Now that section header cache is persistent, it can be used for
finding sections. Add a new function, similar to llext_find_section()
to use it and deprecate llext_find_section().
Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
When building an LLEXT-enabled kernel, 62b19ef65c added weak aliases
of all syscall implementation functions to a pointer to NULL, with the
assumption that LLEXT would check the required symbols at link time and
fail if any of them were found.
This check, however, is ineffective in the current implementation: the
actual address that is exported is the rather normal-looking location of
the variable containing the NULL pointer. This defeats the NULL symbol
validity checks in llext_link.c and causes the extension to crash at
runtime by jumping to a location containing a few zeroes in read-only
data memory.
This commit makes sure the alias target is actually placed at address 0
using the llext-sections.ld linker fragment, so that undefined syscall
implementations are exported as NULLs and as such properly flagged at
link time.
The test for this functionality is also updated to reflect the change.
Signed-off-by: Luca Burelli <l.burelli@arduino.cc>
LLEXT tests currently fail on platforms where a console was chosen but
its underlying device is not enabled. This might happen for example on
multicore SoCs, where the console is only enabled on the most powerful
CPU.
Fix this by only allowing testing the device API if the console device
is both chosen and enabled in the final DT.
Signed-off-by: Luca Burelli <l.burelli@arduino.cc>
This commit introduces architecture-specific ELF relocations for RISC-V,
in accordance with the RISC-V PSABI specification:
https://github.com/riscv-non-isa/riscv-elf-psabi-doc/blob/master/riscv-elf.adoc
Also, the necessary compiler configurations for compiling LLEXT
extensions on RISC-V are added, and the llext tests are executed on
RISC-V targets.
Calling llext extensions from user threads in RISC-V is still
unsupported as of this commit.
Signed-off-by: Eric Ackermann <eric.ackermann@cispa.de>
Add a new test to the simple llext test suite that verifies that DT
devices are properly exported to extensions. This is done by obtaining a
device handle from the DT and then trying to get the same handle at
runtime using the device_get_binding() API.
This test is optional because the current qemu_xtensa platform does not
have any usable DT devices to test with.
The LLEXT heap size is also increased to 64 kbytes to avoid heap
exhaustion when running the new test on mps2_an385.
Signed-off-by: Luca Burelli <l.burelli@arduino.cc>
test_dependent() was changed to test_entry() but one location was
missed - in a warning message string. Fix it too.
Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
This test uses two LLEXT objects: one of them exports a symbol and
the other one calls the exported function.
Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
When identifying the size of an extension image we simply need its
size in bytes, not the number of array elements, that contain it.
Both happen to return the same value, because we use arrays of bytes,
but using sizeof() is clearer.
Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
Add a test to check the proper support of ELF init arrays, using
the new llext_bootstrap function in place of llext_entry in the
LLEXT test suite.
Signed-off-by: Luca Burelli <l.burelli@arduino.cc>
This commit renames the perm_setup callback to test_setup and provides
the extension as an additional parameter. It also adds a test_cleanup
callback that is called after each test completes.
Setup and cleanup functions are now called regardless of whether
CONFIG_USERSPACE is enabled or not.
Signed-off-by: Luca Burelli <l.burelli@arduino.cc>
This patch refactors the macro for test case definition to simplify the
generic case. By using variable macro arguments and default-0 fields,
the most common case does not require arguments at all and tests that
have special requirements can be defined with only the necessary fields.
Signed-off-by: Luca Burelli <l.burelli@arduino.cc>
Extend tests with case where extension is copied from RAM buffer
into filesystem and then loaded from filesystem.
Signed-off-by: Adam Wojasinski <awojasinski@baylibre.com>
Replace LL_EXTENSION_SYMBOL() with EXPORT_SYMBOL() in all tests and
samples.
Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
Add a new test case that verifies that llext_find_section() returns the
correct offset for a symbol in a loadable extension. This exploits the
fact that in the STORAGE_WRITABLE cases, the symbol addresses will be
inside the ELF file buffer, so they can be easily compared.
Signed-off-by: Luca Burelli <l.burelli@arduino.cc>
The ELF files should be aligned to at least sizeof(elf_word) to avoid
issues. Use a larger value to ease debugging, since it reduces the
differences in addresses between similar runs.
Signed-off-by: Luca Burelli <l.burelli@arduino.cc>
Some tests depend on global initialized variables, which are then
modified during the test. When enabling user mode, the functions are
called multiple times but the variables are not reinitialized, resulting
in a test failure.
This patch adds a run_id variable to the tests to reinitialize the
variables when the test is called multiple times. It is purposefully
initialized to 41 at startup to detect if the variable is not set up
properly by the llext loader.
Signed-off-by: Luca Burelli <l.burelli@arduino.cc>
Most of the current llext tests did not actually check the results of
the tests, so the CI reported a pass even if something was wrong when
looking at the logs. This patch adds the appropriate zassert_* macros to
all the tests, to ensure that they actually perform correctly.
Signed-off-by: Luca Burelli <l.burelli@arduino.cc>
This commit introduces support for an alternate linking method in the
LLEXT subsystem, called "SLID" (short for Symbol Link Identifier),
enabled by the CONFIG_LLEXT_EXPORT_BUILTINS_BY_SLID Kconfig option.
SLID-based linking uses a unique identifier (integer) to identify
exported symbols, instead of using the symbol name as done currently.
This approach provides several benefits:
* linking is faster because the comparison operation to determine
whether we found the correct symbol in the export table is now an
integer compare, instead of a string compare
* binary size is reduced as symbol names can be dropped from the binary
* confidentiality is improved as a side-effect, as symbol names are no
longer present in the binary
Signed-off-by: Mathieu Choplain <mathieu.choplain@st.com>
This test checks that the pre_located feature works as expected. It
creates a new extension that is manually relocated to a specific address
via the add_llext_command() CMake function invoking a custom linker
command. The test then loads the extension setting the pre_located
option and checks that the symbol is resolved properly.
Signed-off-by: Luca Burelli <l.burelli@arduino.cc>
Namespaced the generated headers with `zephyr` to prevent
potential conflict with other headers.
Introduce a temporary Kconfig `LEGACY_GENERATED_INCLUDE_PATH`
that is enabled by default. This allows the developers to
continue the use of the old include paths for the time being
until it is deprecated and eventually removed. The Kconfig will
generate a build-time warning message, similar to the
`CONFIG_TIMER_RANDOM_GENERATOR`.
Updated the includes path of in-tree sources accordingly.
Most of the changes here are scripted, check the PR for more
info.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
Add a case, testing building an llext from multiple files, calling
functions and accessing data across files.
Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
The latest commits allow llexts to be built on any architecture, even on
those where llext loading is still unsupported, which is very useful for
testing purposes. This means that the LOADER_BUILD_ONLY logic that was
added for this purpose in 1408d1e5b8 is no longer needed.
Signed-off-by: Luca Burelli <l.burelli@arduino.cc>
LLEXT on Xtensa now supports both shared and relocatable (partially
linked) extensions. This commit adds a copy of the LLEXT test for the
relocatable case.
Signed-off-by: Luca Burelli <l.burelli@arduino.cc>
Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
Check if syscalls can be accessed from both kernel and userspace, and if
optimised away ones indeed point to NULL.
Signed-off-by: Ederson de Souza <ederson.desouza@intel.com>
Fix conflict between commit ce24394437 ("llext: add object test case")
and commit 1408d1e5b8 ("tests: llext: compile architectures not
supported yet") which were tested separately but merged at the same
time.
Github "Merge Queues" can avoid this (and save resources) but:
- they're not used by Zephyr CI
- they provide confusing feedback
Signed-off-by: Marc Herbert <marc.herbert@intel.com>
For now llext supports a very limited number of architectures. This
restriction is enforced by add_llext_target() in CMake at configuration
time.
Add a new `LOADER_BUILD_ONLY` conditional in tests/subsys/llext/simple/
and a new `llext.simple.loader_build`, `build_only` test that does not
invoke `add_llext_target()` and only compiles the llext framework code.
This helps find and fix bugs in `subsys/llext/*.c` and make it ready to
be used when add_llext_target() limitations are lifted.
Note this is pure `tests/` change without any change in the actual llext
framework code. The existing test is only modified to conditionally
invoke add_llext_target().
Signed-off-by: Marc Herbert <marc.herbert@intel.com>
Validate the new relocations for BL and BLX instructions by
creating a new test extension which contains a chain of global
functions in a pseudo random order to (hopefully) generate
relative jumps in both positive and negative directions.
Signed-off-by: Bjarki Arge Andreasen <bjarki@arge-andreasen.me>
The simple test is there to test the API and simple extensions in
unison. Hello world was intended to be the first not the only extension
being tested.
Also refactors the entry thread to allow for usermode potentially by
passing the pointer to the function symbol rather than having it look it
up directly.
Signed-off-by: Tom Burdick <thomas.burdick@intel.com>
