The ELF format allows for multiple string and symbol tables with
complex references between them. This is especially evident when
debugging information is included.
This patch fixes the issues that have been identified with multiple
string tables to allow LLEXT to properly parse those files:
* The symbol table used by LLEXT (LLEXT_MEM_SYMTAB) is now chosen
depending on the loaded file type, and other tables are ignored.
This change is also applied to the SLID injection script.
* The LLEXT string table (LLEXT_MEM_SYMTAB) is now correctly identified
by the symbol table reference, instead of picking the first one.
* VMA range checks only make sense for allocated sections.
Signed-off-by: Luca Burelli <l.burelli@arduino.cc>
Make scripts with an interpreter line executable so that they can be
invoked directly.
Signed-off-by: Sebastian Huber <sebastian.huber@embedded-brains.de>
Mostly a revert of commit b1def7145f ("arch: deprecate `_current`").
This commit was part of PR #80716 whose initial purpose was about providing
an architecture specific optimization for _current. The actual deprecation
was sneaked in later on without proper discussion.
The Zephyr core always used _current before and that was fine. It is quite
prevalent as well and the alternative is proving rather verbose.
Furthermore, as a concept, the "current thread" is not something that is
necessarily architecture specific. Therefore the primary abstraction
should not carry the arch_ prefix.
Hence this revert.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Add UF2 Family ID for Raspberry Pi 2350 and build
UF2 image by default for Pico 2 board
Signed-off-by: Ryan Grachek <grachek@gmail.com>
Signed-off-by: Andrew Featherstone <andrew.featherstone@gmail.com>
The weak syscall symbols generated by gen_syscalls.py are currently
compiled in the LLEXT subsystem library, which is then linked among all
other Zephyr libraries in an unspecified order. This can cause the weak
symbols to override the actual syscall implementations, leading to
undefined behaviour.
To fix this, the currently generated file is split in two elements:
- syscall_exports_llext.c contains the EXPORT_SYMBOL directives for all
syscalls. This part can be compiled with the LLEXT library and linked
among all other Zephyr libraries, and ensures all syscalls symbols
are preserved by the linker.
- syscall_weakdefs_llext.c contains the weak definitions for all syscalls.
This file is compiled in a separate library that is linked last, so
that the weak symbols are only used if no other implementation is
available.
Signed-off-by: Luca Burelli <l.burelli@arduino.cc>
Some linkers (e.g. the ARC MWDT one) don't recognize wildcards in file
names if they are not enclosed in quotes. Looks like the quotes have
no negative effect on the GNU ld linker, so just do it unconditionally.
Signed-off-by: Ilya Tagunov <Ilya.Tagunov@synopsys.com>
The gperf tool automatically selects the optimal data type for the
asso_values table, depending on MAX_HASH_VALUE. However, there is
a corner case when the tables generated on different stages of the
build process have different data types, causing a link-time error.
Upgrade the data type for the table from unsigned char to unsigned
short to at least exclude this 8-bit to 16-bit transition. There is
another potential issue with the 16-bit to 32-bit transition, but
it seems not very likely to have 65k kernel objects anytime soon.
Signed-off-by: Ilya Tagunov <Ilya.Tagunov@synopsys.com>
Extend the device subsystem enumeration script to produce a CMake
pre-load script.
This allow CMake linker generator scripts to create iterable sections
based on output from device subsystem enumeration.
This ensures that same functionality is available in both ld linker
templates and the linker generator.
Update linker generators to support the use of the device subsystem
enumeration CMake pre-load script.
Signed-off-by: Torsten Rasmussen <Torsten.Rasmussen@nordicsemi.no>
The device enumeration feature requires all devices
to place their API implementation in linker sections
by api type. This commit adds a script which uses
the tag __subsystem to identify all existing driver
API types and generate iterable sections for them.
The script is invoked from the top CMakeLists.txt
Signed-off-by: Bjarki Arge Andreasen <baa@trackunit.com>
Co-authored-by: Pieter De Gendt <pieter.degendt@basalte.be>
Signed-off-by: Pieter De Gendt <pieter.degendt@basalte.be>
`_current` is now functionally equals to `arch_curr_thread()`, remove
its usage in-tree and deprecate it instead of removing it outright,
as it has been with us since forever.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
Signed-off-by: Yong Cong Sin <yongcong.sin@gmail.com>
Some auto formatters will wrap long lines of code and insert newlines
that are part of function decls outside of arguments. This change strips
out all newlines so syscall typename regex function as expected.
Signed-off-by: Mark Inderhees <markind@meta.com>
the script shows an incorrect file name, I check the wrong file first, then
I find the script did not print the current reading file name.
Fix this to prevent others from wasting their time on this.
The new error message:
```
[1/179] Generating syscalls.json, struct_tags.json
Error decoding zmk/.../altera_msgdma.c (included in zephyr/.../ethernet.c)
```
Signed-off-by: Sa Sasu <i@sasa.su>
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>
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>
Employs the same linkonce magic of sw_isr_table to fix the
multiple definition of the symtab variables issue that I
get in my application build that doesn't use `west`.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
`start_addr` is the address of the first symbol, rename it to
`first_addr` instead as it seems more intuitive and relatable
to the comments.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
Append new entry to the symtab list only if it has unique
address.
Added a bit more comments and move the debug print to after
the list is sorted.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
The `symtab_find_symbol_name()` is using an adapted binary
search function to get the entry between 2 addresses, we need
to add a dummy entry at the end so that the search function
can remain simple and straightforward without doing
out-of-bound checks:
20 \
|
|
50 x
|
|
90 x
. |
. |
. |
dummy /
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
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>
Use pyelftools to extract the symbol table from the
link stage executable. Then, filter out the function names
and sort them based on their offsets before writing into the
`symtab.c`, this is similar to how the `isr_tables` works.
To access the structure, simply include the new header:
```c
#include <zephyr/debug/symtab.h>
```
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
A new Kconfig option which generates syscall stubs assuming that
extensions will always run on userspace, thus simplifying linking
them, as there's no need for z_impl_ stubs (used for direct syscalls),
CONFIG_LLEXT_EDK_USERSPACE_ONLY.
While defining __ZEPHYR_USER__ could have the same effect for optmised
builds, people building extensions on debug environments - thus
non-optimised - would suffer, as they'd need to somehow make the stubs
available (by either exporting the symbol or implementing dummy stubs).
Signed-off-by: Ederson de Souza <ederson.desouza@intel.com>
Pull in the latest uf2conv.py and newly needed uf2families.json file
to address Python warning on Python 3.12 and keep us in sync.
Signed-off-by: Peter Johanson <peter@peterjohanson.com>
The gen_kobject_list.py script looks at DWARF debug information in the
elf file to determine the address of variables. Make sure that when
looking at DW_FORM_exprloc, it looks at both DW_OP_addr and
DW_OP_plus_uconst.
Signed-off-by: Nick Goote <ngoote@gmail.com>
Since there is a K_THREAD_STACK_LEN, its kernel counterpart
should also be prefixed with K_ for consistency.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
commit 67bb6db3f8 ("syscall: Export all emitted syscalls, enabling
them for extensions") exports all emitted syscalls, however, it does
that only for the `z_mrsh` symbols, effectively only available for
userspace. If an extension running at kernel level tries to use a
syscall, it will fail to load.
This patch fixes that by exposing the `z_impl` symbols instead. However,
this is not as straightforward as the `z_mrsh` ones. As, in their
signatures, they can basically contain any type, it's not just a matter
of emitting `EXPORT_SYMBOL(z_impl_<syscall>)`, as the compiler will
complain about the undefined types. Here, there are a few approaches.
One of them is to have the `EXPORT_SYMBOL` being generated on the same
files where the syscall is implemented - injecting it there would allow
it to access all known symbols. But changing a lot of files is
undesirable, and it was one of the nice points of first patch.
Another one would be to reconstruct - or simply use the absolute path -
for the includes where the syscalls are defined. Reconstruct the paths
seems fragile and I'm not sure using absolute paths is portable.
Finally, the approach used in this patch is to declare, on a different
generated file, all `z_impl_` symbols as `void *` - after all, only the
address (and the name) to the function is relevant to EXPORT_SYMBOL. By
living in an compilation unit that doesn't include any header which
would expose any of the syscalls, there shouldn't be any conflicts. And
to account for the possibility that a syscall is not compiled - due
being configured out via Kconfig - all those symbols are also weak
aliases to a pointer to NULL. This file is then included in
`llext_export.c` (which should naturally not include any conflicting
header).
Signed-off-by: Ederson de Souza <ederson.desouza@intel.com>
Linkable loadable extensions can only use syscalls if they are exported
via EXPORT_SYSCALL (or EXPORT_SYMBOL). Instead of enabling used syscalls
one by one, this patch exports all of them automatically via
`gen_syscalls.py`. If CONFIG_LLEXT=n, the section where the exported
symbols live is discarded, so it should be a non-op when llext is not
enabled.
This patch also removes the now redundant EXPORT_SYSCALL macro. Note
that EXPORT_SYMBOL is still useful on different situations (and is
indeed used by the code generated by `gen_syscalls.py`).
Signed-off-by: Ederson de Souza <ederson.desouza@intel.com>
Remove a message that had a typo in it but, more importantly that was
not needed.
The CONFIG_ISR_TABLES_LOCAL_DECLARATION already selects EXPERIMENTAL so
a user is properly alerted about the consequences just like any other
EXPERIMENTAL features if they enable WARN_EXPERIMENTAL.
Signed-off-by: Benjamin Cabé <benjamin@zephyrproject.org>
According to the Zephyr documentation, the multi-level interrupts
are encoded as follows:
`L1_ID | ((L2_ID + 1) << L2_BITS) | ((L3_ID + 1) << (L3_BITS + L2_BITS))`
This means that when L3 interrupts are enabled, the gen_isr_table.py
script will receive the value of L2_ID + 1. Currently, the script
takes this value and directly compares it with the offsets set via
`CONFIG_3RD_LVL_INTR_xx_OFFSET`. This is wrong because the values from
said configurations are the same as L2_ID and because of that the
script will generate an error. To fix this, use the value of L2_ID
instead of L2_ID + 1.
Signed-off-by: Laurentiu Mihalcea <laurentiu.mihalcea@nxp.com>
This commit moves all the functionality related to the current
interrupt parser into gen_isr_tables_parser_carrays.py file.
The new parser file gen_isr_tables_parser_local.py file is
implemented with the new parser that.
Additional information added to the generated interrupt header
that contains data required by the new parser.
Signed-off-by: Radosław Koppel <radoslaw.koppel@nordicsemi.no>
This commit breaks the code into functional classes.
This way the functionality is visibly splitted into
functional parts and it is easier to replace the
specific parser part to implement new code generators.
There is also common functionality to handle multi level interrupts
moved to configuration class.
Signed-off-by: Radosław Koppel <radoslaw.koppel@nordicsemi.no>
This commit cleans up the gen_isr_tables code for better clarity
and easier modification.
Changes include:
- Separate functions to load intList section.
- Replace spurious handlers in internal data with None.
Now it is the output generator work to choose right function.
- All the work to generate vt and swt separated into its own functions.
- Remove the need for internal shared array - all the information
is here in swt array.
- The update_masks function - more functionality moved here.
- Simplify bit_mask function.
Signed-off-by: Radosław Koppel <radoslaw.koppel@nordicsemi.no>
When using direct isrs, a vector table is needed. However, if none is
present , i.e. `CONFIG_GEN_IRQ_VECTOR_TABLE=n`, this script failed. The
given error message was not helpful (`'NoneType' has no len()`).
This change makes it clearer, where to look for the problem.
Signed-off-by: Greter Raffael <rgreter@baumer.com>
When using the code and data relocation feature, every relocated symbol
would be marked with `KEEP()` in the generated linker script. Therefore,
if any input files contained unused code, then it wouldn't be discarded
by the linker, even when invoked with `--gc-sections`.
This can cause unexpected bloat, or other link-time issues stemming from
some symbols being discarded and others not.
On the other hand, this behavior has been present since the feature's
introduction, so it should remain default for the users who rely on it.
This patch introduces support for `zephyr_code_relocate(... NOKEEP)`.
This will suppress the generation of `KEEP()` statements for all symbols
in a particular library or set of files.
Much like `NOCOPY`, the `NOKEEP` flag is passed to `gen_relocate_app.py`
in string form. The script is now equipped to handle multiple such flags
when passed from CMake as a semicolon-separated list, like so:
"SRAM2:NOCOPY;NOKEEP:/path/to/file1.c;/path/to/file2.c"
Documentation and tests are updated here as well.
Signed-off-by: Grzegorz Swiderski <grzegorz.swiderski@nordicsemi.no>
Rename the bitmask variables from `*_LVL_INTERRUPTS` to
`INTERRUPT_LVL_BITMASK[]` array to be consistent with
`INTERRUPT_BITS`, making it easier to loop over the bitmasks.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
The calculation of `THIRD_LVL_INTERRUPTS` bitmask in the
`update_masks()` function is wrong, the number of bits to shift
should be the sum of the first two levels.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
It's not good to see struct_tags.json change from one build to the next
when nothing changes.
Python's sets are not deterministic, see long story for older commit
f896fc2306 ("scripts: gen_handles: Sort the device handles")
Simply convert multiple_directories to a sorted list before using it.
Fixes commit 80e78208e6 ("kernel: syscalls: no need to include all
syscalls in binary")
Signed-off-by: Marc Herbert <marc.herbert@intel.com>
Python's sets are not deterministic.
`devices` were already sorted but `dev_supports` is still a
non-deterministic set. Sort dev_supports to make the graph output
deterministic.
Fixes commit 29942475c5 ("scripts: gen_handles: output dependency graph")
It is quite ironic that this initial and non-deterministic graph commit
was concurrent with and slightly delayed other commit
f896fc2306 ("scripts: gen_handles: Sort the device handles") which
fixed another, similar non-determinism issue in the same area. A true
"whack-a-mole"!
Signed-off-by: Marc Herbert <marc.herbert@intel.com>
This allows sorting objects consistently without having to specific a
key.
No functional change yet.
Signed-off-by: Marc Herbert <marc.herbert@intel.com>
Luca Burelli
Sebastian Huber
Nicolas Pitre
Ryan Grachek
Ilya Tagunov
Torsten Rasmussen
Bjarki Arge Andreasen
Yong Cong Sin
Pieter De Gendt
Mark Inderhees
Sa Sasu
Florian Grandel
Lingao Meng
Jordan Yates
Mathieu Choplain
Ederson de Souza
Peter Johanson
Nick Goote
Daniel Leung
Benjamin Cabé
Laurentiu Mihalcea
Radoslaw Koppel
Greter Raffael
Grzegorz Swiderski
Marc Herbert
Anas Nashif