treewide: Merge all of the SDK 1.0 changes together #94368
+227
−84
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The picolibc changes were merged upstream; switch to using that revision. Signed-off-by: Anas Nashif <[email protected]>
Now that rpi_pico can use picolibc, let's switch to it. Signed-off-by: Keith Packard <[email protected]>
This version along the zephyr branch allows cmsis_6 to compile against picolibc. Signed-off-by: Keith Packard <[email protected]>
psa-arch-tests includes device drivers that failed to mark registers with 'volatile'. GCC 14.3 cleverly optimized sequential register accesses using strd/ldrd instructions which caused the drivers to fail. Signed-off-by: Keith Packard <[email protected]>
This patch (submitted upstream) avoids an incorrect warning generated by gcc 14.3 about array bounds. It should not change the generated code at all. Signed-off-by: Keith Packard <[email protected]> Signed-off-by: Anas Nashif <[email protected]>
Disable the default startup file with -nostartfiles. Signed-off-by: Keith Packard <[email protected]>
When the target has only a single CPU, this function cannot ever succeed. Skip all of the drama and just return -EINVAL. This makes GCC 14 happy as it doesn't get confused about possible out of bounds access of the soc_cpus_active array. Signed-off-by: Keith Packard <[email protected]>
The stm32u3x header files defines LL_ADC_SINGLE_ENDED but not LL_ADC_DIFFERENTIAL as the device doesn't support differential mode. The driver only checked for LL_ADC_SINGLE_ENDED and assumed that when that was defined, LL_ADC_DIFFERENTIAL would also be defined. Check for both when figuring out which calibration type will be required. Signed-off-by: Keith Packard <[email protected]>
Zephyr replaces the toolchain version of stdint.h to define uint32_t and int32_t as int rather than long. This breaks the ARM MVE intrinics which require uint32_t to be defined as unsigned long. Define 'ZEPHYR_USE_TOOLCHAIN_STDINT_H_' as that causes zephyr_stdint.h to be skipped, ensuring that the normal stdint.h types are used instead. Signed-off-by: Keith Packard <[email protected]>
The linker may want to know the desired optimization level as when the multilib configuration includes -Os as a selector. Adopt the same pattern as the assembler optimization flags by allowing toolchains to specify linker optimization flags which default to the compiler optimization flags and the apply those using add_link_options. Signed-off-by: Keith Packard <[email protected]>
With inclusion of the optimization flag into the multilib selection process, we cannot compute the compiler library path when the compiler's target.cmake is processed as LINKER_OPTIMIZATION_FLAG is not computed until much later. Instead, add a function (compiler_file_path) which can be used to locate the appropriate crtbegin.o and crtend.o files. Delay computation of lib_include_dir and rt_library until after all compiler flags have been computed by adding compiler_set_linker_properties and calling that just before toolchain_linker_finalize is invoked. Place default implementations of both of these functions in a new file, cmake/compiler/target_template.cmake, where we assume the compiler works like gcc or clang and handlers the --print-file-name and --print-libgcc-file-name options. Compilers needing alternate implementations can override these functions in their target.cmake files. These implementations require that no generator expressions are necessary for the compiler to compute the right library paths. This mechanism is also used to take any additional compiler options by adding a new (optional) linker function, toolchain_linker_add_compiler_options, which maps compiler options to equivalent linker options, discarding any that aren't applicable. Signed-off-by: Keith Packard <[email protected]>
GCC version 14.3 does more extensive checking for potentially uninitialized values and warns about a couple of arrays. Initialize them to zero to make the compiler happy. Signed-off-by: Keith Packard <[email protected]>
This patch has been accepted upstream ARM-software/CMSIS-DSP#265. The module PR, zephyrproject-rtos/cmsis-dsp#7 is pending review. Signed-off-by: Keith Packard <[email protected]>
GCC 14.3 will happily delete any code that appears before __builtin_unreachable that isn't separated with an obvious branch. That includes __asm__ statements, even those which generate traps. The failure case that I debugged was on x86 in z_check_stack_sentinel. There is a store to restore the sentinel to the correct value just before the ARCH_EXCEPT, and that macro emits 'int $32' followed by CODE_UNREACHABLE. Because the compiler didn't understand that ARCH_EXCEPT was changing execution flow, it decided that the sentinel restoring store 'couldn't' be reached and elided it. I added the "memory" clobber to the asm statement in ARCH_EXCEPT before CODE_UNREACHABLE to enforce that all pending store operations be performed before the asm statement occurs. This ensures that they are not deleted by the compiler. I think this might be a GCC bug. The GCC documentation explicitly documents that asm statements which change the flow of control should be followed by __builtin_unreachable. Signed-off-by: Keith Packard <[email protected]>
On arm64, GCC feels free to use floating point registers for essentially anything unless we build with -mgeneral-regs-only. After the FPU gets used in an interrupt handler, interrupts are disabled as there's no place to save the registers during a nested exception. As the C library may be built separately without this flag, we cannot use any of its functions or risk having nested exceptions fail. Switch printk usage to k_str_out and stick to (mostly) Zephyr internal functions to ensure nested interrupts work correctly. Signed-off-by: Keith Packard <[email protected]>
GCC 14 takes a look these functions which are just 'return true' and inlines them even though they has the noinline attribute set. This happens because the compiler computes the possible range of values from the function call, and as that is a single value (true), it replaces the call with that value. So it's not strictly inlining the function? Insist a bit harder by adding a compiler_barrier to the functions which seems to solve the issue. Signed-off-by: Keith Packard <[email protected]>
…stems RX consistently fails this test by a tiny amount. Loosen the tolerance to let it work. Signed-off-by: Keith Packard <[email protected]>
This adds linker script bits and compiler options so that trusted-firmware-m will build with picolibc. This has not been merged to the Zephyr trusted-firmware-m repository yet: zephyrproject-rtos/trusted-firmware-m#134 Signed-off-by: Keith Packard <[email protected]>
zephyrbot
requested review from
57300,
andyross,
ceolin,
cfriedt,
dcpleung,
edersondisouza and
jeremybettis
zephyrbot
requested review from
laurenmurphyx64,
nashif,
nordicjm,
npitre,
peter-mitsis,
stephanosio,
TaiJuWu and
tejlmand
keith-packard
force-pushed
the
keithp-sdk-1.0-all
branch
from
a43add5 to
e825ef3
Compare
github-actions
bot
added
manifest-psa-arch-tests
manifest-cmsis-dsp
manifest-mbedtls
manifest-trusted-firmware-m
manifest-cmsis_6
DNM (manifest)
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Let's see if merging all of the proposed updates for SDK 1.0 support build together.