Kernel mode support - #13
Draft
raiden00pl wants to merge 14 commits into
Draft
Conversation
raiden00pl
force-pushed
the
kernel-mode-support
branch
3 times, most recently
from
August 11, 2026 13:43
beb9b3e to
38d33ba
Compare
Member
Author
|
this patchset is too big, I'll break it into smaller PRs |
raiden00pl
force-pushed
the
kernel-mode-support
branch
2 times, most recently
from
August 24, 2026 11:54
2a28bfb to
32aec96
Compare
Kernel-mode CMake builds install applications to <build>/bin. On qemu arm/riscv the target mounts them over a semihosting hostfs relative to the spawned process working directory. When the generated .config selects CONFIG_BUILD_KERNEL=y, register app_bindir=<build>/bin and exec_cwd=<build>, keeping values set in YAML. Signed-off-by: raiden00pl <raiden00@railab.me> Assisted-by: Claude Code
Requirements only supported exact [key, value] matching, which cannot express init configuration that differs between build modes: flat builds use CONFIG_INIT_ENTRYPOINT, kernel builds CONFIG_INIT_FILEPATH. Add, backward compatibly: - [key, "*"]: any truthy value satisfies - [[key, value], ...]: alternatives, any one satisfies Also name the product, core and unmet requirement in the abort message. Signed-off-by: raiden00pl <raiden00@railab.me> Assisted-by: Claude Code
With app_bindir configured, cmd_check resolves commands against the application binary directory first and falls back to symbol lookups over bin_debug for names that are not applications (NSH builtins like cmd_df, cmocka entries). Flat-mode behavior is unchanged. Signed-off-by: raiden00pl <raiden00@railab.me> Assisted-by: Claude Code
Runtime checks searched the core ELF or the NSH help output, neither of which lists kernel-mode applications. Route check_cmd through CoreConfig.cmd_check when the core is a kernel build with a known application directory, so runtime checks agree with collection-time filtering. Signed-off-by: raiden00pl <raiden00@railab.me> Assisted-by: Claude Code
With exec_cwd set, QEMU resolved a relative '-kernel' path against the spawn directory instead of the NTFC working directory it is expressed in, so boot timed out. Absolutize the image path in the qemu and sim start paths when a custom spawn directory is configured. Signed-off-by: raiden00pl <raiden00@railab.me> Assisted-by: Claude Code
In kernel builds a faulting user task is killed while the kernel keeps running and the prompt returns, so the console message is the only crash evidence; such faults were reported as command timeouts. Register the fault messages as SEGFAULT signatures: - 'Segmentation fault in' (risc-v) - 'PANIC: Unhandled user exception' (arm64) Signed-off-by: raiden00pl <raiden00@railab.me> Assisted-by: Claude Code
Real hardware has no hostfs, so kernel-mode applications must be shipped as a filesystem image. Add the per-core apps_image option, which generates a ROMFS image from app_bindir with genromfs after the build, and the $APPS_BINDIR and $APPS_IMG flash command placeholders. Signed-off-by: raiden00pl <raiden00@railab.me> Assisted-by: Claude Code
Prebuilt kernel-mode images have no host application directory, so check_cmd could not resolve commands. List CONFIG_PATH_INITIAL (default /system/bin) once on the running target, cache it, and match cmd_check patterns against it with the matcher shared with AppBinDir. A failed listing is not cached and is retried. Signed-off-by: raiden00pl <raiden00@railab.me> Assisted-by: Claude Code
CMake registers application targets at configure time from .config, but kv overrides are applied after configuring, so an override enabling a new application changed .config without ever creating its build target: code-level options took effect through the .config -> config.h rule while the application never appeared in the image. Overrides can also unlock suboptions (e.g. *_PROGNAME) whose missing defaults make cmake drop the application silently. Run olddefconfig and configure again after applying overrides. Signed-off-by: raiden00pl <raiden00@railab.me> Assisted-by: Claude Code
First kernel-mode reference target. The S-mode knsh64 build boots through the QEMU bundled OpenSBI firmware, so '-bios none' is not passed. The stock defconfig mounts hostfs 'fs=../apps', which does not match the CMake layout; override it with 'fs=.' so /system maps to the build directory holding bin/. Signed-off-by: raiden00pl <raiden00@railab.me> Assisted-by: Claude Code
Board-agnostic template for kernel-mode targets on a serial console, documenting the apps_image/$APPS_IMG flashing pattern, prebuilt image discovery and the boot_timeout requirement. Board specifics are filled in once the hardware target is selected. Signed-off-by: raiden00pl <raiden00@railab.me> Assisted-by: Claude Code
Second kernel-mode reference target, exercising the ROMFS deployment path (no hostfs on x86_64): applications are packed into a ROMFS image linked into the kernel. kv overrides supply what the stock defconfig lacks, each annotated in the file. Requires the CMake kernel-build fixes for x86_64 from the vendored NuttX branch fix-x86_64-cmake-kernel-build. Signed-off-by: raiden00pl <raiden00@railab.me> Assisted-by: Claude Code
Kernel-mode reference targets for arm64 and arm, mirroring the riscv knsh64 hostfs pattern: semihosting hostfs with 'fs=.' mapped to the build directory. Signed-off-by: raiden00pl <raiden00@railab.me> Assisted-by: Claude Code
raiden00pl
force-pushed
the
kernel-mode-support
branch
from
August 25, 2026 11:19
32aec96 to
b14ce38
Compare
The gcore tests wrote the "Saved corefile" and GCORE_MARKER reply lines into the pipe before calling generate_coredump(). If the reader thread consumed those lines before generate_coredump() reset _gcore_done and _last_corefile, the reset discarded them and the wait timed out, returning None (seen as a flaky CI failure on the Python 3.14 job). Feed the reply via a proc.stdin.write side effect triggered by the GCORE_MARKER echo command instead, so it arrives only after the controller has issued the gcore command -- mirroring real gdb and making the ordering deterministic. Signed-off-by: raiden00pl <raiden00@railab.me> Assisted-by: Claude Code
raiden00pl
force-pushed
the
kernel-mode-support
branch
from
August 25, 2026 11:44
81b32f0 to
cb413d8
Compare
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
1 participant
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.
changes required for testing nuttx in kernel build