Multiple reader threads unlocking the read lock simultaneously might
cause the program hang because it's possible that no thread is
identified as the last one to active the writer thread.
To fix the issue, swap the k_sem_give sequence.
Signed-off-by: Jaxson Han <jaxson.han@arm.com>
The `pthread_once_lock` `k_mutex` is statically initialized and
only visible within file scope. Coverity identified it as unsafe
because the return values of `pthread_mutex_lock()` and
`pthread_mutex_unlock()` were unchecked. However, if those
functions were to fail here, it would be indicative that
something far worse has happened.
In any case, we add assertions that these functions
succeed rather than silently ignoring with `(void)`, which
ensures that we have coverage when assertions are enabled,
in test, while removing unneeded code with assertions disable,
in production.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
The `struct k_spinlock` size is zero bytes under certain
circumstances. This is a bit of a problem, because it breaks a
number of assumptions about things in C.
That should be fixed when #59922 is addressed.
This change is just a hotfix to correct the specific condition
where we will need workarounds in other source files.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
This change is setting up for switching over to proper POSIX
option requirements, feature test macros, and a dependency
structure that is reflective of the standard.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
Most of the posix source files can be easily identified by a
short name. I.e. most of the `pthread_` prefixed files do not
need the `pthread_` prefix.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
A significant enough portion of these files has been
changed to justify adding Meta copyright as well as
that of the original author.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
Fixes#58911. Previously the stat command returned
information on the filesystem, but not the file itself.
Because block size is still set this function is
backwards compatible with the previous faulty
behavior.
Signed-off-by: Vincent van Beveren <v.van.beveren@nikhef.nl>
The POSIX API compatibility shim can be used for some
of the POSIX ARCH targets.
Narrow the Kconfig filtering accordingly.
Note that the recommended configuration when building
with the native simulator is still to use an embedded
C library. Using the host C library will in some cases
cause undesired behaviour.
Signed-off-by: Alberto Escolar Piedras <alberto.escolar.piedras@nordicsemi.no>
Fix a regression introduced by commit e6eb0a705b ("posix: eventfd: revise
locking, signaling, and allocation"), which was a complete rewrite stating
that:
The `wait_q` and `k_poll_signal` entries were removed from
`struct eventfd` as they were unnecessary.
In fact, `k_poll_signal` (both `read_sig` and `write_sig`) were used to
wake-up blocking `poll()` invocation in another thread. This is no longer
the case now, i.e. `poll(..., POLLIN)` does not return after calling
`eventfd_write()` on the observed (polled) FD.
Fix this regression by bringing back `read_sig` and `write_sig` to very
similar state as it was before.
Fixes: e6eb0a705b ("posix: eventfd: revise locking, signaling, and
allocation")
Signed-off-by: Marcin Niestroj <m.niestroj@emb.dev>
This adds a few line use zephyr_syscall_header() to include
headers containing syscall function prototypes.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Previously, the `posix_internal.h` header needed to be exposed
to the application because we had non-trivial details for
most posix types (pthread, mutex, cond, ...). Since most of
those have been simplified to a typedef'ed integer, we
no longer need to expose that header to the applicaiton.
Additionally, it means that we can adopt normalized
header order in posix.
Additionally, keep more implementation details hidden
and prefer the static keyword on internal symbols where
possible.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
To enable testing, introduce `CONFIG_PTHREAD_CREATE_BARRIER`.
Some observations were made that running several Qemu SMP targets
concurrently could lead to synchronization problems. On such
targets, it was found that the synchronization issues were
mitigated by introducing a `pthread_barrier_t` shared between
`pthread_create()` and the spawned thread.
It is suggested to enable the option when running many
SMP tests concurrently in several parallel Qemu processes,
e.g. with `twister`.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
Previously, pthreads suffered from some race conditions.
This was almost inevitable given that it was maintained in
parallel to Zephyr's threading and synchronization API.
The unfortunate side-effect of with that is that it did not
receive the reliability and other improvements that
`k_thread`s did.
Here, we perform a significant update of pthread code so
that it depends directly on public Zephyr API. With that,
we reuse as many concepts as possible and pthreads benefits for
free from any improvement made to Zephyr's threading and
synchronization APIs.
Included with this change, we
* implement state with `ready_q`, `run_q`, and `done_q`
* use `pthread_barrier_wait()` to sync `pthread_create()`
* synchronize internal state with a spinlock
These pthreads are considerably more reliable than
before.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
The internal representation of `pthread_cond_t`,
`struct posix_cond`, is an identical clone of `struct k_condvar`
but without the benefit of being able to use all of the
existing `k_condvar_*()` suite of functions.
The first step in the right direction was switching
the external representation of `pthread_cond_t` to a simple
`int`. Let's take the next step in the right direction, which
is getting rid of `struct posix_cond`.
For now, let's keep this change as a simple type
substitution. Eventually, we should be able to fully switch
to Zephyr API internally.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
The internal representation of `pthread_mutex_t`,
`struct posix_mutex`, is basically a clone of `struct k_mutex`
but without the benefit of being able to use all of the
existing `k_mutex_*()` suite of functions.
The first step in the right direction was switching
the external representation of `pthread_mutex_t` to a simple
`int`. Let's take the next step in the right direction, which
is getting rid of `struct posix_mutex`.
The only significant difference between `struct k_mutex` and
`struct posix_mutex` is that the latter needs a `type` field.
Since there were a fixed number of `struct posix_mutex`, we
can just externalize the `type` field and reuse
`struct k_mutex` as-is.
For now, let's keep this change as a simple type
substitution. Eventually, we should be able to fully switch
to Zephyr API internally.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
Previously pthread_barrier_t was implemented in terms of wait
queues and internal scheduler functions.
This introduced some obstacles and inconsistency. In order
to be more consistent, rely only on Zephyr's public API and
reuse as many concepts as possible.
Deprecate `PTHREAD_BARRIER_DEFINE()` since it's non-standard.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
Deprecate `EFD_IN_USE` and `EFD_FLAGS_SET` as they are not
specified as part of any public `eventfd()` API.
While those are being deprecated, use `_INTERNAL` variants.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
TL;DR - a complete rewrite.
Previously, the prototypical `eventfd()` usage (one thread
performing a blocking `read()`, followed by another thread
performing a `write()`) would deadlock Zephyr. This shortcoming
has existed in Zephyr's `eventfd()` implementation from the
start and the suggested workaround was to use `poll()`.
However, that is not sufficient for integrating 3rd-party
libraries that may rely on proper `eventfd()` blocking
operations such as `eventfd_read()` and `eventfd_write()`.
The culprit was the per-fdtable-entry `struct k_mutex`.
Here we perform a minor revision of the locking strategy
and employ `k_condvar_broadcast()` and `k_condvar_wait()`
to signal and wait on the holder of a given `struct k_mutex`.
It is important to note, however, that the primary means of
synchronizing the eventfd state is actually the eventfd
spinlock. The fdtable mutex and condition variable are mainly
used for the purposes of blocking io (r,w,close) and are not
used in the code path of non-blocking reads.
The `wait_q` and `k_poll_signal` entries were removed from
`struct eventfd` as they were unnecessary.
Additionally, switch to using a bitarray because it is
possibly faster than linear search for allocating and
deallocating eventfd resources.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
There is no major benefit to haveing `eventfd_read()` and
`eventfd_write()` as `static inline` functions and it is
arguably much easier to solve the deadlock bug when they
are not `inline`.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
Until now iterable sections APIs have been part of the toolchain
(common) headers. They are not strictly related to a toolchain, they
just rely on linker providing support for sections. Most files relied on
indirect includes to access the API, now, it is included as needed.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
Have eventfd Kconfig select POLL is the code utilizes the polling
API. We get a link error for tests/lib/fdtable/libraries.os.fdtable
when building on arm-clang without this.
Signed-off-by: Kumar Gala <kumar.gala@intel.com>
The `SCHED_OTHER` scheduling priority is mandatory as part of
POSIX. It must be numerically distinct from `SCHED_FIFO`,
`SCHED_RR`, and `SCHED_SPORADIC`, but is implementation-
defined and may behave identically to `SCHED_FIFO` or
`SCHED_RR`.
Signed-off-by: Chris Friedt <cfriedt@meta.com>
MISRA Rule 5.7 requires uniqueness of tag identifiers. Shell is
frequently problematic because many code uses `const struct shell
*shell`. This causes CI noise every time one of these shell files is
edited, so let's update all of them with `const struct shell *sh`
instead.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
The init infrastructure, found in `init.h`, is currently used by:
- `SYS_INIT`: to call functions before `main`
- `DEVICE_*`: to initialize devices
They are all sorted according to an initialization level + a priority.
`SYS_INIT` calls are really orthogonal to devices, however, the required
function signature requires a `const struct device *dev` as a first
argument. The only reason for that is because the same init machinery is
used by devices, so we have something like:
```c
struct init_entry {
int (*init)(const struct device *dev);
/* only set by DEVICE_*, otherwise NULL */
const struct device *dev;
}
```
As a result, we end up with such weird/ugly pattern:
```c
static int my_init(const struct device *dev)
{
/* always NULL! add ARG_UNUSED to avoid compiler warning */
ARG_UNUSED(dev);
...
}
```
This is really a result of poor internals isolation. This patch proposes
a to make init entries more flexible so that they can accept sytem
initialization calls like this:
```c
static int my_init(void)
{
...
}
```
This is achieved using a union:
```c
union init_function {
/* for SYS_INIT, used when init_entry.dev == NULL */
int (*sys)(void);
/* for DEVICE*, used when init_entry.dev != NULL */
int (*dev)(const struct device *dev);
};
struct init_entry {
/* stores init function (either for SYS_INIT or DEVICE*)
union init_function init_fn;
/* stores device pointer for DEVICE*, NULL for SYS_INIT. Allows
* to know which union entry to call.
*/
const struct device *dev;
}
```
This solution **does not increase ROM usage**, and allows to offer clean
public APIs for both SYS_INIT and DEVICE*. Note that however, init
machinery keeps a coupling with devices.
**NOTE**: This is a breaking change! All `SYS_INIT` functions will need
to be converted to the new signature. See the script offered in the
following commit.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
init: convert SYS_INIT functions to the new signature
Conversion scripted using scripts/utils/migrate_sys_init.py.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
manifest: update projects for SYS_INIT changes
Update modules with updated SYS_INIT calls:
- hal_ti
- lvgl
- sof
- TraceRecorderSource
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
tests: devicetree: devices: adjust test
Adjust test according to the recently introduced SYS_INIT
infrastructure.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
tests: kernel: threads: adjust SYS_INIT call
Adjust to the new signature: int (*init_fn)(void);
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
* add opening and closing braces around several blocks
* convert dos line endings to unix
* move assignments out of conditionals
* ensure adequate spacing after control blocks
Signed-off-by: Chris Friedt <cfriedt@meta.com>
Although the eventfd API is not (yet) a part of POSIX,
it's pretty well ubiquitous on POSIX systems now.
Enable it by default when `CONFIG_POSIX_API=y`.
Signed-off-by: Chris Friedt <cfriedt@meta.com>
The `getopt()` function is part of POSIX and should be
available when applications choose to enable general POSIX
API support.
Signed-off-by: Chris Friedt <cfriedt@meta.com>
With the `<zephyr/posix/...> prefix, it became
exponentially more difficult to integrate 3rd-party
libraries that depend on the POSIX API.
Standard POSIX headers should be available in standard
include paths - and that should most certainly the case
when `CONFIG_POSIX_API=y`.
With this change:
* When `CONFIG_POSIX_API=y`
- applications have explicitly chosen to use
POSIX APIs.
- all standard POSIX includes are in the default
include path.
* When `CONFIG_POSIX_API=n`
- applications *may* include POSIX headers
explicitly with the namespaced prefix
- e.g. `#include <zephyr/posix/unistd.h>`
- individual Kconfig options can be used to
enable POSIX features selectively, such as
`getopt` or `eventfd`.
Signed-off-by: Chris Friedt <cfriedt@meta.com>
The remaining types that needed to be harmonized between
Newlib and Zephyr's POSIX definitions are:
* `struct sched_param`
- don't re-define if using minimal libc
* `pthread_attr_t`
- convert to `struct pthread_attr`
- define type if using minimal libc
- assert acceptible object size
* `pthread_mutexattr_t`
- convert to `struct pthread_mutexattr`
- define type if using minimal libc
- assert acceptible object size
* `pthred_condattr_t`
- convert to `struct pthread_condattr`
- define type if using minimal libc
- assert acceptible object size
* `pthread_once_t`
- adopt newlib definition
- define type if using minimal libc
Signed-off-by: Chris Friedt <cfriedt@meta.com>
Define `PTHREAD_CREATE_DETACHED` and
`PTHREAD_CREATE_JOINABLE` to be compatible with the Newlib
definitions.
This is a temporary workaround for #51211 until Newlib
headers are pulled in.
Signed-off-by: Chris Friedt <cfriedt@meta.com>
The `fcntl.h` header has never been a part of ISO C so move it to
`include/zephyr/posix`.
To ensure a smooth migration, a header was left in
`lib/libc/minimal/include` that prints a deprecation warning.
Users should either include `<zephyr/posix/fcntl.h>` or switch to
`CONFIG_POSIX_API=y`.
Signed-off-by: Chris Friedt <cfriedt@meta.com>
The `sys/stat.h` header has never been a part of ISO C so move it
to `zephyr/include/posix/sys/`.
To ensure a smooth migration, leave a stub header in
`lib/libc/minimal/include/sys/` that prints a deprecation warning
suggesting developers either include `<zephyr/posix/sys/stat.h>`
or use `CONFIG_POSIX_API=y`.
Signed-off-by: Chris Friedt <cfriedt@meta.com>
The previous method used to calculate seconds in `clock_gettime()`
seemed to have an inaccuracy that grew with time causing the
seconds to be off by an order of magnitude when ticks would roll
over.
This change fixes the method used to calculate seconds.
Signed-off-by: Chris Friedt <cfriedt@meta.com>
should be able to configure the time spent waiting
for available resources when calling timer_create()
to not cause a hiccup in applications that require
faster response times than the original hard-coded 100 ms.
Signed-off-by: Nicholas Lowell <nlowell@lexmark.com>
Declarations for `getopt()` should be in `<unistd.h>`
according to the spec. The extended versions `getopt_long()`
and `getopt_long_only()` are declared in `<getopt.h>`.
Signed-off-by: Chris Friedt <cfriedt@meta.com>
This seems to have caused build failures in spite of CI being
green in PR 52653.
This reverts commit fc92eb1b37.
Signed-off-by: Chris Friedt <cfriedt@meta.com>
