Restructure the kernel's object tracing implementation
to provide a public API that allows debug tools to use
the debug hooks easier and allows kernel developers
to extend the kernel's object tracing scope and include
new kernel objects easier.
The API provides the trace list abstraction to keep track
of different types of kernel objects. The API contains
a simple single-linked list implementation that allows
to save space and simplifies the access to the data for
debug tools such as gdb.
Change-Id: Ic4d393d584576f67f2c5b706e61bae08869debba
Signed-off-by: Juan Manuel Cruz <juan.m.cruz.alcaraz@intel.com>
Specifying HEAP_SIZE keyword in an app's MDEF file, results in
creating a new memory pool, which can be accessed using the
task_malloc() and task_free() APIs, which have the usual malloc/free
like semantics.
Expected format in MDEF file
HEAP_SIZE <value>
Change-Id: I0569cffeecf8a2c23c20c7b359256123ece91982
Signed-off-by: Jithu Joseph <jithu.joseph@intel.com>
Adds C++ support to the build system.
Change-Id: Ice1e57a13598e7a48b0bf3298fc318f4ce012ee6
Signed-off-by: Dmitriy Korovkin <dmitriy.korovkin@windriver.com>
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
While adding private object supports, there were lots of
copy and paste. This results in lines writing to kernel_main.c
while it should be writing to sysgen.h. So update those lines
to write to sysgen.h instead, and add more new lines there
so sysgen.h is easier to read.
Change-Id: I38496b58d6dc4ed8a50eab4e09ac7bff8c59026f
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
This patch enables defining microkernel events within source code.
This is similar to other private kernel object patches.
The test has been modified a little bit due to the fact that
the event ID is now a memory address, instead of numeric ID.
Change-Id: Ie3c8d4f4e459d9c631e50bb242cf7a05ca8ea82c
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
use zephyr.h as the main include in applications, no need to have nano/micro
includes in applications and samples.
Inclusion of the proper kernel headers is be handled in the zephyr.h based
on the configured and used kernel.
Change-Id: If5275cef5d2ad1f475dfb39102cb71cfe5630f6c
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Add a list for each kind of microkernel object. Add the register of
the public objects to the tracing lists in the dynamic initialization
function of sysgen.
Change-Id: I8c413812e4db0f443c9dd3a5501e0096270dc70e
Signed-off-by: Yonattan Louise <yonattan.a.louise.mendoza@intel.com>
rename the sections macro to avoid conflicts with the same
macro being defined in sys/cdefs.h of an external libc.
Change-Id: I4d9e060eeff788ca4112c0ad3e98f0bea135f145
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Change all the Intel and Wind River code license from BSD-3 to Apache 2.
Change-Id: Id8be2c1c161a06ea8a0b9f38e17660e11dbb384b
Signed-off-by: Javier B Perez Hernandez <javier.b.perez.hernandez@linux.intel.com>
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
Revises microkernel initialization code by incorporating the existing
_k_kernel_init() and main() routines into _main(). This optimizes kernel
initialization a bit, and allows application code to use main()
if desired -- for example, as the entry point to an application task.
The change also eliminates the need for sysgen to generate a routine
whose content is always the same.
This change preserves the existing order of operations done during
kernel initialization, and leaves further improvements for later.
Change-Id: Ie03d8a6f38f8a311f398667ed977fd8478719d70
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Due to similar padding issue as pipe, the list of task object
may not be used directly. As mentioned before, some compiler/linker
may pad the large struct. For example, compiling under gcc and
march=i686 pads the struct to 32-byte alignment (march=atom to
64-byte alignment). This causes issue with sizeof() and pointer
arithmetic because they have no idea about the padding.
When the stars align in a certain way, these task structs may be
corrupted. So add a task pointer list and use it for task
manipulation. The task list remains as it is beneficial to group
them together to take advantage of cache locality.
Change-Id: I0e86bfe05742040f4540d7854c1ac14e76162776
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
It was badly named: this routine initializes what cannot be done
statically, i.e. what has to be done dynamically.
We also do not have concepts of "node" anymore since the kernel is
single-node only.
Change-Id: Idde3183ca01ca3f70c5ae0e948734a965582ded8
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
k_proc was a legacy leftover that does not make any sense, since there
is no concept of a "process" in the system. Rather, that data structure
refers to a 'task control block', i.e. the representation of a task
execution context from the microkernel's point-of-view (not to confuse
with the 'struct ccs', the representation of a thread execution context,
from the nanokernel's point-of-view).
Change-Id: Ic29db565af023be629ce740bbcb652ece7dc359f
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
This enable defining memory maps in source code in addition to
defining in MDEF files. This introduces the macro
DEFINE_MEM_MAP(mem_map_name, ...). The memory maps created this
way are the same, in functionality, as those defined in MDEF
files. They can be manipulated by the standard microkernel
memory map APIs.
Define the memory map using:
DEFINE_MEM_MAP(mem_map1, blocks, block_size);
and "mem_map1" can be used, for example:
task_mem_map_alloc(mem_map1, ...);
or,
task_mem_map_free(mem_map1, ...);
etc.
To use the memory map defined in another source file, simply add:
extern const kmemory_map_t mem_map1;
to the desired C or header file.
Change-Id: I9c551b90f9d0a95f961fd8ec1c5278c2ea44312d
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
The _k_mem_map_list was a static array generated by sysgen,
where it containing all pre-defined memory maps from MDEF file.
To support private memory map objects (aka, defining them within
source files), the list has to accommodate memory maps that
are not only processed through sysgen, but also those defined
within source files.
This is done by creating a new section in binary, and all memory
map pointers go into this section. By doing this, the list
can still be manipulated as an array.
Change-Id: I1f3414b72f685fef4b99850749178661f14d9345
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
The opaque memory map object id type is now a pointer to
the associated memory map structure, rather than an index
into the microkernel's array of memory map structures.
This change is a pre-requisite to support for private
memory map objects, which are defined in source code.
Change-Id: I82ecb59eeed00efa54f781f775710c92ff9c9fc9
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
This enable defining pipes in source code in addition to
defining in MDEF files. This introduces the macro
DEFINE_PIPE(pipe_name, ...). The pipes created this
way are the same, in functionality, as those defined in MDEF
files. They can be manipulated by the standard microkernel
pipe APIs.
Define the pipe using:
DEFINE_PIPE(pipe1, size);
and "pipe1" can be used, for example:
task_pipe_put(pipe1, ...);
or,
task_pipe_get(pipe1, ...);
etc.
To use the pipe defined in another source file, simply add:
extern const kpipe_t pipe1;
to the desired C or header file.
Change-Id: Iae8e04706359bc18aae51acc75df3e3d26388882
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
The _k_pipe_list was a static array generated by sysgen,
where it containing all pre-defined pipes from MDEF file.
To support private pipe objects (aka, defining pipes within
source files), the pipe list has to accommodate pipes that
are not only processed through sysgen, but also those defined
within source files.
This is done by creating a new section in binary, and all pipe
pointers go into this section. By doing this, the pipe list
can still be manipulated as an array. The reason behind
putting the pointers to pipe, instead of the pipe objects
themselves, is that some compiler/linker may pad the large pipe
struct. For example, compiling under gcc and march=i686 pads
the struct to 32-byte alignment (march=atom to 64-byte alignment).
This causes issue with sizeof() and pointer arithmetic because
they have no idea about the padding. So use pointers here to
prevent padding.
Change-Id: I6d3b75614c4d8760c037a5c26746410d4e4b17cb
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
The opaque pipe object id type is now a pointer to
the associated pipe structure, rather than an index
into the microkernel's array of pipe structures.
This change is a pre-requisite to support for private
pipes, which are defined in source code.
This also moves the required struct into more visible
headers such that private pipes can be declared.
Renaming the struct is also being done to conform to
naming convention for private kernel objects.
Since a couple structs have to be moved anyway, so
do the moving and renaming here too (contrary to what
have been done in the past, with separated patches).
Change-Id: Ibb6ec7f62745a81439ae3ea2616688b757439843
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
The _k_task_list was a static array generated by sysgen,
where it containing all pre-defined tasks from MDEF file.
To support private task objects (aka, defining tasks within
source files), the task list has to accommodate tasks that
are not only processed through sysgen, but also those defined
within source files.
This is done by creating a new section in binary, and all task
objects go into this section. By doing this, the task list
can still be manipulated as an array, which is required for
task group operation.
Change-Id: I799d6967567079498bc414e0cb809e8af856b53e
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
The opaque task object id type is now a pointer to
the associated task structure, rather than an index
into the microkernel's array of task structures.
This change is a pre-requisite to support for private
task objects, which are defined in source code.
Change-Id: Idb53ea7f8a8a5b7e6477a74273930b08fc77dcfe
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
This enable defining mailboxes in source code in addition to
defining in MDEF files. This introduces the macro
DEFINE_MAILBOX(mailbox_name). The mailboxes created this
way are the same, in functionality, as those defined in MDEF
files. They can be manipulated by the standard microkernel
mailbox APIs.
Define the mailbox using:
DEFINE_MAILBOX(mailbox1);
and "mailbox1" can be used, for example:
task_mbox_put(mailbox1, &msg);
or,
task_mbox_get(mailbox1, &msg);
etc.
To use the mailbox defined in another source file, simply add:
extern const kmox_t mailbox1;
to the desired C or header file.
Change-Id: I32d4194af740f96fc8df8c6cd3144a0c9accb4ec
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
This is in preparation to enable private mailbox support.
The struct needs to be moved into more visible header
so private mailboxes can be declared. This also renames
according to naming convention to be private kernel
objects.
Change-Id: Ibd75497e726efd447d27f3bfd0b4695ed1695693
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
The opaque mailbox object id type is now a pointer to
the associated mailbox structure, rather than an index
into the microkernel's array of mailbox structures.
This change is a pre-requisite to support for private
mailboxes, which are defined in source code.
Change-Id: Ide832eee2a0762e601847ad07afba380bd79ed8b
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
This enable defining FIFOs in source code in addition to
defining in MDEF files. This introduces the macro
DEFINE_FIFO(fifo_name). The FIFOs created this
way are the same, in functionality, as those defined in MDEF
files. They can be manipulated by the standard microkernel
FIFO APIs.
Define the FIFO using:
DEFINE_FIFO(fifo1, depth, width);
and "fifo1" can be used, for example:
task_fifo_put(fifo1);
or,
task_fifo_get(fifo1);
etc.
To use the FIFO defined in another source file, simply add:
extern const kfifo_t fifo1;
to the desired C or header file.
Change-Id: I29667c4cfdcd0e6d189358478acf1a6149ecb826
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
This is in preparation to enable private FIFO support.
The struct needs to be moved into more visible header
so private FIFOs can be declared. This also renames
according to naming convention to be private kernel
objects.
Change-Id: I9b90ddccbaf01ff8c7e2ef03c926d0328dd7ec39
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
The opaque FIFO object id type is now a pointer to
the associated FIFO structure, rather than an index
into the microkernel's array of FIFO structures.
This change is a pre-requisite to support for private
FIFOs, which are defined in source code.
Change-Id: Ieb1343d6d5dd1b747063603457d47fab2710f557
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
This enable defining semaphores in source code in addition to
defining in MDEF files. This introduces the macro
DEFINE_SEMAPHORE(semaphore_name). The semaphores created this
way are the same, in functionality, as those defined in MDEF
files. They can be manipulated by the standard microkernel
semaphore APIs.
Define the semaphore using:
DEFINE_SEMAPHORE(sem1);
and "sem1" can be used, for example:
task_sem_give(sem1);
or,
task_sem_take(sem1);
etc.
To use the semaphore defined in another source file, simply add:
extern const ksem_t sem11;
to the desired C or header file.
Change-Id: Ia9c128648f7a26fc776eeb8f6fcc4158b1712a97
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
This is in preparation to enable private semaphore support.
The struct needs to be moved into more visible header
so private semaphores can be declared. This also renames
according to naming convention to be private kernel
objects.
Change-Id: I84ac7d580404ac5e1753c1bd446d38993fd23310
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
The opaque semaphore object id type is now a pointer to
the associated semaphore structure, rather than an index
into the microkernel's array of semaphore structures.
This change is a pre-requisite to support for private
semaphores, which are defined in source code.
Change-Id: I3821360be35237bfe3bf090efce84f99e335d309
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
It is no longer necessary to generate variables containing the
object id each public mutex. This capability was once needed to
allow kernel code to reference mutexes generated by the build
system (i.e. pseudo-private mutexes); however, the existence of
private mutexes renders this obsolete.
Change-Id: I594be31268c9a336ae2a43d48c1b32dbf3d7cde0
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
This enable defining mutexes in source code in addition to
defining in MDEF files. This introduces the macro
DEFINE_MUTEX(mutex_name). The mutexes created this way are
the same, in functionality, as those defined in MDEF files.
They can be manipulated by the standard microkernel
mutex APIs.
Sample usage:
DEFINE_MUTEX(mutex1);
void one_function(void)
{
task_mutex_lock_wait(mutex1);
...
task_mutex_unlock(mutex1);
}
To use the mutex defined in another source file, simply add:
extern const kmutex_t mutex1;
to the desired C or header file.
[ DL: This is a slightly modified version of similar patch
by Allan Stephens available at
https://oic-review.01.org/gerrit/3255 ]
Change-Id: Ib9cd8193eaf849a8aad1d217912759324ee8818e
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
This is in preparation to enable private mutex support.
The struct needs to be moved into more visible header
so private mutexes can be declared. This also renames
according to naming convention to be private kernel
objects.
Change-Id: Ifccb60a837b44e443be0b091c2df4f06373718fe
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
The opaque mutex object id type is now a pointer to the associated
mutex structure, rather than an index into the microkernel's
array of mutex structures.
This change is a pre-requisite to support for private mutexes
(i.e. mutexes not defined using a project's MDEF file).
[DL: fix some whitespace issues.]
Change-Id: Ida419f1674df245f51a36d28ca7d4631239f1edd
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
This change removes the internal number-to-function mapping
of microkernel services. Instead, function pointers are used
to specify which service to use.
This is in preparation for private kernel objects. Before this,
only kernel objects that are defined in MDEF files would have
corresponding functions included in the final binary, via sysgen
by populating an array of number-to-function mapping. This
causes an issue when a certain type of objects are all defined
with source code, and never in MDEF file. The corresponding
mapping would be deleted, and the functions are never included
in the binary. For example, if no mutexes are defined in MDEF
file, the _k_mutex_*() functions would not be included.
With this change, any usage of private kernel objects will hint
to the linker that those functions are needed, and should not be
removed from final binary.
Change-Id: If48864abcd6471bcb7964ec00fe668bcabe3239b
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Update microkernel systems to use the new driver initialization model on the
timer driver.
Change-Id: Ida9ef2a395d0dddf4104d490d78b13b11ea3c347
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
Prepares the _sys_clock_driver_init() routine to interface with the new driver
model. To this end, it must accept a pointer to a device structure and return
an integer.
Change-Id: I3c600ce1efb49c0864aae7b09663ca40d6289372
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
The USERDRIVER line is currently unused, and recent driver
initialization enhancements have rendered it obsolete.
Change-Id: If14089a52871a8891223156fde99b3fc9b650b10
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
SysGen generates code to invoke _sys_clock_driver_init() when
CONFIG_SYS_CLOCK_EXISTS is defined thereby reducing the number
items the user must remember to configure.
Change-Id: I6f6909378c8a865a6ef50a99179800d58be5dd68
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
This private API now follows Zephyr OS naming conventions.
Change-Id: Ic8b1d4673862263d90d240e749a272d01a53679e
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
These private APIs now follow Zephyr OS naming conventions.
Change-Id: Ia11a082e46456d12a3d6af7dfb2f34c29c137cb2
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
These private APIs now follow Zephyr OS naming conventions.
Change-Id: I63815013be2d3b01463563d17f05aac28b27ce04
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
The new name better reflects that this file contains private
microkernel APIs (types, structs, etc.) that are used by various
kernel subsystems.
Change-Id: I7a89be893455b3daaf30baa40a0ec8e0cde7cc36
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Gets rid of the Python file extension, since the fact the code is
written in this language could conceivably change someday (and
doesn't really need to be advertised to anyone invoking it).
Change-Id: I0d01a89fd25e4a0f033099a4bcea674bc302c658
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Juan Manuel Cruz