Previously, Zephyr's POSIX API did not differentiate between
deferred and asynchronous pthread cancellation. In fact all
pthread cancellation was asynchronous. According to the spec,
all pthreads should be created with deferred cancellation by
default.
Note: PTHREAD_CANCEL_ASYNCHRONOUS means cancel asynchronously
with respect to cancellation points (but synchronously with
respect to the thread that callse pthread_cancel(), which is
perhaps unintuitive).
The POSIX timer relied on this non-standard convention.
Oddly, this change prevents what would have otherwise been a
regression that would have been caused by fixing pthread
behaviour (in a separate commit).
We are effectively uncovering bugs which were probably always
present in the pthread.c and timer.c implementations going
back quite a few years.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
There is no requirement that says e.g. CLOCK_REALTIME cannot be
used for timer_create(). In fact, the spec explicitly requires
it. It might not be ideal, but users should still be able to
use it.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
Use the existing `is_posix_policy_prio_valid()` function to
verify the POSIX's priority in the conversion functions.
Changed the `priority` arg of `is_posix_policy_prio_valid` to
`int` since that is the output of `sched_get_priority_min` &
`sched_get_priority_max`.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
Made the conversion functions non-static and added ztests for
them to make sure that they work across the full range of
Zephyr priorities.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
Changed the variable type of the priority in the args and the
return type of the conversion functions to `int`, as both
Zephyr's priority & POSIX's `sched_priority` has type `int`.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
If `z_prio` is negative and we want to make sure that it is
within `[-CONFIG_NUM_COOP_PRIORITIES, -1]`, we should invert
its sign and make sure that it is `<=`
`CONFIG_NUM_COOP_PRIORITIES` instead of `<`.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
Added a top level `posix` shell command for other POSIX
commands. Currently only `uname` is supported.
New POSIX commands can be added by including the
`posix_shell.h` header and use the `POSIX_CMD_ADD` helper
macro.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
Relocate the `uname` shell implementation from uname sample, so
that it can be reused by other application and the uname sample
only uses POSIX APIs.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
The POSIX standard doesn't specify if the argument to store
previous state/type in
`pthread_setcancelstate`/`pthread_setcancelstate` can be
`NULL`, but threading implementations in Linux & Apache NuttX
permit the arguments to be `NULL`.
This commit changes Zephyr's implementation to mimic that of
Linux & NuttX, so that user do not get caught off-guard by
NULL pointer dereferencing when porting code over from those
OSes.
Updated test accordingly.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
Make sure that the POSIX clock base is reset after every
testsuite so that the initial time is the same for every test.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
Instead of relying on a fault occurring when NULL
parameters are passed in, explicitly check and return
EINVAL.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
Initial implementation of `sched_getparam()` and `sched_getscheduler()`
POSIX APIs as a part of PSE53 `_POSIX_PRIORITY_SCHEDULING` option group.
Both functions are actually placeholders and just return `ENOSYS`
since Zephyr does not yet support processes or process scheduling.
Signed-off-by: Dmitrii Golovanov <dmitrii.golovanov@intel.com>
- Regroup refcount decrement and semaphore destruction by making the
linked state into a counted reference for it. This allows for
simplifying the code and cleanly adding a few assertions in a common
location.
- Remove redundant initialization to NULL on memory about to be freed
and local pointer in nsem_cleanup().
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Improve a depends on the host libC.
It is technically correct, but NATIVE_LIBC
is shorter and clearer than "ARCH_POSIX && EXTERNAL_LIBC"
Signed-off-by: Alberto Escolar Piedras <alberto.escolar.piedras@nordicsemi.no>
Let's try to use the host unistd.h when building with
the host library only, instead of assuming
that the native boards are always built with it.
This fixes a build error when building for native boards
using minimal libc.
Signed-off-by: Alberto Escolar Piedras <alberto.escolar.piedras@nordicsemi.no>
We try to implement Zephyr's POSIX API as regular library
functions, so remove the __syscall annotation from
clock_gettime() and implement the syscall portion of it under
the hood.
This also adds a bit of a micro-optimization in that
we can do a lot of processing outside of the system call.
In fact, processing CLOCK_MONOTONIC likely does not
require any syscall other than k_uptime_ticks().
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
The clock_settime() function should not accept an input timespec
with nanosecond values < 0 or >= NSEC_PER_SEC.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
Added CRC4 and CRC4_TI to the supported crc_types[] array.
On some SoCs, like ESP32-S3, missing values can cause hardfaults
due to attempts to access the zero address.
Signed-off-by: Pisit Sawangvonganan <pisit@ndrsolution.com>
This makes sure the heap posix_timer_slab provides objects aligned
compatible with the type timer_obj. It was previously set to align at 4
bytes. One example where this failed was on the SPARC which requires
access to int64_t to be 8-byte aligned.
In particular, struct timer_obj contains fields of type k_timer_t and
struct _timeout.
With this commit we now get the information on required alignment for
struct timer_obj from the compiler by using __alignof__().
Signed-off-by: Martin Åberg <martin.aberg@gaisler.com>
There was some discussion about whether it was suitable to have
an architecture-specific workaround in mqueue.c after that
workaround was copied to a different source file in a PR.
The original issue was that newlib and picolibc declare mode_t
to be unsigned short instead of unsigned long when __svr4__
is not defined along with __sparc__. This is specifically
impactful, because va_arg() deals (mainly) with 32-bit and
64-bit values that are all naturally aligned to 4 bytes.
#if defined(__sparc__) && !defined(__sparc_v9__)
#ifdef __svr4__
typedef unsigned long __mode_t;
#else
typedef unsigned short __mode_t;
#endif
A uint16_t is naturally aligned to 2 bytes, so not only would
a 16-bit mode_t be corrupted, it would also generate a warning
with recent gcc versions which is promoted to error (rightfully
so) when run through CI.
mqueue.c:61:35: error: 'mode_t' {aka 'short unsigned int'} is
promoted to 'int' when passed through '...' [-Werror]
61 | mode = va_arg(va, mode_t);
Instead of using an architecture-specific workaround, simply
add a build assert that the size of mode_t is less than or
equal to the size of an int, and use an int to retrieve it
via va_arg().
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
# Please enter the commit message for your changes. Lines starting
# with '#' will be kept; you may remove them yourself if you want to.
# An empty message aborts the commit.
#
# Date: Fri Dec 29 10:06:44 2023 -0500
#
# On branch posix-mqueue-always-use-int-for-mode-t-va-arg
# Changes to be committed:
# modified: lib/posix/mqueue.c
# modified: tests/posix/common/testcase.yaml
#
Use the __fallthrough directive on the switch fallthrough case and drop
the comment instead. This informs the compiler that the fallthrough is
intentional and silence any possible warning about it.
Drop the not reached case as I think that that part can actually be
reached since there are breaks in some of the cases.
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
In theory one may want to build the POSIX compability shim
with a different C library than one provided with Zephyr,
so let's not prevent it.
Signed-off-by: Alberto Escolar Piedras <alberto.escolar.piedras@nordicsemi.no>
This reverts commit 6342aa3cc0.
This commit should never have been merged.
Apart from the fact that this change was rejected in previous
review,
this change is wrong, for 2 reasons:
1. The POSIX_API (POSIX compatibility shim) cannot be
built if the host libC is used.
2. The Zephyr libC CMake files were guarded so they
would not be dragged when CONFIG_EXTERNAL_LIBC was selected.
Signed-off-by: Alberto Escolar Piedras <alberto.escolar.piedras@nordicsemi.no>
* remove unneeded line of code that duplicated the first part
of the SYS_SLIST_FOR_EACH_NODE() expansion.
* return NULL if pthread_self() is not a valid pthread
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
When package contained RO string positions and flag indicates that
they shall not be kept, length was miscalculated which could lead
to failures (e.g. memory corruption).
Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
Previously it was not possible to link to Zephyr's
POSIX API under lib/posix when building with
`CONFIG_EXTERNAL_LIBC=y`.
This small change allows that to work as expected.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
A previous size optimization capped the pthread_attr_t stacksize
property at 65536. Some Zephyr users felt that was not large
enough for specific use cases.
Modify struct pthread_attr to support large stack sizes by
default with the flexibility to allow users to vary the number
of bits used for both stacksizes and guardsizes.
The default guardsize remains zero sinze Zephyr's stack
allocators already pad stacks with a guard area based on other
config parameters, and since Zephyr is already designed to
support both SW and HW stack protection at the kernel layer.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
Move various utilities out of lib into own folder for better assignement
and management in the maintainer file. lib/os has become another dumping
ground for everything and it the Kconfig and contents in that folder
became difficult to manage, configure and test.
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Implement pthread_getguardsize() and pthread_setguardsize().
pthread_getguardsize() and pthread_setguardsize() are required
by the POSIX_THREADS_EXT Option Group as detailed in Section
E.1 of IEEE-1003.1-2017. However, they were formerly part of
XSI_THREADS_EXT.
The XSI_THREADS_EXT Option Group was required for PSE51, PSE52,
PSE53, and PSE54 conformance.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
This change reduces the space occupied by struct pthread_attr
which is the internal type used for pthread_attr_t.
We cap the stack size at 16 bits (so up to 65536 bytes) and
since a stack size of 0 is invalid, we can encode the stack
size by simply subtracting 1 or adding 1 when setting or
getting.
The schedpolicy is capped at 2 bits and initialized,
cancellable, and detached are given 1 bit.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
Use the new HEAP_MEM_POOL_ADD_SIZE_ prefix to construct a minimum
requirement for posix message queue usage. This way we can remove the
"special case" default values from the HEAP_MEM_POOL_SIZE Kconfig
definition.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
All x86 boards have so far set a custom heap memory pool size because of
their dependency on ACPI. It makes more sense to introduce a new
ACPI-specific option, utilizing the recently added
HEAP_MEM_POOL_ADD_SIZE_ Kconfig option prefix, and adjust the default
value as necessary for each board.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Add function walking though all DMAR subtables, at the moment only
first subtable is taking into account, which causes bugs for some
boards.
Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>
Since pthread_once() is both the initializer and executor of
pthread_once_t, it can have maximally two states. Since the
implementation in Zephyr previously aimed to maximize libc
compatibility, we opted to use the definition of pthread_once_t
from newlib, which is a structure with two ints.
It does not make sense to use 64 bits to manage 2 possible
states. The control for that should effectively be a bool.
We maintain compatibility with newlib by asserting (at build
time), that newlib's pthread_once_t is larger than Zephyr's
new struct pthread_once (which just contains a bool).
This allows us to delete the non-standard pthread_key.h
header file (finally).
Reuse the pthread_pool_lock in order to synchronize the related
init function (so that it is only called maximally once from any
thread). The spinlock is only used to test the state and the
init function is not called with the spinlock held.
The alternative was to use an atomic inside of
struct pthread_once. But again, that could be up to 64-bits with
Zephyr's atomics implementation.
Ideally we would use C11 generics or something to support atomics
on 8, 16, 32, and 64-bit primitives.
Lastly, also update declarations for C11 threads as they mostly
mirror our pthread implementation.
This needed to be done as a single commit in order to ensure
continuity of build.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
