On Fri, Jul 18, 2025 at 05:44:12PM +0100, Al Viro wrote:
On Thu, Jul 17, 2025 at 10:48:08AM +0200, Thomas Weißschuh wrote:
The KUnit UAPI infrastructure starts userspace processes. As it should be able to be built as a module, export the necessary symbols.
Signed-off-by: Thomas Weißschuh thomas.weissschuh@linutronix.de
No. This is just plain wrong. This is way too low-level; teach kernel/umh.c to provide what you need, but do *not* add more kernel_execve() callers.
Sounds good.
And the situation with ramfs needs cleaning up, but "export put_filesystem()" is not a solution.
Cleaning up would mean to stop calling put_filesystem(), as it is a no-op here anyways, right?
This would still leave the exports for replace_fd(), create_pipe_files() and set_fs_pwd(). Instead of using kernel/umh.c, I can also extend kernel/usermode_driver.c to provide these in a way that works for me. But kernel/usermode_driver.c is dead code, unused since commit 98e20e5e13d2 ("bpfilter: remove bpfilter") Would it be fine to export those symbols? And delete usermode_driver.c, as carrying around an unused generic framework seems pointless.
Thomas