list an anonymous type as a
dictionary. Listing a type name inside an array refers to a
single-dimension array of that type; multi-dimension arrays are not
directly supported (although an array of a complex struct that
contains an array member is possible).
Types, commands, and events share a common namespace. Therefore,
generally speaking, type definitions should always use CamelCase for
user-defined type names, while built-in types are lowercase. Type
definitions should not end in 'Kind', as this namespace is used for
creating implicit C enums for visiting union types, or in 'List', as
this namespace is used for creating array types. Command names,
and member names within a type, should be all lower case with words
separated by a hyphen. However, some existing older commands and
complex types use underscore; when extending such expressions,
consistency is preferred over blindly avoiding underscore. Event
names should be ALL_CAPS with words separated by underscore. Member
qapi: Reserve 'q_*' and 'has_*' member names c_name() produces names starting with 'q_' when protecting a dictionary member name that would fail to directly compile, but in doing so can cause clashes with any member name already beginning with 'q-' or 'q_'. Likewise, we create a C name 'has_' for any optional member that can clash with any member name beginning with 'has-' or 'has_'. Technically, rather than blindly reserving the namespace, we could try to complain about user names only when an actual collision occurs, or even teach c_name() how to munge names to avoid collisions. But it is not trivial, especially when collisions can occur across multiple types (such as via inheritance or flat unions). Besides, no existing .json files are trying to use these names. So it's easier to just outright forbid the potential for collision. We can always relax things in the future if a real need arises for QMP to express member names that have been forbidden here. 'has_' only has to be reserved for struct/union member names, while 'q_' is reserved everywhere (matching the fact that only members can be optional, while we use c_name() for munging both members and entities). Note that we could relax 'q_' restrictions on entities independently from member names; for example, c_name('qmp_' + 'unix') would result in a different function name than our current 'qmp_' + c_name('unix'). Update and add tests to cover the new error messages. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1445898903-12082-6-git-send-email-eblake@redhat.com> [Consistently pass protect=False to c_name(); commit message tweaked slightly] Signed-off-by: Markus Armbruster <armbru@redhat.com>
2015-10-26 23:34:44 +01:00
names cannot start with 'has-' or 'has_', as this is reserved for
tracking optional members.
Any name (command, event, type, member, or enum value) beginning with
"x-" is marked experimental, and may be withdrawn or changed
incompatibly in a future release. All names must begin with a letter,
and contain only ASCII letters, digits, dash, and underscore. There
are two exceptions: enum values may start with a digit, and any
extensions added by downstream vendors should start with a prefix
matching "__RFQDN_" (for the reverse-fully-qualified-domain-name of
the vendor), even if the rest of the name uses dash (example:
__com.redhat_drive-mirror). Names beginning with 'q_' are reserved
for the generator: QMP names that resemble C keywords or other
problematic strings will be munged in C to use this prefix. For
example, a member named "default" in qapi becomes "q_default" in the
generated C code.
In the rest of this document, usage lines are given for each
expression type, with literal strings written in lower case and
placeholders written in capitals. If a literal string includes a
prefix of '*', that key/value pair can be omitted from the expression.
For example, a usage statement that includes '*base':STRUCT-NAME
means that an expression has an optional key 'base', which if present
must have a value that forms a struct name.
=== Built-in Types ===
The following types are predefined, and map to C as follows:
Schema C JSON
str char * any JSON string, UTF-8
number double any JSON number
int int64_t a JSON number without fractional part
that fits into the C integer type
int8 int8_t likewise
int16 int16_t likewise
int32 int32_t likewise
int64 int64_t likewise
uint8 uint8_t likewise
uint16 uint16_t likewise
uint32 uint32_t likewise
uint64 uint64_t likewise
size uint64_t like uint64_t, except StringInputVisitor
accepts size suffixes
bool bool JSON true or false
any QObject * any JSON value
QType QType JSON string matching enum QType values
=== Includes ===
Usage: { 'include': STRING }
The QAPI schema definitions can be modularized using the 'include' directive:
{ 'include': 'path/to/file.json' }
The directive is evaluated recursively, and include paths are relative to the
file using the directive. Multiple includes of the same file are
idempotent. No other keys should appear in the expression, and the include
value should be a string.
As a matter of style, it is a good idea to have all files be
self-contained, but at the moment, nothing prevents an included file
from making a forward reference to a type that is only introduced by
an outer file. The parser may be made stricter in the future to
prevent incomplete include files.
=== Struct types ===
Usage: { 'struct': STRING, 'data': DICT, '*base': STRUCT-NAME }
A struct is a dictionary containing a single 'data' key whose value is
a dictionary; the dictionary may be empty. This corresponds to a
struct in C or an Object in JSON. Each value of the 'data' dictionary
must be the name of a type, or a one-element array containing a type
name. An example of a struct is:
{ 'struct': 'MyType',
'data': { 'member1': 'str', 'member2': 'int', '*member3': 'str' } }
The use of '*' as a prefix to the name means the member is optional in
the corresponding JSON protocol usage.
The default initialization value of an optional argument should not be changed
between versions of QEMU unless the new default maintains backward
compatibility to the user-visible behavior of the old default.
With proper documentation, this policy still allows some flexibility; for
example, documenting that a default of 0 picks an optimal buffer size allows
one release to declare the optimal size at 512 while another release declares
the optimal size at 4096 - the user-visible behavior is not the bytes used by
the buffer, but the fact that the buffer was optimal size.
On input structures (only mentioned in the 'data' side of a command), changing
from mandatory to optional is safe (older clients will supply the option, and
newer clients can benefit from the default); changing from optional to
mandatory is backwards incompatible (older clients may be omitting the option,
and must continue to work).
On output structures (only mentioned in the 'returns' side of a command),
changing from mandatory to optional is in general unsafe (older clients may be
expecting the member, and could crash if it is missing), although it
can be done if the only way that the optional argument will be omitted
is when it is triggered by the presence of a new input flag to the
command that older clients don't know to send. Changing from optional
to mandatory is safe.
A structure that is used in both input and output of various commands
must consider the backwards compatibility constraints of both directions
of use.
A struct definition can specify another struct as its base.
In this case, the members of the base type are included as top-level members
of the new struct's dictionary in the Client JSON Protocol wire
format. An example definition is:
{ 'struct': 'BlockdevOptionsGenericFormat', 'data': { 'file': 'str' } }
{ 'struct': 'BlockdevOptionsGenericCOWFormat',
'base': 'BlockdevOptionsGenericFormat',
'data': { '*backing': 'str' } }
An example BlockdevOptionsGenericCOWFormat object on the wire could use
both members like this:
{ "file": "/some/place/my-image",
"backing": "/some/place/my-backing-file" }
=== Enumeration types ===
Usage: { 'enum': STRING, 'data': ARRAY-OF-STRING }
{ 'enum': STRING, '*prefix': STRING, 'data': ARRAY-OF-STRING }
An enumeration type is a dictionary containing a single 'data' key
whose value is a list of strings. An example enumeration is:
{ 'enum': 'MyEnum', 'data': [ 'value1', 'value2', 'value3' ] }
Nothing prevents an empty enumeration, although it is probably not
useful. The list of strings should be lower case; if an enum name
represents multiple words, use '-' between words. The string 'max' is
not allowed as an enum value, and values should not be repeated.
The enum constants will be named by using a heuristic to turn the
type name into a set of underscore separated words. For the example
above, 'MyEnum' will turn into 'MY_ENUM' giving a constant name
of 'MY_ENUM_VALUE1' for the first value. If the default heuristic
does not result in a desirable name, the optional 'prefix' member
can be used when defining the enum.
The enumeration values are passed as strings over the Client JSON
Protocol, but are encoded as C enum integral values in generated code.
While the C code starts numbering at 0, it is better to use explicit
comparisons to enum values than implicit comparisons to 0; the C code
will also include a generated enum member ending in _MAX for tracking
the size of the enum, useful when using common functions for
converting between strings and enum values. Since the wire format
always passes by name, it is acceptable to reorder or add new
enumeration members in any location without breaking clients of Client
JSON Protocol; however, removing enum values would break
compatibility. For any struct that has a member that will only contain
a finite set of string values, using an enum type for that member is
better than open-coding the member to be type 'str'.
=== Union types ===
Usage: { 'union': STRING, 'data': DICT }
or: { 'union': STRING, 'data': DICT, 'base': STRUCT-NAME-OR-DICT,
'discriminator': ENUM-MEMBER-OF-BASE }
Union types are used to let the user choose between several different
variants for an object. There are two flavors: simple (no
discriminator or base), and flat (both discriminator and base). A union
type is defined using a data dictionary as explained in the following
paragraphs. The data dictionary for either type of union must not
be empty.
A simple union type defines a mapping from automatic discriminator
values to data types like in this example:
{ 'struct': 'BlockdevOptionsFile', 'data': { 'filename': 'str' } }
{ 'struct': 'BlockdevOptionsQcow2',
'data': { 'backing': 'str', '*lazy-refcounts': 'bool' } }
{ 'union': 'BlockdevOptionsSimple',
'data': { 'file': 'BlockdevOptionsFile',
'qcow2': 'BlockdevOptionsQcow2' } }
In the Client JSON Protocol, a simple union is represented by a
dictionary that contains the 'type' member as a discriminator, and a
'data' member that is of the specified data type corresponding to the
discriminator value, as in these examples:
{ "type": "file", "data": { "filename": "/some/place/my-image" } }
{ "type": "qcow2", "data": { "backing": "/some/place/my-image",
"lazy-refcounts": true } }
The generated C code uses a struct containing a union. Additionally,
an implicit C enum 'NameKind' is created, corresponding to the union
'Name', for accessing the various branches of the union. No branch of
the union can be named 'max', as this would collide with the implicit
enum. The value for each branch can be of any type.
A flat union definition avoids nesting on the wire, and specifies a
set of common members that occur in all variants of the union. The
'base' key must specify either a type name (the type must be a
struct, not a union), or a dictionary representing an anonymous type.
All branches of the union must be complex types, and the top-level
members of the union dictionary on the wire will be combination of
members from both the base type and the appropriate branch type (when
merging two dictionaries, there must be no keys in common). The
'discriminator' member must be the name of a non-optional enum-typed
member of the base struct.
The following example enhances the above simple union example by
adding an optional common member 'read-only', renaming the
discriminator to something more applicable than the simple union's
default of 'type', and reducing the number of {} required on the wire:
{ 'enum': 'BlockdevDriver', 'data': [ 'file', 'qcow2' ] }
{ 'union': 'BlockdevOptions',
'base': { 'driver': 'BlockdevDriver', '*read-only': 'bool' },
'discriminator': 'driver',
'data': { 'file': 'BlockdevOptionsFile',
'qcow2': 'BlockdevOptionsQcow2' } }
Resulting in these JSON objects:
{ "driver": "file", "read-only": true,
"filename": "/some/place/my-image" }
{ "driver": "qcow2", "read-only": false,
"backing": "/some/place/my-image", "lazy-refcounts": true }
Notice that in a flat union, the discriminator name is controlled by
the user, but because it must map to a base member with enum type, the
code generator can ensure that branches exist for all values of the
enum (although the order of the keys need not match the declaration of
the enum). In the resulting generated C data types, a flat union is
represented as a struct with the base members included directly, and
then a union of structures for each branch of the struct.
A simple union can always be re-written as a flat union where the base
class has a single member named 'type', and where each branch of the
union has a struct with a single member named 'data'. That is,
{ 'union': 'Simple', 'data': { 'one': 'str', 'two': 'int' } }
is identical on the wire to:
{ 'enum': 'Enum', 'data': ['one', 'two'] }
{ 'struct': 'Branch1', 'data': { 'data': 'str' } }
{ 'struct': 'Branch2', 'data': { 'data': 'int' } }
{ 'union': 'Flat': 'base': { 'type': 'Enum' }, 'discriminator': 'type',
'data': { 'one': 'Branch1', 'two': 'Branch2' } }
=== Alternate types ===
Usage: { 'alternate': STRING, 'data': DICT }
An alternate type is one that allows a choice between two or more JSON
data types (string, integer, number, or object, but currently not
array) on the wire. The definition is similar to a simple union type,
where each branch of the union names a QAPI type. For example:
{ 'alternate': 'BlockdevRef',
'data': { 'definition': 'BlockdevOptions',
'reference': 'str' } }
Unlike a union, the discriminator string is never passed on the wire
for the Client JSON Protocol. Instead, the value's JSON type serves
as an implicit discriminator, which in turn means that an alternate
can only express a choice between types represented differently in
JSON. If a branch is typed as the 'bool' built-in, the alternate
accepts true and false; if it is typed as any of the various numeric
built-ins, it accepts a JSON number; if it is typed as a 'str'
built-in or named enum type, it accepts a JSON string; and if it is
typed as a complex type (struct or union), it accepts a JSON object.
Two different complex types, for instance, aren't permitted, because
both are represented as a JSON object.
The example alternate declaration above allows using both of the
following example objects:
{ "file": "my_existing_block_device_id" }
{ "file": { "driver": "file",
"read-only": false,
"filename": "/tmp/mydisk.qcow2" } }
=== Commands ===
Usage: { 'command': STRING, '*data': COMPLEX-TYPE-NAME-OR-DICT,
qapi: Implement boxed types for commands/events Turn on the ability to pass command and event arguments in a single boxed parameter, which must name a non-empty type (although the type can be a struct with all optional members). For structs, it makes it possible to pass a single qapi type instead of a breakout of all struct members (useful if the arguments are already in a struct or if the number of members is large); for other complex types, it is now possible to use a union or alternate as the data for a command or event. The empty type may be technically feasible if needed down the road, but it's easier to forbid it now and relax things to allow it later, than it is to allow it now and have to special case how the generated 'q_empty' type is handled (see commit 7ce106a9 for reasons why nothing is generated for the empty type). An alternate type is never considered empty, but now that a boxed type can be either an object or an alternate, we have to provide a trivial QAPISchemaAlternateType.is_empty(). The new call to arg_type.is_empty() during QAPISchemaCommand.check() requires that we first check the type in question; but there is no chance of introducing a cycle since objects do not refer back to commands. We still have a split in syntax checking between ad-hoc parsing up front (merely validates that 'boxed' has a sane value) and during .check() methods (if 'boxed' is set, then 'data' must name a non-empty user-defined type). Generated code is unchanged, as long as no client uses the new feature. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1468468228-27827-10-git-send-email-eblake@redhat.com> Reviewed-by: Markus Armbruster <armbru@redhat.com> [Test files renamed to *-boxed-*] Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-07-14 05:50:20 +02:00
'*returns': TYPE-NAME, '*boxed': true,
'*gen': false, '*success-response': false }
Commands are defined by using a dictionary containing several members,
where three members are most common. The 'command' member is a
mandatory string, and determines the "execute" value passed in a
Client JSON Protocol command exchange.
The 'data' argument maps to the "arguments" dictionary passed in as
part of a Client JSON Protocol command. The 'data' member is optional
and defaults to {} (an empty dictionary). If present, it must be the
string name of a complex type, or a dictionary that declares an
anonymous type with the same semantics as a 'struct' expression, with
one exception noted below when 'gen' is used.
The 'returns' member describes what will appear in the "return" member
of a Client JSON Protocol reply on successful completion of a command.
The member is optional from the command declaration; if absent, the
"return" member will be an empty dictionary. If 'returns' is present,
it must be the string name of a complex or built-in type, a
one-element array containing the name of a complex or built-in type,
with one exception noted below when 'gen' is used. Although it is
permitted to have the 'returns' member name a built-in type or an
array of built-in types, any command that does this cannot be extended
to return additional information in the future; thus, new commands
should strongly consider returning a dictionary-based type or an array
of dictionaries, even if the dictionary only contains one member at the
present.
All commands in Client JSON Protocol use a dictionary to report
failure, with no way to specify that in QAPI. Where the error return
is different than the usual GenericError class in order to help the
client react differently to certain error conditions, it is worth
documenting this in the comments before the command declaration.
Some example commands:
{ 'command': 'my-first-command',
'data': { 'arg1': 'str', '*arg2': 'str' } }
{ 'struct': 'MyType', 'data': { '*value': 'str' } }
{ 'command': 'my-second-command',
'returns': [ 'MyType' ] }
which would validate this Client JSON Protocol transaction:
=> { "execute": "my-first-command",
"arguments": { "arg1": "hello" } }
<= { "return": { } }
=> { "execute": "my-second-command" }
<= { "return": [ { "value": "one" }, { } ] }
qapi: Implement boxed types for commands/events Turn on the ability to pass command and event arguments in a single boxed parameter, which must name a non-empty type (although the type can be a struct with all optional members). For structs, it makes it possible to pass a single qapi type instead of a breakout of all struct members (useful if the arguments are already in a struct or if the number of members is large); for other complex types, it is now possible to use a union or alternate as the data for a command or event. The empty type may be technically feasible if needed down the road, but it's easier to forbid it now and relax things to allow it later, than it is to allow it now and have to special case how the generated 'q_empty' type is handled (see commit 7ce106a9 for reasons why nothing is generated for the empty type). An alternate type is never considered empty, but now that a boxed type can be either an object or an alternate, we have to provide a trivial QAPISchemaAlternateType.is_empty(). The new call to arg_type.is_empty() during QAPISchemaCommand.check() requires that we first check the type in question; but there is no chance of introducing a cycle since objects do not refer back to commands. We still have a split in syntax checking between ad-hoc parsing up front (merely validates that 'boxed' has a sane value) and during .check() methods (if 'boxed' is set, then 'data' must name a non-empty user-defined type). Generated code is unchanged, as long as no client uses the new feature. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1468468228-27827-10-git-send-email-eblake@redhat.com> Reviewed-by: Markus Armbruster <armbru@redhat.com> [Test files renamed to *-boxed-*] Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-07-14 05:50:20 +02:00
The generator emits a prototype for the user's function implementing
the command. Normally, 'data' is a dictionary for an anonymous type,
or names a struct type (possibly empty, but not a union), and its
members are passed as separate arguments to this function. If the
command definition includes a key 'boxed' with the boolean value true,
then 'data' is instead the name of any non-empty complex type
(struct, union, or alternate), and a pointer to that QAPI type is
passed as a single argument.
The generator also emits a marshalling function that extracts
arguments for the user's function out of an input QDict, calls the
user's function, and if it succeeded, builds an output QObject from
its return value.
In rare cases, QAPI cannot express a type-safe representation of a
corresponding Client JSON Protocol command. You then have to suppress
generation of a marshalling function by including a key 'gen' with
boolean value false, and instead write your own function. Please try
to avoid adding new commands that rely on this, and instead use
type-safe unions. For an example of this usage:
{ 'command': 'netdev_add',
'data': {'type': 'str', 'id': 'str'},
'gen': false }
Normally, the QAPI schema is used to describe synchronous exchanges,
where a response is expected. But in some cases, the action of a
command is expected to change state in a way that a successful
response is not possible (although the command will still return a
normal dictionary error on failure). When a successful reply is not
possible, the command expression should include the optional key
'success-response' with boolean value false. So far, only QGA makes
use of this member.
=== Events ===
qapi: Implement boxed types for commands/events Turn on the ability to pass command and event arguments in a single boxed parameter, which must name a non-empty type (although the type can be a struct with all optional members). For structs, it makes it possible to pass a single qapi type instead of a breakout of all struct members (useful if the arguments are already in a struct or if the number of members is large); for other complex types, it is now possible to use a union or alternate as the data for a command or event. The empty type may be technically feasible if needed down the road, but it's easier to forbid it now and relax things to allow it later, than it is to allow it now and have to special case how the generated 'q_empty' type is handled (see commit 7ce106a9 for reasons why nothing is generated for the empty type). An alternate type is never considered empty, but now that a boxed type can be either an object or an alternate, we have to provide a trivial QAPISchemaAlternateType.is_empty(). The new call to arg_type.is_empty() during QAPISchemaCommand.check() requires that we first check the type in question; but there is no chance of introducing a cycle since objects do not refer back to commands. We still have a split in syntax checking between ad-hoc parsing up front (merely validates that 'boxed' has a sane value) and during .check() methods (if 'boxed' is set, then 'data' must name a non-empty user-defined type). Generated code is unchanged, as long as no client uses the new feature. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1468468228-27827-10-git-send-email-eblake@redhat.com> Reviewed-by: Markus Armbruster <armbru@redhat.com> [Test files renamed to *-boxed-*] Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-07-14 05:50:20 +02:00
Usage: { 'event': STRING, '*data': COMPLEX-TYPE-NAME-OR-DICT,
'*boxed': true }
Events are defined with the keyword 'event'. It is not allowed to
name an event 'MAX', since the generator also produces a C enumeration
of all event names with a generated _MAX value at the end. When
'data' is also specified, additional info will be included in the
event, with similar semantics to a 'struct' expression. Finally there
will be C API generated in qapi-event.h; when called by QEMU code, a
message with timestamp will be emitted on the wire.
An example event is:
{ 'event': 'EVENT_C',
'data': { '*a': 'int', 'b': 'str' } }
Resulting in this JSON object:
{ "event": "EVENT_C",
"data": { "b": "test string" },
"timestamp": { "seconds": 1267020223, "microseconds": 435656 } }
qapi: Implement boxed types for commands/events Turn on the ability to pass command and event arguments in a single boxed parameter, which must name a non-empty type (although the type can be a struct with all optional members). For structs, it makes it possible to pass a single qapi type instead of a breakout of all struct members (useful if the arguments are already in a struct or if the number of members is large); for other complex types, it is now possible to use a union or alternate as the data for a command or event. The empty type may be technically feasible if needed down the road, but it's easier to forbid it now and relax things to allow it later, than it is to allow it now and have to special case how the generated 'q_empty' type is handled (see commit 7ce106a9 for reasons why nothing is generated for the empty type). An alternate type is never considered empty, but now that a boxed type can be either an object or an alternate, we have to provide a trivial QAPISchemaAlternateType.is_empty(). The new call to arg_type.is_empty() during QAPISchemaCommand.check() requires that we first check the type in question; but there is no chance of introducing a cycle since objects do not refer back to commands. We still have a split in syntax checking between ad-hoc parsing up front (merely validates that 'boxed' has a sane value) and during .check() methods (if 'boxed' is set, then 'data' must name a non-empty user-defined type). Generated code is unchanged, as long as no client uses the new feature. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1468468228-27827-10-git-send-email-eblake@redhat.com> Reviewed-by: Markus Armbruster <armbru@redhat.com> [Test files renamed to *-boxed-*] Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-07-14 05:50:20 +02:00
The generator emits a function to send the event. Normally, 'data' is
a dictionary for an anonymous type, or names a struct type (possibly
empty, but not a union), and its members are passed as separate
arguments to this function. If the event definition includes a key
'boxed' with the boolean value true, then 'data' is instead the name of
any non-empty complex type (struct, union, or alternate), and a
pointer to that QAPI type is passed as a single argument.
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
== Client JSON Protocol introspection ==
Clients of a Client JSON Protocol commonly need to figure out what
exactly the server (QEMU) supports.
For this purpose, QMP provides introspection via command
query-qmp-schema. QGA currently doesn't support introspection.
While Client JSON Protocol wire compatibility should be maintained
between qemu versions, we cannot make the same guarantees for
introspection stability. For example, one version of qemu may provide
a non-variant optional member of a struct, and a later version rework
the member to instead be non-optional and associated with a variant.
Likewise, one version of qemu may list a member with open-ended type
'str', and a later version could convert it to a finite set of strings
via an enum type; or a member may be converted from a specific type to
an alternate that represents a choice between the original type and
something else.
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
query-qmp-schema returns a JSON array of SchemaInfo objects. These
objects together describe the wire ABI, as defined in the QAPI schema.
qapi-introspect: Document lack of sorting qapi-code-gen.txt already claims that types, commands, and events share a common namespace; set this in stone by further documenting that our introspection output will never have collisions with the same name tied to more than one meta-type. Our largest QMP enum currently has 125 values, our largest object type has 27 members, and the mean for each is less than 10. These sizes are small enough that the per-element overhead of O(log n) binary searching probably outweighs the speed possible with direct O(n) linear searching (a better algorithm with more overhead will only beat a leaner naive algorithm only as you scale to larger input sizes). Arguably, the overall SchemaInfo array could be sorted by name; there, we currently have 531 entities, large enough for a binary search to be faster than linear. However, remember that we have mutually-recursive types, which means there is no topological ordering that will allow clients to learn all information about that type in a single linear pass; thus clients will want to do random access over the data, and they will probably read the introspection output into a hashtable for O(1) lookup rather than O(log n) binary searching, at which point, pre-sorting our introspection output doesn't help the client. It doesn't help that sorting can be subjective if you introduce locales into the mix (I'm not experienced enough with Python to know for sure, but at least it looks like it defaults to sorting in the C locale even when run under a different locale). And while our current introspection output is deterministic (because we visit entities in a sorted order), we may want to change that order in the future (such as using OrderedDict to stick to .json declaration order). For these reasons, we simply document that clients should not rely on any particular order of items in introspection output. And since it is now a documented part of the contract, we have the freedom to later rearrange output if needed, without worrying about breaking well-written clients. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1446791754-23823-13-git-send-email-eblake@redhat.com> [Commit message tweaked] Signed-off-by: Markus Armbruster <armbru@redhat.com>
2015-11-06 07:35:36 +01:00
There is no specified order to the SchemaInfo objects returned; a
client must search for a particular name throughout the entire array
to learn more about that name, but is at least guaranteed that there
will be no collisions between type, command, and event names.
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
However, the SchemaInfo can't reflect all the rules and restrictions
that apply to QMP. It's interface introspection (figuring out what's
there), not interface specification. The specification is in the QAPI
schema. To understand how QMP is to be used, you need to study the
QAPI schema.
Like any other command, query-qmp-schema is itself defined in the QAPI
schema, along with the SchemaInfo type. This text attempts to give an
overview how things work. For details you need to consult the QAPI
schema.
SchemaInfo objects have common members "name" and "meta-type", and
additional variant members depending on the value of meta-type.
Each SchemaInfo object describes a wire ABI entity of a certain
meta-type: a command, event or one of several kinds of type.
SchemaInfo for commands and events have the same name as in the QAPI
schema.
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
Command and event names are part of the wire ABI, but type names are
not. Therefore, the SchemaInfo for types have auto-generated
meaningless names. For readability, the examples in this section use
meaningful type names instead.
To examine a type, start with a command or event using it, then follow
references by name.
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
QAPI schema definitions not reachable that way are omitted.
The SchemaInfo for a command has meta-type "command", and variant
members "arg-type" and "ret-type". On the wire, the "arguments"
member of a client's "execute" command must conform to the object type
named by "arg-type". The "return" member that the server passes in a
success response conforms to the type named by "ret-type".
If the command takes no arguments, "arg-type" names an object type
without members. Likewise, if the command returns nothing, "ret-type"
names an object type without members.
Example: the SchemaInfo for command query-qmp-schema
{ "name": "query-qmp-schema", "meta-type": "command",
"arg-type": "q_empty", "ret-type": "SchemaInfoList" }
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
Type "q_empty" is an automatic object type without members, and type
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
"SchemaInfoList" is the array of SchemaInfo type.
The SchemaInfo for an event has meta-type "event", and variant member
"arg-type". On the wire, a "data" member that the server passes in an
event conforms to the object type named by "arg-type".
If the event carries no additional information, "arg-type" names an
object type without members. The event may not have a data member on
the wire then.
Each command or event defined with dictionary-valued 'data' in the
QAPI schema implicitly defines an object type.
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
Example: the SchemaInfo for EVENT_C from section Events
{ "name": "EVENT_C", "meta-type": "event",
"arg-type": "q_obj-EVENT_C-arg" }
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
Type "q_obj-EVENT_C-arg" is an implicitly defined object type with
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
the two members from the event's definition.
The SchemaInfo for struct and union types has meta-type "object".
The SchemaInfo for a struct type has variant member "members".
The SchemaInfo for a union type additionally has variant members "tag"
and "variants".
"members" is a JSON array describing the object's common members, if
any. Each element is a JSON object with members "name" (the member's
name), "type" (the name of its type), and optionally "default". The
member is optional if "default" is present. Currently, "default" can
only have value null. Other values are reserved for future
qapi-introspect: Document lack of sorting qapi-code-gen.txt already claims that types, commands, and events share a common namespace; set this in stone by further documenting that our introspection output will never have collisions with the same name tied to more than one meta-type. Our largest QMP enum currently has 125 values, our largest object type has 27 members, and the mean for each is less than 10. These sizes are small enough that the per-element overhead of O(log n) binary searching probably outweighs the speed possible with direct O(n) linear searching (a better algorithm with more overhead will only beat a leaner naive algorithm only as you scale to larger input sizes). Arguably, the overall SchemaInfo array could be sorted by name; there, we currently have 531 entities, large enough for a binary search to be faster than linear. However, remember that we have mutually-recursive types, which means there is no topological ordering that will allow clients to learn all information about that type in a single linear pass; thus clients will want to do random access over the data, and they will probably read the introspection output into a hashtable for O(1) lookup rather than O(log n) binary searching, at which point, pre-sorting our introspection output doesn't help the client. It doesn't help that sorting can be subjective if you introduce locales into the mix (I'm not experienced enough with Python to know for sure, but at least it looks like it defaults to sorting in the C locale even when run under a different locale). And while our current introspection output is deterministic (because we visit entities in a sorted order), we may want to change that order in the future (such as using OrderedDict to stick to .json declaration order). For these reasons, we simply document that clients should not rely on any particular order of items in introspection output. And since it is now a documented part of the contract, we have the freedom to later rearrange output if needed, without worrying about breaking well-written clients. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1446791754-23823-13-git-send-email-eblake@redhat.com> [Commit message tweaked] Signed-off-by: Markus Armbruster <armbru@redhat.com>
2015-11-06 07:35:36 +01:00
extensions. The "members" array is in no particular order; clients
must search the entire object when learning whether a particular
member is supported.
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
Example: the SchemaInfo for MyType from section Struct types
{ "name": "MyType", "meta-type": "object",
"members": [
{ "name": "member1", "type": "str" },
{ "name": "member2", "type": "int" },
{ "name": "member3", "type": "str", "default": null } ] }
"tag" is the name of the common member serving as type tag.
"variants" is a JSON array describing the object's variant members.
Each element is a JSON object with members "case" (the value of type
tag this element applies to) and "type" (the name of an object type
qapi-introspect: Document lack of sorting qapi-code-gen.txt already claims that types, commands, and events share a common namespace; set this in stone by further documenting that our introspection output will never have collisions with the same name tied to more than one meta-type. Our largest QMP enum currently has 125 values, our largest object type has 27 members, and the mean for each is less than 10. These sizes are small enough that the per-element overhead of O(log n) binary searching probably outweighs the speed possible with direct O(n) linear searching (a better algorithm with more overhead will only beat a leaner naive algorithm only as you scale to larger input sizes). Arguably, the overall SchemaInfo array could be sorted by name; there, we currently have 531 entities, large enough for a binary search to be faster than linear. However, remember that we have mutually-recursive types, which means there is no topological ordering that will allow clients to learn all information about that type in a single linear pass; thus clients will want to do random access over the data, and they will probably read the introspection output into a hashtable for O(1) lookup rather than O(log n) binary searching, at which point, pre-sorting our introspection output doesn't help the client. It doesn't help that sorting can be subjective if you introduce locales into the mix (I'm not experienced enough with Python to know for sure, but at least it looks like it defaults to sorting in the C locale even when run under a different locale). And while our current introspection output is deterministic (because we visit entities in a sorted order), we may want to change that order in the future (such as using OrderedDict to stick to .json declaration order). For these reasons, we simply document that clients should not rely on any particular order of items in introspection output. And since it is now a documented part of the contract, we have the freedom to later rearrange output if needed, without worrying about breaking well-written clients. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1446791754-23823-13-git-send-email-eblake@redhat.com> [Commit message tweaked] Signed-off-by: Markus Armbruster <armbru@redhat.com>
2015-11-06 07:35:36 +01:00
that provides the variant members for this type tag value). The
"variants" array is in no particular order, and is not guaranteed to
list cases in the same order as the corresponding "tag" enum type.
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
Example: the SchemaInfo for flat union BlockdevOptions from section
Union types
{ "name": "BlockdevOptions", "meta-type": "object",
"members": [
{ "name": "driver", "type": "BlockdevDriver" },
{ "name": "read-only", "type": "bool", "default": null } ],
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
"tag": "driver",
"variants": [
{ "case": "file", "type": "BlockdevOptionsFile" },
{ "case": "qcow2", "type": "BlockdevOptionsQcow2" } ] }
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
Note that base types are "flattened": its members are included in the
"members" array.
A simple union implicitly defines an enumeration type for its implicit
discriminator (called "type" on the wire, see section Union types).
A simple union implicitly defines an object type for each of its
variants.
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
Example: the SchemaInfo for simple union BlockdevOptionsSimple from section
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
Union types
{ "name": "BlockdevOptionsSimple", "meta-type": "object",
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
"members": [
{ "name": "type", "type": "BlockdevOptionsSimpleKind" } ],
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
"tag": "type",
"variants": [
{ "case": "file", "type": "q_obj-BlockdevOptionsFile-wrapper" },
{ "case": "qcow2", "type": "q_obj-BlockdevOptionsQcow2-wrapper" } ] }
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
Enumeration type "BlockdevOptionsSimpleKind" and the object types
"q_obj-BlockdevOptionsFile-wrapper", "q_obj-BlockdevOptionsQcow2-wrapper"
are implicitly defined.
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
The SchemaInfo for an alternate type has meta-type "alternate", and
variant member "members". "members" is a JSON array. Each element is
a JSON object with member "type", which names a type. Values of the
qapi-introspect: Document lack of sorting qapi-code-gen.txt already claims that types, commands, and events share a common namespace; set this in stone by further documenting that our introspection output will never have collisions with the same name tied to more than one meta-type. Our largest QMP enum currently has 125 values, our largest object type has 27 members, and the mean for each is less than 10. These sizes are small enough that the per-element overhead of O(log n) binary searching probably outweighs the speed possible with direct O(n) linear searching (a better algorithm with more overhead will only beat a leaner naive algorithm only as you scale to larger input sizes). Arguably, the overall SchemaInfo array could be sorted by name; there, we currently have 531 entities, large enough for a binary search to be faster than linear. However, remember that we have mutually-recursive types, which means there is no topological ordering that will allow clients to learn all information about that type in a single linear pass; thus clients will want to do random access over the data, and they will probably read the introspection output into a hashtable for O(1) lookup rather than O(log n) binary searching, at which point, pre-sorting our introspection output doesn't help the client. It doesn't help that sorting can be subjective if you introduce locales into the mix (I'm not experienced enough with Python to know for sure, but at least it looks like it defaults to sorting in the C locale even when run under a different locale). And while our current introspection output is deterministic (because we visit entities in a sorted order), we may want to change that order in the future (such as using OrderedDict to stick to .json declaration order). For these reasons, we simply document that clients should not rely on any particular order of items in introspection output. And since it is now a documented part of the contract, we have the freedom to later rearrange output if needed, without worrying about breaking well-written clients. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1446791754-23823-13-git-send-email-eblake@redhat.com> [Commit message tweaked] Signed-off-by: Markus Armbruster <armbru@redhat.com>
2015-11-06 07:35:36 +01:00
alternate type conform to exactly one of its member types. There is
no guarantee on the order in which "members" will be listed.
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
Example: the SchemaInfo for BlockdevRef from section Alternate types
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
{ "name": "BlockdevRef", "meta-type": "alternate",
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
"members": [
{ "type": "BlockdevOptions" },
{ "type": "str" } ] }
The SchemaInfo for an array type has meta-type "array", and variant
member "element-type", which names the array's element type. Array
types are implicitly defined. For convenience, the array's name may
resemble the element type; however, clients should examine member
"element-type" instead of making assumptions based on parsing member
"name".
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
Example: the SchemaInfo for ['str']
{ "name": "[str]", "meta-type": "array",
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
"element-type": "str" }
The SchemaInfo for an enumeration type has meta-type "enum" and
qapi-introspect: Document lack of sorting qapi-code-gen.txt already claims that types, commands, and events share a common namespace; set this in stone by further documenting that our introspection output will never have collisions with the same name tied to more than one meta-type. Our largest QMP enum currently has 125 values, our largest object type has 27 members, and the mean for each is less than 10. These sizes are small enough that the per-element overhead of O(log n) binary searching probably outweighs the speed possible with direct O(n) linear searching (a better algorithm with more overhead will only beat a leaner naive algorithm only as you scale to larger input sizes). Arguably, the overall SchemaInfo array could be sorted by name; there, we currently have 531 entities, large enough for a binary search to be faster than linear. However, remember that we have mutually-recursive types, which means there is no topological ordering that will allow clients to learn all information about that type in a single linear pass; thus clients will want to do random access over the data, and they will probably read the introspection output into a hashtable for O(1) lookup rather than O(log n) binary searching, at which point, pre-sorting our introspection output doesn't help the client. It doesn't help that sorting can be subjective if you introduce locales into the mix (I'm not experienced enough with Python to know for sure, but at least it looks like it defaults to sorting in the C locale even when run under a different locale). And while our current introspection output is deterministic (because we visit entities in a sorted order), we may want to change that order in the future (such as using OrderedDict to stick to .json declaration order). For these reasons, we simply document that clients should not rely on any particular order of items in introspection output. And since it is now a documented part of the contract, we have the freedom to later rearrange output if needed, without worrying about breaking well-written clients. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1446791754-23823-13-git-send-email-eblake@redhat.com> [Commit message tweaked] Signed-off-by: Markus Armbruster <armbru@redhat.com>
2015-11-06 07:35:36 +01:00
variant member "values". The values are listed in no particular
order; clients must search the entire enum when learning whether a
particular value is supported.
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
Example: the SchemaInfo for MyEnum from section Enumeration types
{ "name": "MyEnum", "meta-type": "enum",
"values": [ "value1", "value2", "value3" ] }
The SchemaInfo for a built-in type has the same name as the type in
the QAPI schema (see section Built-in Types), with one exception
detailed below. It has variant member "json-type" that shows how
values of this type are encoded on the wire.
Example: the SchemaInfo for str
{ "name": "str", "meta-type": "builtin", "json-type": "string" }
The QAPI schema supports a number of integer types that only differ in
how they map to C. They are identical as far as SchemaInfo is
concerned. Therefore, they get all mapped to a single type "int" in
SchemaInfo.
As explained above, type names are not part of the wire ABI. Not even
the names of built-in types. Clients should examine member
"json-type" instead of hard-coding names of built-in types.
== Code generation ==
Schemas are fed into five scripts to generate all the code/files that,
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
paired with the core QAPI libraries, comprise everything required to
take JSON commands read in by a Client JSON Protocol server, unmarshal
the arguments into the underlying C types, call into the corresponding
C function, map the response back to a Client JSON Protocol response
to be returned to the user, and introspect the commands.
As an example, we'll use the following schema, which describes a
single complex user-defined type, along with command which takes a
list of that type as a parameter, and returns a single element of that
type. The user is responsible for writing the implementation of
qmp_my_command(); everything else is produced by the generator.
$ cat example-schema.json
{ 'struct': 'UserDefOne',
'data': { 'integer': 'int', '*string': 'str' } }
{ 'command': 'my-command',
'data': { 'arg1': ['UserDefOne'] },
'returns': 'UserDefOne' }
{ 'event': 'MY_EVENT' }
For a more thorough look at generated code, the testsuite includes
tests/qapi-schema/qapi-schema-tests.json that covers more examples of
what the generator will accept, and compiles the resulting C code as
part of 'make check-unit'.
=== scripts/qapi-types.py ===
Used to generate the C types defined by a schema, along with
supporting code. The following files are created:
$(prefix)qapi-types.h - C types corresponding to types defined in
the schema you pass in
$(prefix)qapi-types.c - Cleanup functions for the above C types
The $(prefix) is an optional parameter used as a namespace to keep the
generated code from one schema/code-generation separated from others so code
can be generated/used from multiple schemas without clobbering previously
created code.
Example:
$ python scripts/qapi-types.py --output-dir="qapi-generated" \
--prefix="example-" example-schema.json
$ cat qapi-generated/example-qapi-types.h
[Uninteresting stuff omitted...]
#ifndef EXAMPLE_QAPI_TYPES_H
#define EXAMPLE_QAPI_TYPES_H
[Built-in types omitted...]
typedef struct UserDefOne UserDefOne;
typedef struct UserDefOneList UserDefOneList;
struct UserDefOne {
int64_t integer;
bool has_string;
char *string;
};
void qapi_free_UserDefOne(UserDefOne *obj);
struct UserDefOneList {
UserDefOneList *next;
UserDefOne *value;
};
void qapi_free_UserDefOneList(UserDefOneList *obj);
#endif
$ cat qapi-generated/example-qapi-types.c
[Uninteresting stuff omitted...]
qapi-types: Convert to QAPISchemaVisitor, fixing flat unions Fixes flat unions to get the base's base members. Test case is from commit 2fc0043, in qapi-schema-test.json: { 'union': 'UserDefFlatUnion', 'base': 'UserDefUnionBase', 'discriminator': 'enum1', 'data': { 'value1' : 'UserDefA', 'value2' : 'UserDefB', 'value3' : 'UserDefB' } } { 'struct': 'UserDefUnionBase', 'base': 'UserDefZero', 'data': { 'string': 'str', 'enum1': 'EnumOne' } } { 'struct': 'UserDefZero', 'data': { 'integer': 'int' } } Patch's effect on UserDefFlatUnion: struct UserDefFlatUnion { /* Members inherited from UserDefUnionBase: */ + int64_t integer; char *string; EnumOne enum1; /* Own members: */ union { /* union tag is @enum1 */ void *data; UserDefA *value1; UserDefB *value2; UserDefB *value3; }; }; Flat union visitors remain broken. They'll be fixed next. Code is generated in a different order now, but that doesn't matter. The two guards QAPI_TYPES_BUILTIN_STRUCT_DECL and QAPI_TYPES_BUILTIN_CLEANUP_DECL are replaced by just QAPI_TYPES_BUILTIN. Two ugly special cases for simple unions now stand out like sore thumbs: 1. The type tag is named 'type' everywhere, except in generated C, where it's 'kind'. 2. QAPISchema lowers simple unions to semantically equivalent flat unions. However, the C generated for a simple unions differs from the C generated for its equivalent flat union, and we therefore need special code to preserve that pointless difference for now. Mark both TODO. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Daniel P. Berrange <berrange@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:09 +02:00
void qapi_free_UserDefOne(UserDefOne *obj)
{
Visitor *v;
if (!obj) {
return;
}
v = qapi_dealloc_visitor_new();
visit_type_UserDefOne(v, NULL, &obj, NULL);
visit_free(v);
}
qapi-types: Convert to QAPISchemaVisitor, fixing flat unions Fixes flat unions to get the base's base members. Test case is from commit 2fc0043, in qapi-schema-test.json: { 'union': 'UserDefFlatUnion', 'base': 'UserDefUnionBase', 'discriminator': 'enum1', 'data': { 'value1' : 'UserDefA', 'value2' : 'UserDefB', 'value3' : 'UserDefB' } } { 'struct': 'UserDefUnionBase', 'base': 'UserDefZero', 'data': { 'string': 'str', 'enum1': 'EnumOne' } } { 'struct': 'UserDefZero', 'data': { 'integer': 'int' } } Patch's effect on UserDefFlatUnion: struct UserDefFlatUnion { /* Members inherited from UserDefUnionBase: */ + int64_t integer; char *string; EnumOne enum1; /* Own members: */ union { /* union tag is @enum1 */ void *data; UserDefA *value1; UserDefB *value2; UserDefB *value3; }; }; Flat union visitors remain broken. They'll be fixed next. Code is generated in a different order now, but that doesn't matter. The two guards QAPI_TYPES_BUILTIN_STRUCT_DECL and QAPI_TYPES_BUILTIN_CLEANUP_DECL are replaced by just QAPI_TYPES_BUILTIN. Two ugly special cases for simple unions now stand out like sore thumbs: 1. The type tag is named 'type' everywhere, except in generated C, where it's 'kind'. 2. QAPISchema lowers simple unions to semantically equivalent flat unions. However, the C generated for a simple unions differs from the C generated for its equivalent flat union, and we therefore need special code to preserve that pointless difference for now. Mark both TODO. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Daniel P. Berrange <berrange@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:09 +02:00
void qapi_free_UserDefOneList(UserDefOneList *obj)
{
Visitor *v;
if (!obj) {
return;
}
v = qapi_dealloc_visitor_new();
visit_type_UserDefOneList(v, NULL, &obj, NULL);
visit_free(v);
}
=== scripts/qapi-visit.py ===
Used to generate the visitor functions used to walk through and
convert between a native QAPI C data structure and some other format
(such as QObject); the generated functions are named visit_type_FOO()
and visit_type_FOO_members().
The following files are generated:
$(prefix)qapi-visit.c: visitor function for a particular C type, used
to automagically convert QObjects into the
corresponding C type and vice-versa, as well
as for deallocating memory for an existing C
type
$(prefix)qapi-visit.h: declarations for previously mentioned visitor
functions
Example:
$ python scripts/qapi-visit.py --output-dir="qapi-generated"
--prefix="example-" example-schema.json
$ cat qapi-generated/example-qapi-visit.h
[Uninteresting stuff omitted...]
#ifndef EXAMPLE_QAPI_VISIT_H
#define EXAMPLE_QAPI_VISIT_H
[Visitors for built-in types omitted...]
void visit_type_UserDefOne_members(Visitor *v, UserDefOne *obj, Error **errp);
void visit_type_UserDefOne(Visitor *v, const char *name, UserDefOne **obj, Error **errp);
void visit_type_UserDefOneList(Visitor *v, const char *name, UserDefOneList **obj, Error **errp);
#endif
$ cat qapi-generated/example-qapi-visit.c
[Uninteresting stuff omitted...]
void visit_type_UserDefOne_members(Visitor *v, UserDefOne *obj, Error **errp)
{
Error *err = NULL;
visit_type_int(v, "integer", &obj->integer, &err);
qapi: Replace uncommon use of the error API by the common one We commonly use the error API like this: err = NULL; foo(..., &err); if (err) { goto out; } bar(..., &err); Every error source is checked separately. The second function is only called when the first one succeeds. Both functions are free to pass their argument to error_set(). Because error_set() asserts no error has been set, this effectively means they must not be called with an error set. The qapi-generated code uses the error API differently: // *errp was initialized to NULL somewhere up the call chain frob(..., errp); gnat(..., errp); Errors accumulate in *errp: first error wins, subsequent errors get dropped. To make this work, the second function does nothing when called with an error set. Requires non-null errp, or else the second function can't see the first one fail. This usage has also bled into visitor tests, and two device model object property getters rtc_get_date() and balloon_stats_get_all(). With the "accumulate" technique, you need fewer error checks in callers, and buy that with an error check in every callee. Can be nice. However, mixing the two techniques is confusing. You can't use the "accumulate" technique with functions designed for the "check separately" technique. You can use the "check separately" technique with functions designed for the "accumulate" technique, but then error_set() can't catch you setting an error more than once. Standardize on the "check separately" technique for now, because it's overwhelmingly prevalent. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
2014-05-07 09:53:54 +02:00
if (err) {
goto out;
}
if (visit_optional(v, "string", &obj->has_string)) {
visit_type_str(v, "string", &obj->string, &err);
if (err) {
goto out;
}
qapi: Replace uncommon use of the error API by the common one We commonly use the error API like this: err = NULL; foo(..., &err); if (err) { goto out; } bar(..., &err); Every error source is checked separately. The second function is only called when the first one succeeds. Both functions are free to pass their argument to error_set(). Because error_set() asserts no error has been set, this effectively means they must not be called with an error set. The qapi-generated code uses the error API differently: // *errp was initialized to NULL somewhere up the call chain frob(..., errp); gnat(..., errp); Errors accumulate in *errp: first error wins, subsequent errors get dropped. To make this work, the second function does nothing when called with an error set. Requires non-null errp, or else the second function can't see the first one fail. This usage has also bled into visitor tests, and two device model object property getters rtc_get_date() and balloon_stats_get_all(). With the "accumulate" technique, you need fewer error checks in callers, and buy that with an error check in every callee. Can be nice. However, mixing the two techniques is confusing. You can't use the "accumulate" technique with functions designed for the "check separately" technique. You can use the "check separately" technique with functions designed for the "accumulate" technique, but then error_set() can't catch you setting an error more than once. Standardize on the "check separately" technique for now, because it's overwhelmingly prevalent. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
2014-05-07 09:53:54 +02:00
}
qapi: Replace uncommon use of the error API by the common one We commonly use the error API like this: err = NULL; foo(..., &err); if (err) { goto out; } bar(..., &err); Every error source is checked separately. The second function is only called when the first one succeeds. Both functions are free to pass their argument to error_set(). Because error_set() asserts no error has been set, this effectively means they must not be called with an error set. The qapi-generated code uses the error API differently: // *errp was initialized to NULL somewhere up the call chain frob(..., errp); gnat(..., errp); Errors accumulate in *errp: first error wins, subsequent errors get dropped. To make this work, the second function does nothing when called with an error set. Requires non-null errp, or else the second function can't see the first one fail. This usage has also bled into visitor tests, and two device model object property getters rtc_get_date() and balloon_stats_get_all(). With the "accumulate" technique, you need fewer error checks in callers, and buy that with an error check in every callee. Can be nice. However, mixing the two techniques is confusing. You can't use the "accumulate" technique with functions designed for the "check separately" technique. You can use the "check separately" technique with functions designed for the "accumulate" technique, but then error_set() can't catch you setting an error more than once. Standardize on the "check separately" technique for now, because it's overwhelmingly prevalent. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
2014-05-07 09:53:54 +02:00
out:
error_propagate(errp, err);
}
void visit_type_UserDefOne(Visitor *v, const char *name, UserDefOne **obj, Error **errp)
{
qapi: Replace uncommon use of the error API by the common one We commonly use the error API like this: err = NULL; foo(..., &err); if (err) { goto out; } bar(..., &err); Every error source is checked separately. The second function is only called when the first one succeeds. Both functions are free to pass their argument to error_set(). Because error_set() asserts no error has been set, this effectively means they must not be called with an error set. The qapi-generated code uses the error API differently: // *errp was initialized to NULL somewhere up the call chain frob(..., errp); gnat(..., errp); Errors accumulate in *errp: first error wins, subsequent errors get dropped. To make this work, the second function does nothing when called with an error set. Requires non-null errp, or else the second function can't see the first one fail. This usage has also bled into visitor tests, and two device model object property getters rtc_get_date() and balloon_stats_get_all(). With the "accumulate" technique, you need fewer error checks in callers, and buy that with an error check in every callee. Can be nice. However, mixing the two techniques is confusing. You can't use the "accumulate" technique with functions designed for the "check separately" technique. You can use the "check separately" technique with functions designed for the "accumulate" technique, but then error_set() can't catch you setting an error more than once. Standardize on the "check separately" technique for now, because it's overwhelmingly prevalent. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
2014-05-07 09:53:54 +02:00
Error *err = NULL;
visit_start_struct(v, name, (void **)obj, sizeof(UserDefOne), &err);
if (err) {
goto out;
}
if (!*obj) {
goto out_obj;
}
visit_type_UserDefOne_members(v, *obj, &err);
qapi: Split visit_end_struct() into pieces As mentioned in previous patches, we want to call visit_end_struct() functions unconditionally, so that visitors can release resources tied up since the matching visit_start_struct() without also having to worry about error priority if more than one error occurs. Even though error_propagate() can be safely used to ignore a second error during cleanup caused by a first error, it is simpler if the cleanup cannot set an error. So, split out the error checking portion (basically, input visitors checking for unvisited keys) into a new function visit_check_struct(), which can be safely skipped if any earlier errors are encountered, and leave the cleanup portion (which never fails, but must be called unconditionally if visit_start_struct() succeeded) in visit_end_struct(). Generated code in qapi-visit.c has diffs resembling: |@@ -59,10 +59,12 @@ void visit_type_ACPIOSTInfo(Visitor *v, | goto out_obj; | } | visit_type_ACPIOSTInfo_members(v, obj, &err); |- error_propagate(errp, err); |- err = NULL; |+ if (err) { |+ goto out_obj; |+ } |+ visit_check_struct(v, &err); | out_obj: |- visit_end_struct(v, &err); |+ visit_end_struct(v); | out: and in qapi-event.c: @@ -47,7 +47,10 @@ void qapi_event_send_acpi_device_ost(ACP | goto out; | } | visit_type_q_obj_ACPI_DEVICE_OST_arg_members(v, &param, &err); |- visit_end_struct(v, err ? NULL : &err); |+ if (!err) { |+ visit_check_struct(v, &err); |+ } |+ visit_end_struct(v); | if (err) { | goto out; Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1461879932-9020-20-git-send-email-eblake@redhat.com> [Conflict with a doc fixup resolved] Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-04-28 23:45:27 +02:00
if (err) {
goto out_obj;
}
visit_check_struct(v, &err);
out_obj:
2016-06-09 18:48:34 +02:00
visit_end_struct(v, (void **)obj);
qapi: Change visit_type_FOO() to no longer return partial objects Returning a partial object on error is an invitation for a careless caller to leak memory. We already fixed things in an earlier patch to guarantee NULL if visit_start fails ("qapi: Guarantee NULL obj on input visitor callback error"), but that does not help the case where visit_start succeeds but some other failure happens before visit_end, such that we leak a partially constructed object outside visit_type_FOO(). As no one outside the testsuite was actually relying on these semantics, it is cleaner to just document and guarantee that ALL pointer-based visit_type_FOO() functions always leave a safe value in *obj during an input visitor (either the new object on success, or NULL if an error is encountered), so callers can now unconditionally use qapi_free_FOO() to clean up regardless of whether an error occurred. The decision is done by adding visit_is_input(), then updating the generated code to check if additional cleanup is needed based on the type of visitor in use. Note that we still leave *obj unchanged after a scalar-based visit_type_FOO(); I did not feel like auditing all uses of visit_type_Enum() to see if the callers would tolerate a specific sentinel value (not to mention having to decide whether it would be better to use 0 or ENUM__MAX as that sentinel). Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1461879932-9020-25-git-send-email-eblake@redhat.com> Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-04-28 23:45:32 +02:00
if (err && visit_is_input(v)) {
qapi_free_UserDefOne(*obj);
*obj = NULL;
}
out:
qapi: Replace uncommon use of the error API by the common one We commonly use the error API like this: err = NULL; foo(..., &err); if (err) { goto out; } bar(..., &err); Every error source is checked separately. The second function is only called when the first one succeeds. Both functions are free to pass their argument to error_set(). Because error_set() asserts no error has been set, this effectively means they must not be called with an error set. The qapi-generated code uses the error API differently: // *errp was initialized to NULL somewhere up the call chain frob(..., errp); gnat(..., errp); Errors accumulate in *errp: first error wins, subsequent errors get dropped. To make this work, the second function does nothing when called with an error set. Requires non-null errp, or else the second function can't see the first one fail. This usage has also bled into visitor tests, and two device model object property getters rtc_get_date() and balloon_stats_get_all(). With the "accumulate" technique, you need fewer error checks in callers, and buy that with an error check in every callee. Can be nice. However, mixing the two techniques is confusing. You can't use the "accumulate" technique with functions designed for the "check separately" technique. You can use the "check separately" technique with functions designed for the "accumulate" technique, but then error_set() can't catch you setting an error more than once. Standardize on the "check separately" technique for now, because it's overwhelmingly prevalent. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
2014-05-07 09:53:54 +02:00
error_propagate(errp, err);
}
void visit_type_UserDefOneList(Visitor *v, const char *name, UserDefOneList **obj, Error **errp)
{
Error *err = NULL;
qapi: Simplify semantics of visit_next_list() The semantics of the list visit are somewhat baroque, with the following pseudocode when FooList is used: start() for (prev = head; cur = next(prev); prev = &cur) { visit(&cur->value) } Note that these semantics (advance before visit) requires that the first call to next() return the list head, while all other calls return the next element of the list; that is, every visitor implementation is required to track extra state to decide whether to return the input as-is, or to advance. It also requires an argument of 'GenericList **' to next(), solely because the first iteration might need to modify the caller's GenericList head, so that all other calls have to do a layer of dereferencing. Thankfully, we only have two uses of list visits in the entire code base: one in spapr_drc (which completely avoids visit_next_list(), feeding in integers from a different source than uint8List), and one in qapi-visit.py. That is, all other list visitors are generated in qapi-visit.c, and share the same paradigm based on a qapi FooList type, so we can refactor how lists are laid out with minimal churn among clients. We can greatly simplify things by hoisting the special case into the start() routine, and flipping the order in the loop to visit before advance: start(head) for (tail = *head; tail; tail = next(tail)) { visit(&tail->value) } With the simpler semantics, visitors have less state to track, the argument to next() is reduced to 'GenericList *', and it also becomes obvious whether an input visitor is allocating a FooList during visit_start_list() (rather than the old way of not knowing if an allocation happened until the first visit_next_list()). As a minor drawback, we now allocate in two functions instead of one, and have to pass the size to both functions (unless we were to tweak the input visitors to cache the size to start_list for reuse during next_list, but that defeats the goal of less visitor state). The signature of visit_start_list() is chosen to match visit_start_struct(), with the new parameters after 'name'. The spapr_drc case is a virtual visit, done by passing NULL for list, similarly to how NULL is passed to visit_start_struct() when a qapi type is not used in those visits. It was easy to provide these semantics for qmp-output and dealloc visitors, and a bit harder for qmp-input (several prerequisite patches refactored things to make this patch straightforward). But it turned out that the string and opts visitors munge enough other state during visit_next_list() to make it easier to just document and require a GenericList visit for now; an assertion will remind us to adjust things if we need the semantics in the future. Several pre-requisite cleanup patches made the reshuffling of the various visitors easier; particularly the qmp input visitor. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1461879932-9020-24-git-send-email-eblake@redhat.com> Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-04-28 23:45:31 +02:00
UserDefOneList *tail;
size_t size = sizeof(**obj);
qapi: Simplify semantics of visit_next_list() The semantics of the list visit are somewhat baroque, with the following pseudocode when FooList is used: start() for (prev = head; cur = next(prev); prev = &cur) { visit(&cur->value) } Note that these semantics (advance before visit) requires that the first call to next() return the list head, while all other calls return the next element of the list; that is, every visitor implementation is required to track extra state to decide whether to return the input as-is, or to advance. It also requires an argument of 'GenericList **' to next(), solely because the first iteration might need to modify the caller's GenericList head, so that all other calls have to do a layer of dereferencing. Thankfully, we only have two uses of list visits in the entire code base: one in spapr_drc (which completely avoids visit_next_list(), feeding in integers from a different source than uint8List), and one in qapi-visit.py. That is, all other list visitors are generated in qapi-visit.c, and share the same paradigm based on a qapi FooList type, so we can refactor how lists are laid out with minimal churn among clients. We can greatly simplify things by hoisting the special case into the start() routine, and flipping the order in the loop to visit before advance: start(head) for (tail = *head; tail; tail = next(tail)) { visit(&tail->value) } With the simpler semantics, visitors have less state to track, the argument to next() is reduced to 'GenericList *', and it also becomes obvious whether an input visitor is allocating a FooList during visit_start_list() (rather than the old way of not knowing if an allocation happened until the first visit_next_list()). As a minor drawback, we now allocate in two functions instead of one, and have to pass the size to both functions (unless we were to tweak the input visitors to cache the size to start_list for reuse during next_list, but that defeats the goal of less visitor state). The signature of visit_start_list() is chosen to match visit_start_struct(), with the new parameters after 'name'. The spapr_drc case is a virtual visit, done by passing NULL for list, similarly to how NULL is passed to visit_start_struct() when a qapi type is not used in those visits. It was easy to provide these semantics for qmp-output and dealloc visitors, and a bit harder for qmp-input (several prerequisite patches refactored things to make this patch straightforward). But it turned out that the string and opts visitors munge enough other state during visit_next_list() to make it easier to just document and require a GenericList visit for now; an assertion will remind us to adjust things if we need the semantics in the future. Several pre-requisite cleanup patches made the reshuffling of the various visitors easier; particularly the qmp input visitor. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1461879932-9020-24-git-send-email-eblake@redhat.com> Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-04-28 23:45:31 +02:00
visit_start_list(v, name, (GenericList **)obj, size, &err);
qapi: Replace uncommon use of the error API by the common one We commonly use the error API like this: err = NULL; foo(..., &err); if (err) { goto out; } bar(..., &err); Every error source is checked separately. The second function is only called when the first one succeeds. Both functions are free to pass their argument to error_set(). Because error_set() asserts no error has been set, this effectively means they must not be called with an error set. The qapi-generated code uses the error API differently: // *errp was initialized to NULL somewhere up the call chain frob(..., errp); gnat(..., errp); Errors accumulate in *errp: first error wins, subsequent errors get dropped. To make this work, the second function does nothing when called with an error set. Requires non-null errp, or else the second function can't see the first one fail. This usage has also bled into visitor tests, and two device model object property getters rtc_get_date() and balloon_stats_get_all(). With the "accumulate" technique, you need fewer error checks in callers, and buy that with an error check in every callee. Can be nice. However, mixing the two techniques is confusing. You can't use the "accumulate" technique with functions designed for the "check separately" technique. You can use the "check separately" technique with functions designed for the "accumulate" technique, but then error_set() can't catch you setting an error more than once. Standardize on the "check separately" technique for now, because it's overwhelmingly prevalent. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
2014-05-07 09:53:54 +02:00
if (err) {
goto out;
}
qapi: Simplify semantics of visit_next_list() The semantics of the list visit are somewhat baroque, with the following pseudocode when FooList is used: start() for (prev = head; cur = next(prev); prev = &cur) { visit(&cur->value) } Note that these semantics (advance before visit) requires that the first call to next() return the list head, while all other calls return the next element of the list; that is, every visitor implementation is required to track extra state to decide whether to return the input as-is, or to advance. It also requires an argument of 'GenericList **' to next(), solely because the first iteration might need to modify the caller's GenericList head, so that all other calls have to do a layer of dereferencing. Thankfully, we only have two uses of list visits in the entire code base: one in spapr_drc (which completely avoids visit_next_list(), feeding in integers from a different source than uint8List), and one in qapi-visit.py. That is, all other list visitors are generated in qapi-visit.c, and share the same paradigm based on a qapi FooList type, so we can refactor how lists are laid out with minimal churn among clients. We can greatly simplify things by hoisting the special case into the start() routine, and flipping the order in the loop to visit before advance: start(head) for (tail = *head; tail; tail = next(tail)) { visit(&tail->value) } With the simpler semantics, visitors have less state to track, the argument to next() is reduced to 'GenericList *', and it also becomes obvious whether an input visitor is allocating a FooList during visit_start_list() (rather than the old way of not knowing if an allocation happened until the first visit_next_list()). As a minor drawback, we now allocate in two functions instead of one, and have to pass the size to both functions (unless we were to tweak the input visitors to cache the size to start_list for reuse during next_list, but that defeats the goal of less visitor state). The signature of visit_start_list() is chosen to match visit_start_struct(), with the new parameters after 'name'. The spapr_drc case is a virtual visit, done by passing NULL for list, similarly to how NULL is passed to visit_start_struct() when a qapi type is not used in those visits. It was easy to provide these semantics for qmp-output and dealloc visitors, and a bit harder for qmp-input (several prerequisite patches refactored things to make this patch straightforward). But it turned out that the string and opts visitors munge enough other state during visit_next_list() to make it easier to just document and require a GenericList visit for now; an assertion will remind us to adjust things if we need the semantics in the future. Several pre-requisite cleanup patches made the reshuffling of the various visitors easier; particularly the qmp input visitor. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1461879932-9020-24-git-send-email-eblake@redhat.com> Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-04-28 23:45:31 +02:00
for (tail = *obj; tail;
tail = (UserDefOneList *)visit_next_list(v, (GenericList *)tail, size)) {
visit_type_UserDefOne(v, NULL, &tail->value, &err);
if (err) {
break;
}
}
qapi: Replace uncommon use of the error API by the common one We commonly use the error API like this: err = NULL; foo(..., &err); if (err) { goto out; } bar(..., &err); Every error source is checked separately. The second function is only called when the first one succeeds. Both functions are free to pass their argument to error_set(). Because error_set() asserts no error has been set, this effectively means they must not be called with an error set. The qapi-generated code uses the error API differently: // *errp was initialized to NULL somewhere up the call chain frob(..., errp); gnat(..., errp); Errors accumulate in *errp: first error wins, subsequent errors get dropped. To make this work, the second function does nothing when called with an error set. Requires non-null errp, or else the second function can't see the first one fail. This usage has also bled into visitor tests, and two device model object property getters rtc_get_date() and balloon_stats_get_all(). With the "accumulate" technique, you need fewer error checks in callers, and buy that with an error check in every callee. Can be nice. However, mixing the two techniques is confusing. You can't use the "accumulate" technique with functions designed for the "check separately" technique. You can use the "check separately" technique with functions designed for the "accumulate" technique, but then error_set() can't catch you setting an error more than once. Standardize on the "check separately" technique for now, because it's overwhelmingly prevalent. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
2014-05-07 09:53:54 +02:00
2016-06-09 18:48:34 +02:00
visit_end_list(v, (void **)obj);
qapi: Change visit_type_FOO() to no longer return partial objects Returning a partial object on error is an invitation for a careless caller to leak memory. We already fixed things in an earlier patch to guarantee NULL if visit_start fails ("qapi: Guarantee NULL obj on input visitor callback error"), but that does not help the case where visit_start succeeds but some other failure happens before visit_end, such that we leak a partially constructed object outside visit_type_FOO(). As no one outside the testsuite was actually relying on these semantics, it is cleaner to just document and guarantee that ALL pointer-based visit_type_FOO() functions always leave a safe value in *obj during an input visitor (either the new object on success, or NULL if an error is encountered), so callers can now unconditionally use qapi_free_FOO() to clean up regardless of whether an error occurred. The decision is done by adding visit_is_input(), then updating the generated code to check if additional cleanup is needed based on the type of visitor in use. Note that we still leave *obj unchanged after a scalar-based visit_type_FOO(); I did not feel like auditing all uses of visit_type_Enum() to see if the callers would tolerate a specific sentinel value (not to mention having to decide whether it would be better to use 0 or ENUM__MAX as that sentinel). Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1461879932-9020-25-git-send-email-eblake@redhat.com> Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-04-28 23:45:32 +02:00
if (err && visit_is_input(v)) {
qapi_free_UserDefOneList(*obj);
*obj = NULL;
}
qapi: Replace uncommon use of the error API by the common one We commonly use the error API like this: err = NULL; foo(..., &err); if (err) { goto out; } bar(..., &err); Every error source is checked separately. The second function is only called when the first one succeeds. Both functions are free to pass their argument to error_set(). Because error_set() asserts no error has been set, this effectively means they must not be called with an error set. The qapi-generated code uses the error API differently: // *errp was initialized to NULL somewhere up the call chain frob(..., errp); gnat(..., errp); Errors accumulate in *errp: first error wins, subsequent errors get dropped. To make this work, the second function does nothing when called with an error set. Requires non-null errp, or else the second function can't see the first one fail. This usage has also bled into visitor tests, and two device model object property getters rtc_get_date() and balloon_stats_get_all(). With the "accumulate" technique, you need fewer error checks in callers, and buy that with an error check in every callee. Can be nice. However, mixing the two techniques is confusing. You can't use the "accumulate" technique with functions designed for the "check separately" technique. You can use the "check separately" technique with functions designed for the "accumulate" technique, but then error_set() can't catch you setting an error more than once. Standardize on the "check separately" technique for now, because it's overwhelmingly prevalent. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
2014-05-07 09:53:54 +02:00
out:
error_propagate(errp, err);
}
=== scripts/qapi-commands.py ===
Used to generate the marshaling/dispatch functions for the commands
defined in the schema. The generated code implements
qmp_marshal_COMMAND() (registered automatically), and declares
qmp_COMMAND() that the user must implement. The following files are
generated:
$(prefix)qmp-marshal.c: command marshal/dispatch functions for each
QMP command defined in the schema. Functions
generated by qapi-visit.py are used to
convert QObjects received from the wire into
function parameters, and uses the same
visitor functions to convert native C return
values to QObjects from transmission back
over the wire.
$(prefix)qmp-commands.h: Function prototypes for the QMP commands
specified in the schema.
Example:
$ python scripts/qapi-commands.py --output-dir="qapi-generated"
--prefix="example-" example-schema.json
$ cat qapi-generated/example-qmp-commands.h
[Uninteresting stuff omitted...]
#ifndef EXAMPLE_QMP_COMMANDS_H
#define EXAMPLE_QMP_COMMANDS_H
#include "example-qapi-types.h"
#include "qapi/qmp/qdict.h"
#include "qapi/error.h"
UserDefOne *qmp_my_command(UserDefOneList *arg1, Error **errp);
#endif
$ cat qapi-generated/example-qmp-marshal.c
[Uninteresting stuff omitted...]
static void qmp_marshal_output_UserDefOne(UserDefOne *ret_in, QObject **ret_out, Error **errp)
{
Error *err = NULL;
Visitor *v;
v = qobject_output_visitor_new(ret_out);
visit_type_UserDefOne(v, "unused", &ret_in, &err);
qapi: Add new visit_complete() function Making each output visitor provide its own output collection function was the only remaining reason for exposing visitor sub-types to the rest of the code base. Add a polymorphic visit_complete() function which is a no-op for input visitors, and which populates an opaque pointer for output visitors. For maximum type-safety, also add a parameter to the output visitor constructors with a type-correct version of the output pointer, and assert that the two uses match. This approach was considered superior to either passing the output parameter only during construction (action at a distance during visit_free() feels awkward) or only during visit_complete() (defeating type safety makes it easier to use incorrectly). Most callers were function-local, and therefore a mechanical conversion; the testsuite was a bit trickier, but the previous cleanup patch minimized the churn here. The visit_complete() function may be called at most once; doing so lets us use transfer semantics rather than duplication or ref-count semantics to get the just-built output back to the caller, even though it means our behavior is not idempotent. Generated code is simplified as follows for events: |@@ -26,7 +26,7 @@ void qapi_event_send_acpi_device_ost(ACP | QDict *qmp; | Error *err = NULL; | QMPEventFuncEmit emit; |- QmpOutputVisitor *qov; |+ QObject *obj; | Visitor *v; | q_obj_ACPI_DEVICE_OST_arg param = { | info |@@ -39,8 +39,7 @@ void qapi_event_send_acpi_device_ost(ACP | | qmp = qmp_event_build_dict("ACPI_DEVICE_OST"); | |- qov = qmp_output_visitor_new(); |- v = qmp_output_get_visitor(qov); |+ v = qmp_output_visitor_new(&obj); | | visit_start_struct(v, "ACPI_DEVICE_OST", NULL, 0, &err); | if (err) { |@@ -55,7 +54,8 @@ void qapi_event_send_acpi_device_ost(ACP | goto out; | } | |- qdict_put_obj(qmp, "data", qmp_output_get_qobject(qov)); |+ visit_complete(v, &obj); |+ qdict_put_obj(qmp, "data", obj); | emit(QAPI_EVENT_ACPI_DEVICE_OST, qmp, &err); and for commands: | { | Error *err = NULL; |- QmpOutputVisitor *qov = qmp_output_visitor_new(); | Visitor *v; | |- v = qmp_output_get_visitor(qov); |+ v = qmp_output_visitor_new(ret_out); | visit_type_AddfdInfo(v, "unused", &ret_in, &err); |- if (err) { |- goto out; |+ if (!err) { |+ visit_complete(v, ret_out); | } |- *ret_out = qmp_output_get_qobject(qov); |- |-out: | error_propagate(errp, err); Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1465490926-28625-13-git-send-email-eblake@redhat.com> Reviewed-by: Markus Armbruster <armbru@redhat.com> Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-06-09 18:48:43 +02:00
if (!err) {
visit_complete(v, ret_out);
}
error_propagate(errp, err);
visit_free(v);
v = qapi_dealloc_visitor_new();
visit_type_UserDefOne(v, "unused", &ret_in, NULL);
visit_free(v);
}
static void qmp_marshal_my_command(QDict *args, QObject **ret, Error **errp)
{
Error *err = NULL;
UserDefOne *retval;
Visitor *v;
UserDefOneList *arg1 = NULL;
v = qobject_input_visitor_new(QOBJECT(args), true);
qapi-commands: Wrap argument visit in visit_start_struct The qmp-input visitor was allowing callers to play rather fast and loose: when visiting a QDict, you could grab members of the root dictionary without first pushing into the dict; among the culprit callers was the generated marshal code on the 'arguments' dictionary of a QMP command. But we are about to tighten the input visitor, at which point the generated marshal code MUST follow the same paradigms as everyone else, of pushing into the struct before grabbing its keys. Generated code grows as follows: |@@ -515,7 +641,12 @@ void qmp_marshal_blockdev_backup(QDict * | BlockdevBackup arg = {0}; | | v = qmp_input_get_visitor(qiv); |+ visit_start_struct(v, NULL, NULL, 0, &err); |+ if (err) { |+ goto out; |+ } | visit_type_BlockdevBackup_members(v, &arg, &err); |+ visit_end_struct(v, err ? NULL : &err); | if (err) { | goto out; | } |@@ -527,7 +715,9 @@ out: | qmp_input_visitor_cleanup(qiv); | qdv = qapi_dealloc_visitor_new(); | v = qapi_dealloc_get_visitor(qdv); |+ visit_start_struct(v, NULL, NULL, 0, NULL); | visit_type_BlockdevBackup_members(v, &arg, NULL); |+ visit_end_struct(v, NULL); | qapi_dealloc_visitor_cleanup(qdv); | } The use of 'err ? NULL : &err' is temporary; a later patch will clean that up when it splits visit_end_struct(). Prior to this patch, the fact that there was no final visit_end_struct() meant that even though we are using a strict input visit, the marshalling code was not detecting excess input at the top level (only in nested levels). Fortunately, we have code in monitor.c:qmp_check_client_args() that also checks for no excess arguments at the top level. But as the generated code is more compact than the manual check, a later patch will clean up monitor.c to drop the redundancy added here. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1461879932-9020-9-git-send-email-eblake@redhat.com> Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-04-28 23:45:16 +02:00
visit_start_struct(v, NULL, NULL, 0, &err);
if (err) {
goto out;
}
visit_type_UserDefOneList(v, "arg1", &arg1, &err);
qapi: Split visit_end_struct() into pieces As mentioned in previous patches, we want to call visit_end_struct() functions unconditionally, so that visitors can release resources tied up since the matching visit_start_struct() without also having to worry about error priority if more than one error occurs. Even though error_propagate() can be safely used to ignore a second error during cleanup caused by a first error, it is simpler if the cleanup cannot set an error. So, split out the error checking portion (basically, input visitors checking for unvisited keys) into a new function visit_check_struct(), which can be safely skipped if any earlier errors are encountered, and leave the cleanup portion (which never fails, but must be called unconditionally if visit_start_struct() succeeded) in visit_end_struct(). Generated code in qapi-visit.c has diffs resembling: |@@ -59,10 +59,12 @@ void visit_type_ACPIOSTInfo(Visitor *v, | goto out_obj; | } | visit_type_ACPIOSTInfo_members(v, obj, &err); |- error_propagate(errp, err); |- err = NULL; |+ if (err) { |+ goto out_obj; |+ } |+ visit_check_struct(v, &err); | out_obj: |- visit_end_struct(v, &err); |+ visit_end_struct(v); | out: and in qapi-event.c: @@ -47,7 +47,10 @@ void qapi_event_send_acpi_device_ost(ACP | goto out; | } | visit_type_q_obj_ACPI_DEVICE_OST_arg_members(v, &param, &err); |- visit_end_struct(v, err ? NULL : &err); |+ if (!err) { |+ visit_check_struct(v, &err); |+ } |+ visit_end_struct(v); | if (err) { | goto out; Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1461879932-9020-20-git-send-email-eblake@redhat.com> [Conflict with a doc fixup resolved] Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-04-28 23:45:27 +02:00
if (!err) {
visit_check_struct(v, &err);
}
2016-06-09 18:48:34 +02:00
visit_end_struct(v, NULL);
if (err) {
goto out;
}
qapi: Replace uncommon use of the error API by the common one We commonly use the error API like this: err = NULL; foo(..., &err); if (err) { goto out; } bar(..., &err); Every error source is checked separately. The second function is only called when the first one succeeds. Both functions are free to pass their argument to error_set(). Because error_set() asserts no error has been set, this effectively means they must not be called with an error set. The qapi-generated code uses the error API differently: // *errp was initialized to NULL somewhere up the call chain frob(..., errp); gnat(..., errp); Errors accumulate in *errp: first error wins, subsequent errors get dropped. To make this work, the second function does nothing when called with an error set. Requires non-null errp, or else the second function can't see the first one fail. This usage has also bled into visitor tests, and two device model object property getters rtc_get_date() and balloon_stats_get_all(). With the "accumulate" technique, you need fewer error checks in callers, and buy that with an error check in every callee. Can be nice. However, mixing the two techniques is confusing. You can't use the "accumulate" technique with functions designed for the "check separately" technique. You can use the "check separately" technique with functions designed for the "accumulate" technique, but then error_set() can't catch you setting an error more than once. Standardize on the "check separately" technique for now, because it's overwhelmingly prevalent. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
2014-05-07 09:53:54 +02:00
retval = qmp_my_command(arg1, &err);
if (err) {
qapi: Replace uncommon use of the error API by the common one We commonly use the error API like this: err = NULL; foo(..., &err); if (err) { goto out; } bar(..., &err); Every error source is checked separately. The second function is only called when the first one succeeds. Both functions are free to pass their argument to error_set(). Because error_set() asserts no error has been set, this effectively means they must not be called with an error set. The qapi-generated code uses the error API differently: // *errp was initialized to NULL somewhere up the call chain frob(..., errp); gnat(..., errp); Errors accumulate in *errp: first error wins, subsequent errors get dropped. To make this work, the second function does nothing when called with an error set. Requires non-null errp, or else the second function can't see the first one fail. This usage has also bled into visitor tests, and two device model object property getters rtc_get_date() and balloon_stats_get_all(). With the "accumulate" technique, you need fewer error checks in callers, and buy that with an error check in every callee. Can be nice. However, mixing the two techniques is confusing. You can't use the "accumulate" technique with functions designed for the "check separately" technique. You can use the "check separately" technique with functions designed for the "accumulate" technique, but then error_set() can't catch you setting an error more than once. Standardize on the "check separately" technique for now, because it's overwhelmingly prevalent. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
2014-05-07 09:53:54 +02:00
goto out;
}
qmp_marshal_output_UserDefOne(retval, ret, &err);
qapi: Replace uncommon use of the error API by the common one We commonly use the error API like this: err = NULL; foo(..., &err); if (err) { goto out; } bar(..., &err); Every error source is checked separately. The second function is only called when the first one succeeds. Both functions are free to pass their argument to error_set(). Because error_set() asserts no error has been set, this effectively means they must not be called with an error set. The qapi-generated code uses the error API differently: // *errp was initialized to NULL somewhere up the call chain frob(..., errp); gnat(..., errp); Errors accumulate in *errp: first error wins, subsequent errors get dropped. To make this work, the second function does nothing when called with an error set. Requires non-null errp, or else the second function can't see the first one fail. This usage has also bled into visitor tests, and two device model object property getters rtc_get_date() and balloon_stats_get_all(). With the "accumulate" technique, you need fewer error checks in callers, and buy that with an error check in every callee. Can be nice. However, mixing the two techniques is confusing. You can't use the "accumulate" technique with functions designed for the "check separately" technique. You can use the "check separately" technique with functions designed for the "accumulate" technique, but then error_set() can't catch you setting an error more than once. Standardize on the "check separately" technique for now, because it's overwhelmingly prevalent. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
2014-05-07 09:53:54 +02:00
out:
error_propagate(errp, err);
visit_free(v);
v = qapi_dealloc_visitor_new();
qapi-commands: Wrap argument visit in visit_start_struct The qmp-input visitor was allowing callers to play rather fast and loose: when visiting a QDict, you could grab members of the root dictionary without first pushing into the dict; among the culprit callers was the generated marshal code on the 'arguments' dictionary of a QMP command. But we are about to tighten the input visitor, at which point the generated marshal code MUST follow the same paradigms as everyone else, of pushing into the struct before grabbing its keys. Generated code grows as follows: |@@ -515,7 +641,12 @@ void qmp_marshal_blockdev_backup(QDict * | BlockdevBackup arg = {0}; | | v = qmp_input_get_visitor(qiv); |+ visit_start_struct(v, NULL, NULL, 0, &err); |+ if (err) { |+ goto out; |+ } | visit_type_BlockdevBackup_members(v, &arg, &err); |+ visit_end_struct(v, err ? NULL : &err); | if (err) { | goto out; | } |@@ -527,7 +715,9 @@ out: | qmp_input_visitor_cleanup(qiv); | qdv = qapi_dealloc_visitor_new(); | v = qapi_dealloc_get_visitor(qdv); |+ visit_start_struct(v, NULL, NULL, 0, NULL); | visit_type_BlockdevBackup_members(v, &arg, NULL); |+ visit_end_struct(v, NULL); | qapi_dealloc_visitor_cleanup(qdv); | } The use of 'err ? NULL : &err' is temporary; a later patch will clean that up when it splits visit_end_struct(). Prior to this patch, the fact that there was no final visit_end_struct() meant that even though we are using a strict input visit, the marshalling code was not detecting excess input at the top level (only in nested levels). Fortunately, we have code in monitor.c:qmp_check_client_args() that also checks for no excess arguments at the top level. But as the generated code is more compact than the manual check, a later patch will clean up monitor.c to drop the redundancy added here. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <1461879932-9020-9-git-send-email-eblake@redhat.com> Signed-off-by: Markus Armbruster <armbru@redhat.com>
2016-04-28 23:45:16 +02:00
visit_start_struct(v, NULL, NULL, 0, NULL);
visit_type_UserDefOneList(v, "arg1", &arg1, NULL);
2016-06-09 18:48:34 +02:00
visit_end_struct(v, NULL);
visit_free(v);
}
static void qmp_init_marshal(void)
{
qmp_register_command("my-command", qmp_marshal_my_command, QCO_NO_OPTIONS);
}
qapi_init(qmp_init_marshal);
=== scripts/qapi-event.py ===
Used to generate the event-related C code defined by a schema, with
implementations for qapi_event_send_FOO(). The following files are
created:
$(prefix)qapi-event.h - Function prototypes for each event type, plus an
enumeration of all event names
$(prefix)qapi-event.c - Implementation of functions to send an event
Example:
$ python scripts/qapi-event.py --output-dir="qapi-generated"
--prefix="example-" example-schema.json
$ cat qapi-generated/example-qapi-event.h
[Uninteresting stuff omitted...]
#ifndef EXAMPLE_QAPI_EVENT_H
#define EXAMPLE_QAPI_EVENT_H
#include "qapi/error.h"
#include "qapi/qmp/qdict.h"
#include "example-qapi-types.h"
void qapi_event_send_my_event(Error **errp);
typedef enum example_QAPIEvent {
EXAMPLE_QAPI_EVENT_MY_EVENT = 0,
EXAMPLE_QAPI_EVENT__MAX = 1,
} example_QAPIEvent;
extern const char *const example_QAPIEvent_lookup[];
#endif
$ cat qapi-generated/example-qapi-event.c
[Uninteresting stuff omitted...]
void qapi_event_send_my_event(Error **errp)
{
QDict *qmp;
Error *err = NULL;
QMPEventFuncEmit emit;
emit = qmp_event_get_func_emit();
if (!emit) {
return;
}
qmp = qmp_event_build_dict("MY_EVENT");
emit(EXAMPLE_QAPI_EVENT_MY_EVENT, qmp, &err);
error_propagate(errp, err);
QDECREF(qmp);
}
const char *const example_QAPIEvent_lookup[] = {
[EXAMPLE_QAPI_EVENT_MY_EVENT] = "MY_EVENT",
[EXAMPLE_QAPI_EVENT__MAX] = NULL,
};
qapi: New QMP command query-qmp-schema for QMP introspection qapi/introspect.json defines the introspection schema. It's designed for QMP introspection, but should do for similar uses, such as QGA. The introspection schema does not reflect all the rules and restrictions that apply to QAPI schemata. A valid QAPI schema has an introspection value conforming to the introspection schema, but the converse is not true. Introspection lowers away a number of schema details, and makes implicit things explicit: * The built-in types are declared with their JSON type. All integer types are mapped to 'int', because how many bits we use internally is an implementation detail. It could be pressed into external interface service as very approximate range information, but that's a bad idea. If we need range information, we better do it properly. * Implicit type definitions are made explicit, and given auto-generated names: - Array types, named by appending "List" to the name of their element type, like in generated C. - The enumeration types implicitly defined by simple union types, named by appending "Kind" to the name of their simple union type, like in generated C. - Types that don't occur in generated C. Their names start with ':' so they don't clash with the user's names. * All type references are by name. * The struct and union types are generalized into an object type. * Base types are flattened. * Commands take a single argument and return a single result. Dictionary argument or list result is an implicit type definition. The empty object type is used when a command takes no arguments or produces no results. The argument is always of object type, but the introspection schema doesn't reflect that. The 'gen': false directive is omitted as implementation detail. The 'success-response' directive is omitted as well for now, even though it's not an implementation detail, because it's not used by QMP. * Events carry a single data value. Implicit type definition and empty object type use, just like for commands. The value is of object type, but the introspection schema doesn't reflect that. * Types not used by commands or events are omitted. Indirect use counts as use. * Optional members have a default, which can only be null right now Instead of a mandatory "optional" flag, we have an optional default. No default means mandatory, default null means optional without default value. Non-null is available for optional with default (possible future extension). * Clients should *not* look up types by name, because type names are not ABI. Look up the command or event you're interested in, then follow the references. TODO Should we hide the type names to eliminate the temptation? New generator scripts/qapi-introspect.py computes an introspection value for its input, and generates a C variable holding it. It can generate awfully long lines. Marked TODO. A new test-qmp-input-visitor test case feeds its result for both tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a QmpInputVisitor to verify it actually conforms to the schema. New QMP command query-qmp-schema takes its return value from that variable. Its reply is some 85KiBytes for me right now. If this turns out to be too much, we have a couple of options: * We can use shorter names in the JSON. Not the QMP style. * Optionally return the sub-schema for commands and events given as arguments. Right now qmp_query_schema() sends the string literal computed by qmp-introspect.py. To compute sub-schema at run time, we'd have to duplicate parts of qapi-introspect.py in C. Unattractive. * Let clients cache the output of query-qmp-schema. It changes only on QEMU upgrades, i.e. rarely. Provide a command query-qmp-schema-hash. Clients can have a cache indexed by hash, and re-query the schema only when they don't have it cached. Even simpler: put the hash in the QMP greeting. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com>
2015-09-16 13:06:28 +02:00
=== scripts/qapi-introspect.py ===
Used to generate the introspection C code for a schema. The following
files are created:
$(prefix)qmp-introspect.c - Defines a string holding a JSON
description of the schema.
$(prefix)qmp-introspect.h - Declares the above string.
Example:
$ python scripts/qapi-introspect.py --output-dir="qapi-generated"
--prefix="example-" example-schema.json
$ cat qapi-generated/example-qmp-introspect.h
[Uninteresting stuff omitted...]
#ifndef EXAMPLE_QMP_INTROSPECT_H
#define EXAMPLE_QMP_INTROSPECT_H
extern const char example_qmp_schema_json[];
#endif
$ cat qapi-generated/example-qmp-introspect.c
[Uninteresting stuff omitted...]
const char example_qmp_schema_json[] = "["
"{\"arg-type\": \"0\", \"meta-type\": \"event\", \"name\": \"MY_EVENT\"}, "
"{\"arg-type\": \"1\", \"meta-type\": \"command\", \"name\": \"my-command\", \"ret-type\": \"2\"}, "
"{\"members\": [], \"meta-type\": \"object\", \"name\": \"0\"}, "
"{\"members\": [{\"name\": \"arg1\", \"type\": \"[2]\"}], \"meta-type\": \"object\", \"name\": \"1\"}, "
"{\"members\": [{\"name\": \"integer\", \"type\": \"int\"}, {\"default\": null, \"name\": \"string\", \"type\": \"str\"}], \"meta-type\": \"object\", \"name\": \"2\"}, "
"{\"element-type\": \"2\", \"meta-type\": \"array\", \"name\": \"[2]\"}, "
"{\"json-type\": \"int\", \"meta-type\": \"builtin\", \"name\": \"int\"}, "
"{\"json-type\": \"string\", \"meta-type\": \"builtin\", \"name\": \"str\"}]";