The size of long double on x86-32 is 12 which is not
a power of 2, and this results in build error when it is
being used for alignment of buf32 in log_core.c.
So manually set it to 16.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
With SOF secondary cores are booted later at run-time instead
of the traditional simultaneous booting of all the cores.
Adjust arch_start_cpu() to make that possible.
Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
Currently P4WQ supports queues with sets of user-provided
worked threads of arbitrary numbers. These threads are started
immediately upon initialisation.
This patch adds support for 3 more thread implementation options:
1. queue per thread. It adds a K_P4WQ_ARRAY_DEFINE() macro which
initialises an array of queues and threads of the same number.
These threads are then uniquely assigned to respective queues.
2. delayed start. With this option threads aren't started
immediately upon queue initialisation. Instead a new function
k_p4wq_enable_static_thread() has to be called to enable those
threads individually.
3. queue per CPU. With this option the user can assign CPU masks
to threads when calling k_p4wq_enable_static_thread().
Otherwise the cpu_mask parameter to that function is ignored.
Currently enabling this option implies option 2 above. Also so
far to enable queues per CPU the user has to use
K_P4WQ_ARRAY_DEFINE(), which means this option also implies 1
above, but both these restrictions can be relaxed in the
future if required.
Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
Work items in P4WQ currently belong to the user before submission
and after exit from the handler, therefore, unless the handler
re-submits the item, accessing it in p4wq_loop() in such cases
is racy. To fix this we re-define work item ownership. Now the
item belongs to the P4WQ core until the user calls
k_p4wq_wait(). If the work item has its .sync flag set, the
function will sleep until the handler completes processing the
work item or until the timeout expires. If .sync isn't set and
the handler hasn't processed the item yet, the function returns
-EBUSY.
Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
For non-specified archs, including those out-of-tree, the possibility to
use a specific implementation has been reintroduced.
CONFIG_ATOMIC_OPERATIONS_ARCH must be selected to utilize this.
Signed-off-by: Tommie Skriver <tosk@demant.com>
Reboot functionality has nothing to do with PM, so move it out to the
subsys/os folder.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
Coverity is complaining about sizeof(v + 0) and it is used
here intentionally to promote variable. Added comment that should
suppress this error in the future.
Note that this macro will be used in all log messages so without
solving it before logging v2 is merged there will be a flood of
errors.
Signed-off-by: Krzysztof Chruscinski <krzysztof.chruscinski@nordicsemi.no>
z_tmcvt() was missing final else statement in the
if else if construct. This commit removes the
else if in this small structure to comply with
guideline 15.7.
Signed-off-by: Jennifer Williams <jennifer.m.williams@intel.com>
Z_CBPRINTF_ARG_SIZE macro is called for each argument in
logging macros. If argument is a string literal an intention
of this macro is to return size of a pointer. Suppressing
warning which appears in that case.
Signed-off-by: Krzysztof Chruscinski <krzysztof.chruscinski@nordicsemi.no>
Added module for storing variable length packets in a ring buffer.
Implementation assumes multiple producing contexts and single consumer.
API provides zero copy functionality with alloc, commit, claim, free
scheme.
Additionally, there are functions optimized for storing single word
packets and packets consisting of a word and a pointer. Buffer can work
in two modes: saturation or overwriting the oldest packets when buffer
has no space to allocate for a new buffer.
Signed-off-by: Krzysztof Chruscinski <krzysztof.chruscinski@nordicsemi.no>
There's a typedef for non-pointer values compatible with atomic
non-pointer objects. Add a similar typedef for pointer values, and
the corresponding macro for initializing atomic pointer types.
This also will simplify replacing the Zephyr atomic API with one
based on C11 atomics, should that be desirable. C11 atomic pointer
values are not void*.
Signed-off-by: Peter Bigot <peter.bigot@nordicsemi.no>
Move out of misc/ and put in own folder and add the grouping to doxygen
to be able to reference the doxygen docs into RST.
Move each item into their own file to reduce clutter and to make it
less crowded in one single page.
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Allow NULL data buffers to be provided to `ring_buf_get` and
`ring_buf_item_get`, in which case data will be discarded instead of
copied out to the user.
Fixes#33488.
Signed-off-by: Jordan Yates <jordan.yates@data61.csiro.au>
__ASSERT() macro is used in sys/sflist.h while sys/__assert.h was not
included. Fix that now.
Signed-off-by: Marcin Niestroj <m.niestroj@grinn-global.com>
Predefine ATOMIC_DEFINE in the Doxyfile so that documentation output is
generated correctly. In order to simplify the predefinition
ATOMIC_BITMAP_SIZE has been introduced.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
This adds the bits to record execution time of eviction selection,
and backing store page-in/page-out in histograms.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
This adds more bits to gather statistics on demand paging,
e.g. clean vs dirty pages evicted, # page faults with
IRQ locked/unlocked, etc.
Also extends this to gather per-thread demand paging
statistics.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Usually, GCC builtin or arch-specific atomic functions are being
used. The corresponding kconfigs are "selected" by architecture
or SoC kconfigs. CONFIG_ATOMIC_OPERATIONS_C is usually used to
override the GCC built-in or arch-specific atomic functions when
these two are not supported. So change the priority of #include
so that C version is included first if selected, and skips
the inline versions of the other two variants. Or else there
will be two compiled versions of atomic functions: inline version
and the compiled C version. Note that the arch-specific version
and builtin are swapped, so the builtin one is now the default.
Fixes#33857
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
The builtin atomic header file should not include the syscall
header that is generated from C version of atomic functions.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Unified define used for handling sparc case in static and
runtime packaging. Reworked macro for storing argument in
static packaging.
Signed-off-by: Krzysztof Chruscinski <krzysztof.chruscinski@nordicsemi.no>
Added parameter to CBPRINTF_STATIC_PACKAGE which indicates buffer
alignment offset compared to CBPRINTF_PACKAGE_ALIGNMENT. When offset
is set to 0, macro assumes that input buffer is aligned to
CBPRINTF_PACKAGE_ALIGNMENT. When offset is positive, macro assumes
that buffer address is shifted by given number of bytes to
CBPRINTF_PACKAGE_ALIGNMENT alignment.
Extended cbprintf_package to use len argument as alignment offset
indicator when calculating length only (package pointer is null).
Features are not available for xtensa platform which seems to
require 16 byte alignment from the package. It is only an assumption
due to lack of the documentation and may be fixed in the future.
Feature allows to avoid unnecessary padding when package is part of
a message and preceeded by a header of a known size. For example,
message header on 32 bit architecture has 12 bytes, long doubles are
not used so cbprintf requires 8 byte alignment. Without alignment
offset indicator, package containing just a string with one argument
would need 4 byte padding after the header and 4 byte padding after
the package. Message would be 32 bytes long. With alignment offset
indication both paddings are not needed and message is only 24 bytes
long.
Signed-off-by: Krzysztof Chruscinski <krzysztof.chruscinski@nordicsemi.no>
Sparc architecture is strange. Va_list arguments are packed (1 byte
alignment) while unaligned access fails. Added dedicated handling of
Z_CBPRINTF_STORE_ARG which is copying the data word by word.
Signed-off-by: Krzysztof Chruscinski <krzysztof.chruscinski@nordicsemi.no>
Added missing errno.h include. Fixed Z_CBPRINTF_ARG_SIZE macro for
void * and cleaned up macro description.
Signed-off-by: Krzysztof Chruscinski <krzysztof.chruscinski@nordicsemi.no>
The only user of arch_mem_domain_destroy was the deprecated
k_mem_domain_destroy function which has now been removed. So remove
arch_mem_domain_destroy as well.
Signed-off-by: Kumar Gala <kumar.gala@linaro.org>
We need to do a few things differently if we are to support
a virtual memory map, i.e. CONFIG_MMU where CONFIG_KERNEL_VM_BASE
is not the same as CONFIG_SRAM_BASE_ADDRESS.
- All sections must be specified with a VMA and LMA, where
VMA is the virtual address and LMA is the physical memory
location.
- All sections must be specified with ALIGN_WITH_INPUT to
keep VMAs and LMAs synchronized
To do this, the existing linker macros need some adjustment:
- GROUP_LINK_IN undefined when CONFIG_KERNEL_VM_BASE is not
the same as CONFIG_SRAM_BASE_ADDRESS.
- New macro GROUP_ROM_LINK_IN for text/rodata sections
- New macro GROUP_NOLOAD_LINK_IN for bss/noinit sections
- Implicit ALIGN_WITH_INPUT for all sections
GROUP_FOLLOWS_AT is unused anywhere in the kernel for years
now and has been removed.
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
This reverts commit 7d32e9f9a5.
These functions are needed for linking kernel in virtual address
space when it differs from the physical address space.
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Turn sys_heap_dump() into sys_heap_print_info() to better reflect
what it actually does, and improve the information being printed.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Due to the fact that define was created after including
cbprintf_internal.h, it was not used there. Change the order and fix
the issue that was revealed.
Signed-off-by: Krzysztof Chruscinski <krzysztof.chruscinski@nordicsemi.no>
Added validation of alignment to cbprintf_package. Error is returned if
input buffer is not aligned to the largest argument.
Signed-off-by: Krzysztof Chruscinski <krzysztof.chruscinski@nordicsemi.no>
Remove redundant if() statements that are already included with
k_mutex_unlock()
Relates to issue #32994
Signed-off-by: Guðni Már Gilbert <gudni.m.g@gmail.com>
The xtensa atomics layer was written with hand-coded assembly that had
to be called as functions. That's needlessly slow, given that the low
level primitives are a two-instruction sequence. Ideally the compiler
should see this as an inline to permit it to better optimize around
the needed barriers.
There was also a bug with the atomic_cas function, which had a loop
internally instead of returning the old value synchronously on a
failed swap. That's benign right now because our existing spin lock
does nothing but retry it in a tight loop anyway, but it's incorrect
per spec and would have caused a contention hang with more elaborate
algorithms (for example a spinlock with backoff semantics).
Remove the old implementation and replace with a much smaller inline C
one based on just two assembly primitives.
This patch also contains a little bit of refactoring to address the
scheme has been split out into a separate header for each, and the
ATOMIC_OPERATIONS_CUSTOM kconfig has been renamed to
ATOMIC_OPERATIONS_ARCH to better capture what it means.
Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
In applications like logging the call site where arguments to
formatting are available may not be suitable for performing the
formatting, e.g. when the output operation can sleep. Add API that
supports capturing data that may be transient into a buffer that can
be saved, and API that then produces the output later using the
packaged arguments.
[ Documentation and commit log from Peter Bigot. ]
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Signed-off-by: Peter Bigot <peter.bigot@nordicsemi.no>
The compiler was inserting additional redundant loads in many
`sys_dlist_*` APIs, in case writes aliased with previous reads.
However, these additional reads are unnecessary, as the only cases
where the aliasing would matter would be a violation of the `dlist`
API contract (e.g. if node->next == node but node->prev != node).
This is decidedly a micro-optimization.
Signed-off-by: James Harris <james.harris@intel.com>
