Let's introduce a new option so components which cannot
be used with the native libC can select it.
And at this point let's already have POSIX_API select it
instead of being used in the libC choice default.
Signed-off-by: Alberto Escolar Piedras <alberto.escolar.piedras@nordicsemi.no>
Based on Andy's talk at eoss 2024, use the sys/sem.h api instead
of the spinlock.h api to synchronize pooled elements since it
has minimal overhead like semaphores but also works from
userspace.
Signed-off-by: Chris Friedt <cfriedt@tenstorrent.com>
ftruncate should return -1 on errors and it needs to
set appropriate error values in errno
Signed-off-by: Karthikeyan Krishnasamy <karthikeyan@linumiz.com>
Commit e6eb0a705b ("posix: eventfd: revise locking, signaling, and
allocation") introduced a regression where the internal flags of an
event file descriptor would be erased when calling the F_SETFL ioctl
operation.
This includes the flag EFD_IN_USE_INTERNAL which determines whether
this file descriptor has been opened, thus effectively closing the
eventfd whenever one tries to change a flag.
Signed-off-by: Celina Sophie Kalus <hello@celinakalus.de>
The `BUILD_VERSION` can be defined but empty when built
without git, causing version to be missing from the uname
version string:
```
*** Booting Zephyr OS build 3.5.0 ***
Printing everything in utsname...
sysname[7]: Zephyr
nodename[7]: zephyr
release[13]: 3.5.0
version[61]: Apr 1 2024 23:48:30
machine[8]: riscv64
uart:~$
```
Let's check if it is empty before using it, so that
`KERNEL_VERSION_STRING`, which is generated independently
with cmake can be used as a fallback:
```
*** Booting Zephyr OS build 3.5.0 ***
Printing everything in utsname...
sysname[7]: Zephyr
nodename[7]: zephyr
release[13]: 3.5.0
version[61]: 3.5.0 Apr 1 2024 23:53:48
machine[8]: riscv64
uart:~$
```
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
With CONFIG_NET_POSIX_NAMES being deprecated, convert static
inlines in headers to prototypes, and move implementations
to lib/posix/options/net.c .
Since select and poll should technically also operate on
non-socket file descriptors, these may be relocated in the
future.
Below are some reasons for this change.
1. So posix calls are regular symbols and not directly inlined
as syscalls. This is also the most portable linkage type.
2. Many posix calls are cancellation points and additional
checks are necessary within the implementation.
3. The more we add to an inline call, the less sense it makes
for that call to be inline.
4. When the implementation of a posix function changes, the
interface for it should still remain consistent.
Note: the same principles have been applied in Zephyr's
posix implementation already, e.g. clock_gettime(), so this
is nothing new.
Signed-off-by: Chris Friedt <cfriedt@tenstorrent.com>
Implement `pthread_attr_setinheritsched()` and
`pthread_attr_getinheritsched()`are required
as part of _POSIX_THREAD_PRIORITY_SCHEDULING Option Group.
signed-off-by: Gaetan Perrot <gaetanperrotpro@gmail.com>
Implement if_indextoname(), if_freenameindex(), if_nameindex(),
and if_nametoindex() by wrapping around networking subystem
calls.
Signed-off-by: Chris Friedt <chrisfriedt@gmail.com>
Implement pthread_rwlockattr_setpshared() and
pthread_rwlockattr_getpshared().
Both functions are required by the _POSIX_READER_WRITER_LOCKS
Option as detailed in Section E.1 of IEEE-1003.1-2017.
The _POSIX_READER_WRITER_LOCKS Option is required for PSE51,
PSE52, PSE53, and PSE54 conformance, and is otherwise mandatory
for any POSIX conforming system as per Section A.2.1.3 of
IEEE-1003-1.2017.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
Move the implementation of the following functions into
the posix library rather than having themn as static
inline functions.
* pthread_rwlockattr_init()
* pthread_rwlockattr_destroy()
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
clock_getres() is required by the POSIX_TIMERS Option Group
as detailed in Section E.1 of IEEE-1003.1-2017.
The POSIX_TIMERS Option Group is required for PSE51, PSE52,
PSE53, and PSE54 conformance, and is otherwise mandatory for
any POSIX conforming system as per Section A.2.1.3 of
IEEE-1003-1.2017.
With this, we have complete support for the POSIX_TIMERS
Option Group.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
Support querying POSIX string configuration values (similar to
sysconf()).
confstr() is required by the POSIX_SINGLE_PROCESS Option
Group as detailed in Section E.1 of IEEE-1003.1-2017 and has
been part of the specification since POSIX-2.
The POSIX_SINGLE_PROCESS Option Group is required for PSE51,
PSE52, PSE53, and PSE54 conformance, and is otherwise mandatory
for any POSIX conforming system as per Section A.2.1.3 of
IEEE-1003-1.2017.
With this, we have complete support for the POSIX_SINGLE_PROCESS
Option Group.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
Implement `pthread_attr_setscope()` and `pthread_attr_getscope()`
are required as part of _POSIX_THREAD_PRIORITY_SCHEDULING Option Group.
signed-off-by: Gaetan Perrot <gaetanperrotpro@gmail.com>
This PR fixes the following compiler warning:
No SOURCES given to Zephyr library: lib__posix__shell
Signed-off-by: Matthias Alleman <matthias.alleman@basalte.be>
Previously, sysconf() was only available as a macro (i.e. the
"extra-small" option).
This has the advantage of being compile-time constant, and
optimized for both space and speed. One disadvantage is that
querying an `_SC_` value that was invalid or unsupported
would result in a compile error.
Provide a "small" implementation of sysconf() (via Kconfig
choice) as a normal addressable function.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
Seen this in compiler
lib/posix/options/syslog.c: In function 'setlogmask':
lib/posix/options/syslog.c:66:16: error: 'oldpri' may be used uninitialized
66 | return oldpri;
| ^~~~~~
lib/posix/options/syslog.c:59:13: note: 'oldpri' was declared here
59 | int oldpri;
| ^~~~~~
lib/posix/options/syslog.c: In function 'vsyslog':
lib/posix/options/syslog.c:83:33: error: 'mask' may be used uninitialized
83 | if ((BIT(level) & mask) == 0) {
| ~~~~~~~~~~~~~~~~~~~~^~~~
lib/posix/options/syslog.c:71:17: note: 'mask' was declared here
71 | uint8_t mask;
| ^~~~
Signed-off-by: Jukka Rissanen <jukka.rissanen@nordicsemi.no>
This PR adds declarations for the application conformance
feature test macro _POSIX_C_SOURCE.
It needs to be defined to value greater than or equal to 200112L
by the appplication.
However, Zephyr currently does not have a simple and consistent
means of specifying this value for POSIX samples, tests,
applications, and other libraries.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
This service is used to get, set, and unset system
environment variables.
Note: shell parameter expansion is not supported
at this time.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
Support getting and setting POSIX environment variables.
Additionally, the thread-safe BSD variant getenv_r() is
provided.
environ, getenv(), setenv(), and unsetenv() are required by
the POSIX_SINGLE_PROCESS Option Group as detailed in
Section E.1 of IEEE-1003.1-2017.
The POSIX_SINGLE_PROCESS Option Group is required for PSE51,
PSE52, PSE53, and PSE54 conformance, and is otherwise mandatory
for any POSIX conforming system as per Section A.2.1.3 of
IEEE-1003-1.2017.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
Brings SMF framework closer into alignment with accepted Hierarchical State
Machine operation by:
1. Allowing 'programming by difference' by having some child states handle
events and prevent propagation up to the parent run actions while others
propagate events up to a common handler in a parent state.
2. Optionally allow initial transitions within a parent state to determine
the most nested child state to transition to.
3. Adding a test case for `CONFIG_SMF_INITIAL_TRANSITION` and
`smf_set_handled()`
4. Updating documentation for the new API (and fixing some references)
There was discussion in https://github.com/zephyrproject-rtos/zephyr/issues/55344
about not making the initial transition a Kconfig option, but I'm not sure
of any way else of doing it without permanently adding a pointer to each
`smf_state` entry, which is a problem for resource-constrained devices.
This does not fix https://github.com/zephyrproject-rtos/zephyr/issues/66341
but documentation has been updated to warn users of the issue.
Signed-off-by: Glenn Andrews <glenn.andrews.42@gmail.com>
It turns out that currently LTO is enabled only for the kernel.
This commit updates it to enable it for the whole application
and adds additional LTO exclusions required for the standard
C libraries to build and link properly.
Signed-off-by: Radosław Koppel <radoslaw.koppel@nordicsemi.no>
This is just a simple wrapper around Zephyr's native log api.
Note: the standard LOG_ERR syslog priority does conflict with
Zephyr's LOG_ERR() macro. This will need to be worked-around on
a case-by-case basis.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
gmtime_r() has been in the minimal libc for years, however it was not
added to expcetions due to an overlook. In order to do this however, it
has to be moved first to the common libc area, so that it's available
to any libc that may not implement it.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Previously, pthread_mutexattr_gettype() and
pthread_mutexattr_settype() were non-conformant and also
less safe, as they would not check whether a pthread_mutexattr_t
had been initialized prior to manipulating them. Furthermore,
they would potentially dereference NULL pointers.
Additionally, move the pthread_mutexattr_init() and
pthread_mutexattr_destroy() functions to the library, and add
some level of checking to them so that they are more than simply
static inline / no-op calls.
Lastly, reduce the size of struct pthread_mutexattr to only
what is necessary (one byte should suffice).
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
pthread_exit() does not return and therefore it does not make
sense to return NULL after it in mq_notify_thread(), and that
would constitute dead code.
Rather than explicitly exiting the thread, simply return
gracefully from the thread function, and allow the pthread to
terminate in the usual way.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
In Zephyr, things are often optimized for size first. That's how
we fit into such tight parking spaces.
This change gives more control to the user about whether the
POSIX API does any logging at all, simultaneously shrinking binary
size while improving speed.
No bytes / cycles left behind!
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
update shell and test app with the modifed resource struct such as
acpi_irq_resource and acpi_mmio_resource
Signed-off-by: Najumon B.A <najumon.ba@intel.com>
This fixes CID 340851.
This should never fail, but it's also something that can be
easily verified.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
When CONFIG_LOG_USE_TAGGED_ARGUMENTS is enabled, and
CONFIG_CBPRINTF_COMPLETE is also enabled, we should not be
blindly skipping tags when processing the tagged package
for output. The issue is that if there is a "%%" in
the format string, the specifier is considered invalid but
the code blindly skips ahead in the argument list as if
it is a valid specifier (think "%s"), which resulting in
the next valid specifier using incorrect argument in
the list. So fix it by skipping ahead if and only if
the specifier is not invalid.
Fixes#68271
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
This change mitigates the following cmake warning
No SOURCES given to Zephyr library: lib__posix__options
without needing to have a separate interface library
(as we do not need private headers exposed).
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
Newlib doesn't have Zephyr support, so we need to define these functions
when the application doesn't ask for the right level of POSIX support.
Signed-off-by: Keith Packard <keithp@keithp.com>
Implement `sched_rr_get_interval()` POSIX APIs
as a part of PSE53 `_POSIX_PRIORITY_SCHEDULING` option group.
Functions is actually placeholders and just return `ENOSYS`
since Zephyr does not yet support processes or process scheduling.
signed-off-by: Gaetan Perrot <gaetanperrotpro@gmail.com>
In case the read index points to the end of the packet buffer and thus
wraps around, we must not invalidate the cache, otherwise we draw
outside of the lines and purge data outside of our responsibility.
Signed-off-by: Patrick Wildt <pwildt@google.com>
