Draft
Conversation
This patch changes the build to perform two-pass linking to separate rodata
with relocations (copied to RAM) from rodata with no relocations (pure data)
that can be kept in flash by llext.
For dynamic linking mode:
- First pass: Link with build-dynamic.ld to create temp ELF (unchanged).
- Analyze the elf and extract, sort and write out the sections:
- .rodata: sections WITH relocations → copied to RAM (LLEXT_MEM_RODATA)
- .rodata_noreloc: sections WITHOUT relocations → kept in flash (LLEXT_MEM_RODATA_NO_RELOC)
- Second pass: Link with generated rodata_split.ld + build-dynamic.ld.
For static linking mode:
- Generates empty rodata_split.ld (no section separation needed)
Signed-off-by: iabdalkader <i.abdalkader@gmail.com>
Older toolchains will not use the multi-step linking. Make sure the rodata sections are merged into one to avoid wasting memory on small one-item sections at LLEXT link time.
PR arduino#309 addressed the C++ issues that were causing the Ethernet examples to fail to link. With those fixes in place, these examples can be removed from the known issues list. Signed-off-by: Luca Burelli <l.burelli@arduino.cc>
-Os is already used in cpp flags but was missing from cflags. Signed-off-by: iabdalkader <i.abdalkader@gmail.com>
Co-authored-by: Per Tillisch <accounts@perglass.com> Co-authored-by: Luca Burelli <pil@iol.it>
Explicitly state the license in all source files to comply with SPDX best practices. In addition, update the copyright notice to the new format.
- ea_malloc: MIT (derived from https://github.com/embeddedartistry/c-linked-list) - Arduino_LED_Matrix: Apache-2.0 (new code) - Camera: LGPL-3.0+ (derived) - all others: LGPL-2.1+ (derived)
We cannot depend on Zephyr's action since it actively flags any non-Apache2 contribution. Instead, use the official scancode-action and process the resulting JSON file to point out missing or new licenses.
Fix minor typo in comment. Signed-off-by: TOKITA Hiroshi <tokita.hiroshi@gmail.com>
strcpy, memcmp and memset were missing from export, they are however exported by zephyr llext/llext_export.c. Signed-off-by: Andrea Gilardoni <a.gilardoni@arduino.cc>
added time and sys_clock_settime export Signed-off-by: Andrea Gilardoni <a.gilardoni@arduino.cc>
added ldexp symbol export
Make sure the verification passed before deleting temporary files. These may be used in a re-run of verify-core if failed jobs are retried by the user via the Github interface.
This change fixes the errors that are raised when the workflow is triggered by a push event, which does not have a PR number associated with it. Also, no need to error if the PR number is invalid or if the workflow file is changed in the PR - just issue a warning. Signed-off-by: Luca Burelli <l.burelli@arduino.cc>
They appear as regular files but can't be copied without '-r', causing an error and preventing the scan from running properly. Signed-off-by: Luca Burelli <l.burelli@arduino.cc>
Action crashes if no files are found in the scancode-inputs folder.
Otherwise USB_NEXT complains with <wrn> usbd_class: Class instance already registered and callbacks don't get registered (the only real effect is that 1200bps touch doesn't work)
Fix inconsistency introduced by zephyrproject-rtos/zephyr#93739
PortentaC33 and PortentaH7 still have bugs
Provide long-call trampolines for the LLEXT extension. Precompiled toolchain libraries emit R_ARM_THM_JUMP24 / R_ARM_THM_CALL relocations for the export libc functions. These have a +/-16 MB range limit which is exceeded when the LLEXT is loaded in a different memory region than the kernel. Signed-off-by: iabdalkader <i.abdalkader@gmail.com>
CONFIG_POSIX is now frozen and cause import error on new picolib libc because doesn't implement all posix compliance timers. Replace all CONFIG_POSIX with equivalent, more limited apis as reported in the mainline POSIX#Legacy documentation. Signed-off-by: Gilberto Conti <g.conti@arduino.cc>
The LLEXT modules share the loader's thread pool but can't access Thread Local Storage variables (like errno) directly because the compiler bakes in wrong memory offsets at link time. When Zephyr 4.3 broke CONFIG_POSIX_API=y, fixing the POSIX headers required touching ZephyrClient::read(), which exposed a latent TLS issue with the errno check. Since both branches of that check returned 0 anyway (dead code), the simplest correct fix is if (received <= 0) return 0 — no errno access needed. Signed-off-by: Gilberto Conti <g.conti@arduino.cc>
With local-exec TLS, the generated code is essentially: BLX __aeabi_read_tp ; get thread pointer → r0 LDR r1, [r0, #errno_off] ; add a CONSTANT offset baked in at link time The errno_off constant is computed by the linker at LLEXT link time using an R_ARM_TLS_LE32 relocation. To compute it correctly, the linker needs to know the full TLS block layout of the loader binary — every TLS variable from every module (picolibc, Zephyr kernel, etc.) in their final positions. The LLEXT linker does not have that information. It only sees the llext-edk headers and syms-dynamic.ld stubs. So the offset it bakes in is very likely wrong — the LLEXT would call the right __aeabi_read_tp, get the right thread pointer, then read from the wrong offset and get garbage instead of errno. The new exported function arduino_errno_ptr() runs inside the loader, where local-exec is valid and correct. It computes thread_ptr + errno_off (the loader's errno), and returns a plain int *. An alternative fix is to ignore errno and just return 0 as done before. Signed-off-by: Gilberto Conti <g.conti@arduino.cc>
Signed-off-by: Gilberto Conti <g.conti@arduino.cc>
Signed-off-by: Gilberto Conti <g.conti@arduino.cc>
Signed-off-by: Gilberto Conti <g.conti@arduino.cc>
Signed-off-by: Gilberto Conti <g.conti@arduino.cc>
Signed-off-by: Gilberto Conti <g.conti@arduino.cc>
Signed-off-by: Gilberto Conti <g.conti@arduino.cc>
|
|
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
11 participants
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.
This is a first porting of portenta_ota
The main "change" is that now the ota payload contains
<llext loader + padding + user sketch>Features
Boards
Step to reproduce
extra/ota-pack.otafile from the output dirpython -m http.server 8000in the directory that contains the sketch