UlasSertan opened a new pull request, #20106:
URL: https://github.com/apache/nuttx/pull/20106
### Summary
* The linker script gave `.data` a separate load address
(`.data : AT (_sidata)`, `_sidata = _etext`), but this is a
RAM/remoteproc
image (`CONFIG_BOOT_RUNFROMISRAM=y`) whose `arm_data_initialize` only
zeros
`.bss` — the LMA→VMA copy is compiled out under
`CONFIG_BOOT_RUNFROMFLASH`.
* `_sidata` was taken before `.init_section`, so `.data`'s load address sat
`sizeof(.init_section)` below its run address, and remoteproc loads
segments
by `p_paddr`.
* That section is empty in both in-tree configurations, so the two
addresses
coincide today and the fault is **latent**: `LMA == VMA` already, and
this
change is a no-op for the shipped `nsh` and `netnsh` builds.
* It becomes real as soon as anything lands in `.init_array` — a C++ static
constructor, for instance — which is how it was first seen: initialised
globals read shifted and the console pinmux table took a data abort in
`arm_boot` before the console came up.
* Dropping the `AT()` makes `.data` load at its run address (LMA == VMA),
which is what is required when nothing copies `.data` at boot.
### Impact
* Is new feature added? Is existing feature changed? **NO** — bug fix.
* Impact on user? **NO**.
* Impact on build? **NO** (link layout of one board only).
* Impact on hardware? **YES** — t3-gem-o1 only.
* Impact on documentation? **NO**.
* Impact on security? **NO**.
* Impact on compatibility? **NO**.
* Anything else? No dependency on other PRs in this series.
### Testing
```
Build Host(s): Pardus GNU/Linux 25 (Debian 13 base), x86_64, host gcc 14
Toolchain: arm-none-eabi-gcc 14.2.1 20241119 (15:14.2.rel1-1)
Target: arm / TI AM67 (J722S) Cortex-R5F, board t3-gem-o1:nsh
nuttx-apps: 5a7ab4200
Loaded on the main-domain R5F by Linux remoteproc; console on UART-MAIN1.
```
Testing logs before change: not reproducible in the in-tree configurations --
`.init_section` is empty there, so `_sidata` already equals `_sdata` and the
pre-change image links with `LMA == VMA`. The fault was originally observed
in
a build with C++ static constructors. Verified by inspection instead:
```
$ arm-none-eabi-nm nuttx | grep -E ' _etext| _sinit| _einit| _sdata'
a2235168 A _einit
a2235168 A _etext
a2235168 A _sdata
a2235168 A _sinit
```
Because the fault is latent, the pre-change image boots too. The log below
therefore shows only that this change does no harm -- it is not evidence of a
failure being fixed.
The version string in the logs is from the tested build; the commits were
GPG-signed afterwards, which rewrites hashes. The code is unchanged.
Testing logs after change:
```
make distclean && ./tools/configure.sh t3-gem-o1:nsh && make -j
-> nuttx ELF 511876 bytes, 0 compiler warnings
Console:
nsh> echo hwtest-alive
nsh> cat /proc/version
NuttX version 0.0.0 8ec1e42bf0 Sep 10 2026 13:47:37 t3-gem-o1:nsh
nsh> ls /proc
nsh> cat /proc/meminfo
nsh> ps
Test plan result:
console_alive PASS prompt responsive
version PASS NuttX version 0.0.0 8ec1e42bf0 Sep 10 2026
13:47:37 t3-gem-o1:nsh
procfs PASS procfs readable
tasks PASS 2 task lines
PASS=4
```
### PR verification Self-Check
* [x] This PR introduces only one functional change.
* [x] I have updated all required description fields above.
* [x] My PR adheres to Contributing Guidelines and Documentation.
* [ ] My PR is still work in progress (not ready for review).
* [x] My PR is ready for review and can be safely merged into a codebase.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
