This commit does two things to k_wakeup():
1. It locks the scheduler before marking the thread as not suspended.
As the the clearing of the _THREAD_SUSPENDED bit is not atomic, this
helps ensure that neither another thread nor ISR interrupts this
action (resulting in a corrupted thread_state).
2. The call to flag_ipi() has been removed as it is already being
made within ready_thread().
Signed-off-by: Peter Mitsis <peter.mitsis@intel.com>
cpu_start_fn is global, it's used by the initiator CPU to start or
resume secondary CPUs. However it's possible, that the initiator CPU
goes ahead and starts a second secondary CPU before the first one has
finished using the object. Fix this by creating a local copy of the
global object.
Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
When allowing dynamic thread stack allocation the stack may come from
the heap in coherent memory, trying to use cached memory is over
complicated because of heap meta data and cache line sizes.
Also when userspace is enabled, stacks have to be page aligned and the
address of the stack is used to track kernel objects.
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
It is observed that starting up CPU may result in other CPUs
crashing due to de-referencing NULL pointers. Note that this
happened on the up_squared board, but there was no way to
attach debugger to verify. It started working again after
moving z_dummy_thread_init() before smp_timer_init(), which
was the old behavior before commit
eefaeee061 where the issue
first appeared. So mimic the old behavior to workaround
the issue.
Fixes#68115
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Traditionally, k_thread_create() has required that the application
size the stack correctly. Zephyr doesn't detect or return errors and
treats stack overflow as an application bug (though obviously some
architectures have runtime features to trap on overflows).
At this one spot though, it's possible for the kernel to adjust the
stack for K_THREAD_STACK_RESERVED in such a way that the arch layer's
own stack initialization overflows. That failure can be seen by
static analysis, so we can't just sweep it under the rug as an
application failure.
Unfortunately there aren't any good options for handling it here (no
way to return failure, can't be a build assert as the size is a
runtime argument). A panic will have to do.
Fixes: #67106Fixes: #65584
Signed-off-by: Andy Ross <andyross@google.com>
In many cases suspending or resuming of a device is limited to
just a few register writes. Current solution assumes that those
operations may be blocking, asynchronous and take a lot of time.
Due to this assumption runtime PM API cannot be effectively used
from the interrupt context. Zephyr has few driver APIs which
can be used from an interrupt context and now use of runtime PM
is limited in those cases.
Patch introduces a new type of PM device - synchronous PM. If
device is specified as capable of synchronous PM operations then
device runtime getting and putting is executed in the critical
section. In that case, runtime API can be used from an interrupt
context. Additionally, this approach reduces RAM needed for
PM device (104 -> 20 bytes of RAM on ARM Cortex-M).
Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
k_thread_stack_free() is designed to be called with any pointer
value. We return -EINVAL when an attempt is made to free an
invalid stack pointer.
This change reduces the verbosity in the degenerate case, when
the pointer is not obtained via k_thread_stack_alloc(), but
otherwise does not affect functionality.
If debug log verbosity is not enabled, we save a few bytes in
.text / .rodata / .strtab.
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
This removes z_smp_thread_init() and z_smp_thread_swap() as
SOF has been updated to use k_smp_cpu_custom_start() instead.
This removes the need for SOF to mirror part of the SMP
power up code, and thus these two functions are no longer
needed.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
This provides a path to resume a previously suspended CPU.
This differs from k_smp_cpu_start() where per-CPU kernel
structs are not initialized such that execution context
can be saved during suspend and restored during resume.
Though the actual context saving and restoring are
platform specific.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
This renames z_smp_cpu_start() to k_smp_cpu_start().
This effectively promotes z_smp_cpu_start() into a public API
which allows out of tree usage. Since this is a new API,
we can afford to change it signature, where it allows
an additional initialization steps to be done before a newly
powered up CPU starts participating in scheduling threads
to run.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
This extends the wording so that not only architecture code can
start secondary CPUs at a later time. Also adds a missing 'to'.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
The current z_tick_sleep return directly when building kernel for Single
Thread model. This reorganize the code to use k_busy_wait() to be time
coherent since subsystems may depend on it.
In the case of a K_FOREVER timeout is selected the Single Thread the
implementation will invoke k_cpu_idle() and the system will wait for
an interrupt saving power.
Signed-off-by: Gerson Fernando Budke <gerson.budke@ossystems.com.br>
After a call to k_work_flush returns the sync variable
may still be modified by the workq. This is because
the work queue thread continues to modify the flag in
sync even after k_work_flush returns. This commit adds
K_WORK_FLUSHING_BIT, and with this bit, we moved the
logic of waking up the caller from handle_flush to the
finalize_flush_locked in workq, so that after waking up
the caller, the workqueue will no longer operate on sync.
Fixes: #64530
Signed-off-by: Junfan Song <sjf221100@gmail.com>
It is possible that address + size will overflow the available
address space and the pointer wraps around back to zero. Some
of these have been fixed in previous commits. This fixes
the remaining ones with regard to Z_PHYS_RAM_START/_END,
and Z_VIRT_RAM_START/_END.
Fixes#65542
Signed-off-by: Daniel Leung <daniel.leung@intel.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>
There are several subsystems and boards which require a relatively large
system heap (used by k_malloc()) to function properly. This became even
more notable with the recent introduction of the ACPICA library, which
causes ACPI-using boards to require a system heap of up to several
megabytes in size.
Until now, subsystems and boards have tried to solve this by having
Kconfig overlays which modify the default value of HEAP_MEM_POOL_SIZE.
This works ok, except when applications start explicitly setting values
in their prj.conf files:
$ git grep CONFIG_HEAP_MEM_POOL_SIZE= tests samples|wc -l
157
The vast majority of values set by current sample or test applications
is much too small for subsystems like ACPI, which results in the
application not being able to run on such boards.
To solve this situation, we introduce support for subsystems to specify
their own custom system heap size requirement. Subsystems do
this by defining Kconfig options with the prefix HEAP_MEM_POOL_ADD_SIZE_.
The final value of the system heap is the sum of the custom
minimum requirements, or the value existing HEAP_MEM_POOL_SIZE option,
whichever is greater.
We also introduce a new HEAP_MEM_POOL_IGNORE_MIN Kconfig option which
applications can use to force a lower value than what subsystems have
specficied, however this behavior is disabled by default.
Whenever the minimum is greater than the requested value a CMake warning
will be issued in the build output.
This patch ends up modifying several places outside of kernel code,
since the presence of the system heap is no longer detected using a
non-zero CONFIG_HEAP_MEM_POOL_SIZE value, rather it's now detected using
a new K_HEAP_MEM_POOL_SIZE value that's evaluated at build.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Remove converting bit to string and comparing the string instead of
ready helpers. The "Check if thread is in use" seems to check only
that parameters state_buf and sizeof(state_buf) not zero.
Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>
The pointer parameter 'data' in the function 'k_pipe_put()' ought to
use the const modifier as the contents of the buffer to which it
points never change. Internally, that const modifier is dropped as
both 'k_pipe_get()' and 'k_pipe_put()' share common code for copying
data; however 'k_pipe_put()' never takes a path that modifies those
contents.
Signed-off-by: Peter Mitsis <peter.mitsis@intel.com>
Fix k_sleep compilation error:
[build] ... syscalls/kernel.h:135: undefined reference to `z_impl_k_sleep'
for single thread applications (CONFIG_MULTITHREADING = n).
The shed.c contains source code which must be present also
in single thread applications.
Signed-off-by: Andrej Butok <andrey.butok@nxp.com>
z_free_page_count is only used in one file, so there is
no need to expose it, even to other part of kernel.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Updates both the k_sleep() and k_usleep() return values so that if
the thread was woken up prematurely, they will return the time left
to sleep rounded up to the nearest millisecond (for k_sleep) or
microsecond (for k_usleep) instead of rounding down. This removes
ambiguity should there be a non-zero number of remaining ticks
that correlate to a time of less than 1 millisecond or 1 microsecond.
Signed-off-by: Peter Mitsis <peter.mitsis@intel.com>
Replace open-coded time conversion with the macro which as that will
usually use a constant divide or multiply.
Signed-off-by: Keith Packard <keithp@keithp.com>
Export some symbols for loadable modules. Also add an
EXPORT_SYSCALL() helper macro for exporting system calls by their
official names.
Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
The function _Cstart has already been renamed to z_cstart,
so change the remaining references of it in various docs.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
- issue found with Ztest case of test_thread_timeout_remaining_expires
on Intel ISH platform when adjust CONFIG_SYS_CLOCK_TICKS_PER_SEC
to 10k.
- timeout_rem() return exact remaining ticks which is calibrated by
decrease elapsed(), while z_timeout_expires try to get expire ticks
to be timeout using current tick as base, so need get exact current
ticks by plus elasped().
Signed-off-by: Qipeng Zha <qipeng.zha@intel.com>
Instead of performing a set of relative address comparisons using
pointers of type 'uint8_t *', we leverage the existing IN_RANGE()
macro and perform the comparisons with 'uintptr_t'.
Signed-off-by: Peter Mitsis <peter.mitsis@intel.com>
Add support for mutable devices. Mutable devices are those which
can be modified after declaration, in-place, in kernel mode.
In order for a device to be mutable, the following must be true
* `CONFIG_DEVICE_MUTABLE` must be y-selected
* the Devicetree bindings for the device must include
`mutable.yaml`
* the Devicetree node must include the `zephyr,mutable` property
Signed-off-by: Christopher Friedt <cfriedt@meta.com>
This moves including of demand_paging.h out of kernel/mm.h,
so that users of demand paging APIs must include the header
explicitly. Since the main user is kernel itself, we can be
more discipline about header inclusion.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Add a Kconfig option to tell whether or not using thread
local storage to store current thread.
The function using it can be called from ISR and using
TLS variables in this context may (should ???) not be
allowed
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
This separates demand paging related headers into its own file
instead of being stuffed inside the main kernel memory
management header file.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
This moves the k_* memory management functions from sys/ into
kernel/ includes, as there are kernel public APIs. The z_*
functions are further separated into the kernel internal
header directory.
Also made a quick change to doxygen to group sys_mem_* into
the OS Memory Management group so they will appear in doc.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Some platforms already have .bss section zeroed-out externally before the
Zephyr initialization and there is no sence to zero it out the second time
from the SW.
Such boot-time optimization could be critical e.g. for RTL Simulation.
Signed-off-by: Alexander Razinkov <alexander.razinkov@syntacore.com>
Extends the concept of halting a thread from just aborting a thread
to both aborting and suspending a thread.
Part of this involves updating k_thread_suspend() to operate in a
similar fashion to that of k_thread_abort().
Signed-off-by: Peter Mitsis <peter.mitsis@intel.com>
Extracts the essential thread synchronization logic when aborting
a thread from z_thread_abort() and moves it to its own routine
called z_thread_halt().
Signed-off-by: Peter Mitsis <peter.mitsis@intel.com>
The routine halt_thread() acts nearly identical to end_thread()
except that instead of only halting the thread if the _THREAD_DEAD
state bit is not set, it will halt it if bit specified by the
parameter new_state is not set (which is always _THREAD_DEAD).
Signed-off-by: Peter Mitsis <peter.mitsis@intel.com>
The halt queue will be used to identify threads that are waiting
for a thread on another CPU to finish suspending.
Signed-off-by: Peter Mitsis <peter.mitsis@intel.com>
Basic spinlock implementation is based on single
atomic variable and doesn't guarantee locking fairness
across multiple CPUs. It's even possible that single CPU
will win the contention every time which will result
in a live-lock.
Ticket spinlocks provide a FIFO order of lock aquisition
which resolves such unfairness issue at the cost of slightly
increased memory footprint.
Signed-off-by: Alexander Razinkov <alexander.razinkov@syntacore.com>
