| Age | Commit message (Collapse) | Author |
|---|
|
The IPADataSerializer::deserializer attempts to optimise code paths and
remove potentially unused code where multiple File Descriptors were not
expected to be utilised.
The addition of multiple SharedFD entries in the IPC highlights this as
a bug.
Clean up the conditionals to ensure that all File Descriptors are
correctly deserialized.
Bug: https://bugs.libcamera.org/show_bug.cgi?id=205
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Tested-by: Andrey Konovalov <andrey.konovalov@linaro.org>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
Make the local mojom library the first priority in the sys path, to
avoid mixing the local one with the system one in build.
Tested on chromebook soraka-libcamera.
Signed-off-by: Harvey Yang <chenghaoyang@chromium.org>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Currently, enums that are passed between pipeline handlers and their IPA
must be defined in a mojom file. However, there is a use case for
enum/flags to be defined in a C++ header, such that the enum can be used
in a component other than the pipeline handler and its IPA.
To support this, add support for the skipHeader attribute for enums.
Like structs, it is only allowed in core.mojom.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Add Flags<E> as a supported type in the IPA interface.
It is used in mojom with the [flags] attribute. Any field or parameter
type E that is prefixed with the [flags] attribute will direct the code
generator to generate the type name "Flags<E>" and appropriate
serialization/deserialization code for Flags<E> instead of for E.
It is usable and has been tested in struct members, function input and
output parameters, and Signal parameters. This does not add support for
returning Flags as direct return values.
Additionally, the [scopedEnum] attribute can be used on enum
definitions, which will instruct the code generator to convert it to an
enum class instead of a raw enum.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
|
|
There is already support for enums as struct members, but there was no
support for enums in function parameters. Add it.
This does not add support for returning enums as direct return values.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
|
|
It's to fix build errors with mojom in ChromeOS chroot. The reason is that
ChromeOS recently moves native mojom generators module into site-packages,
which has higher import precedence than the local namespace package and shadows
the local package of the same. Add __init__.py to make it an explicit package
to avoid shadowing.
Signed-off-by: Han-Lin Chen <hanlinchen@chromium.org>
Tested-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul ELder <paul.elder@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Now that we have a UniqueFD class, the name FileDescriptor is ambiguous.
Rename it to SharedFD.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Hirokazu Honda <hiroh@chromium.org>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
|
|
IPCUnixSocket::create() creates two file descriptors. One of
them is stored in IPCUnixSocket and the other is returned to a
caller. This clarifies the ownership using UniqueFD.
Signed-off-by: Hirokazu Honda <hiroh@chromium.org>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Hirokazu Honda <hiroh@chromium.org>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
|
|
Remove the verbose #ifndef/#define/#endif pattern for maintaining
header idempotency, and replace it with a simple #pragma once.
This simplifies the headers, and prevents redundant changes when
header files get moved.
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Jean-Michel Hautbois <jeanmichel.hautbois@ideasonboard.com>
|
|
Two independent instances of the ControlSerializer class are in use at
the IPC boundaries, one in the Proxy class that serializes data from the
pipeline handler to the IPA, and one in the ProxyWorker which serializes
data in the opposite direction.
Each instance operates autonomously, without any centralized point of
control, and each one assigns a numerical handle to each ControlInfoMap
it serializes. This creates a risk of potential collision on the handle
values, as both instances will use the same numerical space and
are not aware of what handles has been already used by the instance "on
the other side".
To fix that, partition the handles numerical space by initializing the
control serializer with a seed according to the role of the component
that creates the serializer and increment the handle number by 2, to
avoid any collision risk.
While this is temporary and rather hacky solution, it solves an issue
with isolated IPA modules without too much complexity added.
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
When running the IPA in isolated mode, each side of the IPC boundary
has an instance of the ControlSerializer class which is used to
serializer/deserialize controls before transmitting them on the wire.
The IPAProxyWorker, which creates and manages the process the IPA runs in,
does not reset its ControlSerializer upon an IPA::configure() call, while
the IPAProxy does, effectively creating a misalignment between the
two sides of the fence.
This obviously creates issues as one side of the IPC runs with a
populated and possibly stale cache of ControlInfoMap references, while the
other side gets reset every time a new configuration is applied to the
Camera.
Fix that by resetting the IPAProxyWorker ControlSerializer on an
IPA configure() call.
This change fixes an issue which is easily triggered by running two
consecutive capture sessions with the IPA running in isolated mode:
ERROR Serializer control_serializer.cpp:520 Can't deserialize ControlList: unknown ControlInfoMap
Fixes: 7832e19a599e ("utils: ipc: add templates for code generation for IPC mechanism")
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Reviewed-by: Umang Jain <umang.jain@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
The indentation of the deserialization call on the proxy worker side
inside the case statement was one level too shallow. Fix it.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Many signals used in internal and public APIs carry the emitter pointer
as a signal argument. This was done to allow slots connected to multiple
signal instances to differentiate between emitters. While starting from
a good intention of facilitating the implementation of slots, it turned
out to be a bad API design as the signal isn't meant to know what it
will be connected to, and thus shouldn't carry parameters that are
solely meant to support a use case specific to the connected slot.
These pointers turn out to be unused in all slots but one. In the only
case where it is needed, it can be obtained by wrapping the slot in a
lambda function when connecting the signal. Do so, and drop the emitter
pointer from all signals.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Umang Jain <umang.jain@ideasonboard.com>
|
|
In many cases, the emitter object passed as a pointer from signals to
slots is also available as a class member. Use the class member when
this occurs, to prepare for removal of the emitter object pointer from
signals.
In test/event.cpp, this additionally requires moving the EventNotifier
to a class member.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Umang Jain <umang.jain@ideasonboard.com>
|
|
Regarding (de)serialization in isolated IPA calls, we have four layers:
- struct
- byte vector + fd vector
- IPCMessage
- IPC payload
The proxy handles the upper three layers (with help from the
IPADataSerializer), and passes an IPCMessage to the IPC mechanism
(implemented as an IPCPipe), which sends an IPC payload to its worker
counterpart.
When a FileDescriptor is involved, previously it was only a
FileDescriptor in the first layer; in the lower three it was an int. To
reduce the risk of potential fd leaks in the future, keep the
FileDescriptor as-is throughout the upper three layers. Only the IPC
mechanism will deal with ints, if it so wishes, when it does the actual
IPC. IPCPipeUnixSocket does deal with ints for sending fds, so the
conversion between IPCMessage and IPCUnixSocket::Payload converts
between FileDescriptor and int.
Additionally, change the data portion of the serialized form of
FileDescriptor to a 32-bit unsigned integer, for alightnment purposes
and in preparation for conversion to an index into the fd array.
Also update the deserializer of FrameBuffer::Plane accordingly.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Tested-by: Umang Jain <umang.jain@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Umang Jain <umang.jain@ideasonboard.com>
|
|
The IPCUnixSocket::send() function may fail, in which case it can be
useful for debugging to log an error message that tells which event was
affected. Do so.
Reported-by: Coverity CID=35483[6-9]
Reported-by: Coverity CID=35484[01]
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Umang Jain <umang.jain@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
The ThreadProxy IPA template does not implement a constructor and the
default compiler generated constructor does not initialise the private
ipa_ pointer.
Whilst this should not be expected to be used while uninitialised, it
does get caught by static analysis for every IPA module constructed, so
lets be clean and fix it.
Reported-by: Coverity CID=350116
Reported-by: Coverity CID=350123
Reported-by: Coverity CID=350140
Reported-by: Coverity CID=350147
Fixes: 7832e19a599e ("utils: ipc: add templates for code generation for IPC mechanism")
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Umang Jain <umang.jain@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
Isolated IPAs are forked to a new process by the proxy worker, which
shares the same process group. This allows the undesired effect that
the proxy worker will receive signals such as SIGINT and will be closed
by a Ctrl-C event before the pipeline handlers have been able to fully
clean up.
Prevent this signal from being delivered to the proxy worker by moving
the process to a new process group, matching the pid of the isolated
proxy.
Bug: https://bugs.libcamera.org/show_bug.cgi?id=60
Tested-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Signed-off-by: Umang Jain <umang.jain@ideasonboard.com>
|
|
ControlSerializer should be reset during IPA (re)configuration,
so that it doesn't look up stale deserialized cache built from
consecutive previous runs. This is already recommended in
ControlSerializer docs but the implementation seems missing.
The stale cache lookup seems to the core issue with Bug #58.
Bug: https://bugs.libcamera.org/show_bug.cgi?id=58
Signed-off-by: Umang Jain <umang.jain@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Tested-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
Move the functionality for the following components to the new
base support library:
- BoundMethod
- EventDispatcher
- EventDispatcherPoll
- Log
- Message
- Object
- Signal
- Semaphore
- Thread
- Timer
While it would be preferable to see these split to move one component
per commit, these components are all interdependent upon each other,
which leaves us with one big change performing the move for all of them.
Reviewed-by: Hirokazu Honda <hiroh@chromium.org>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
The mojo parser is fine if there are types that are used in array/map
members that it does not know about. These are usually caught by the C++
compiler, because the generated code refers to unknown types. This
feature is necessary for us for supporting FrameBuffer::Plane as an
array/map member, since as long as the type has an IPADataSerializer and
the struct defined in C++, the generated code will run fine
(FrameBuffer::Plane is not defined anywhere in mojom but is used as an
array member in IPABuffer).
The types that are defined in controls.h (or any header included in
ipa_interface.h) will all be compiled by the C++ compiler fine, since
the generated files all include controls.h. The types that are there
that are not ControlInfoMap or ControlList (like ControlValue) will
still fail at the linker stage. For example:
struct A {
array<ControlValue> a;
};
will compile fine, but will fail to link, since
IPADataSerializer<ControlValue> doesn't exist. This behavior, although
not the best, is acceptable.
The issue is that if ControlInfoMap or ControlList are used as array/map
members without the libcamera prefix, the compiler will not complain, as
the types are valid, and the linker will also not complain, as
IPADataSerializer<ControlList> and IPADataSerializer<ControlInfoMap>
both exist. However, the code generator will not recognize them as
types that require a ControlSerializer (since mojo doesn't recognize
them, so they are different from the ones that it does recognize with
the libcamera namespace), and so the ControlSerializer will not be
passed to the serializer in the generated code. This is the cause of the
FATAL breakage:
FATAL IPADataSerializer ipa_data_serializer.cpp:437 ControlSerializer
not provided for serialization of ControlInfoMap
Since ControlInfoMap and ControlList are the only types that will run
into this issue, we solve this by simply detecting if they are used
without the prefix, and produce an error at that point in the code
generator. As the code generator stage no longer has information on the
source code file and line, we output the struct name in which the error
was found (ninja will output the file name).
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Take the SPDX header from the mojom file. Error out if the mojom file
has no SPDX header.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Umang Jain <umang.jain@ideasonboard.com>
|
|
Add a script to extract doxygen documentation comments from mojom files.
It matches based on ^\/\*\*$ for start of block and ^ \*\/$ for end of
block, and simply copies the comments to the output file along with a
header and the libcamera namespace.
Also add it to the meson file so it is usable by other meson files.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Acked-by: Umang Jain <umang.jain@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Update mojo from the Chromium repository. The commit from which this was
taken is:
9c138d992bfc1fb8f4f7bcf58d00bf19c219e4e2 "Updating trunk VERSION from
4523.0 to 4524.0"
The update-mojo.sh script was used for this update.
Bug: https://bugs.libcamera.org/show_bug.cgi?id=34
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Moving the core.mojom documentation to its corresponding .cpp file
(core_ipa_interface.cpp). This will allow Doxygen to generate the
documentation for IPABuffer, IPASettings and IPAStream structures.
Since the .mojom files are placed in include/ directory, the .cpp file
will live in $sourcedir/src/libcamera/ipa/ - which can also contain
documentation for other mojom generated IPA interfaces in subsequent
commit.
Also hide the constructors in generated IPA interface from doxygen,
via #ifndef __DOXYGEN__. These constructors provide no major value in
documenting them, instead will spew out doxygen warnings during the
build.
Signed-off-by: Umang Jain <umang.jain@ideasonboard.com>
Acked-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Structs defined in mojom previously used the namespace of the mojom file
that was being used as the source. This is obviously not the correct
namespace for structs that are defined in core.mojom. Fix the jinja
function for getting the element type including namespace, and use it.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Asynchronous tasks can only be submitted while the IPA is running.
Further more, the shutdown sequence can not be tracked with a simple
running flag. We can also be in the state 'Stopping' where we have not
yet completed all events, but we must not commence anything new.
Refactor the running_ boolean into a stateful enum to track this.
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
Events may be queued to the pipeline handler between the pipeline
handler entering the ::stop() function, and before the call to stop the
IPA has completed.
Handle these events by dispatching all pending messages at the proxy
after the IPA has fully stopped.
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
Signals and calls from the IPA should not occur after the IPA has been
put into the stopped state.
Add assertions to catch and prevent any messages being processed after
this.
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
To make it more convenient for synchronous IPA calls to return a status,
convert the first output into a direct return if it is an int32.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Naushir Patuck <naush@raspberrypi.com>
|
|
Add support to the mojom-based code generator for custom parameters to
init(). Remove the parameter type and count validation as well.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
The has_input variable is unused, drop it.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
|
|
Run mojo from meson to generate the header, serializer, and proxy files
for every pipeline's mojom data definition file.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
|
|
On some systems, python2 might still be the default python. Enforce
python3 by wrapping the mojo parser script in a python3 script. This
also has the benefit of not modifying mojo.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
|
|
We want to avoid changing our copy of mojo to make updates easier. Some
parameters in the mojo generator script needs to be changed though; add
a wrapper script that sets these parameters.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
|
|
Add templates to mojo to generate code for the IPC mechanism. These
templates generate:
- module header
- module serializer
- IPA proxy cpp, header, and worker
Given an input data definition mojom file for a pipeline.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Acked-by: Jacopo Mondi <jacopo@jmondi.org>
Acked-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Specify in the readme and meson file that we depend on python3-ply and
python3-jinja2 for generating the IPA interface.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
Import mojo from the Chromium repository, so that we can use it for
generating code for the IPC mechanism. The commit from which this was
taken is:
a079161ec8c6907b883f9cb84fc8c4e7896cb1d0 "Add PPAPI constructs for
sending focus object to PdfAccessibilityTree"
This tree has been pruned to remove directories that didn't have any
necessary code:
- mojo/* except for mojo/public
- mojo core, docs, and misc files
- mojo/public/* except for mojo/public/{tools,LICENSE}
- language bindings for IPC, tests, and some mojo internals
- mojo/public/tools/{fuzzers,chrome_ipc}
- mojo/public/tools/bindings/generators
- code generation for other languages
No files were modified.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Acked-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
