On Mon, Mar 28, 2022 at 4:56 PM Eric W. Biederman <ebiederm@xxxxxxxxxxxx> wrote:
>
> The removal of tracehook.h is quite significant as it has been a major
> source of confusion in recent years. Much of that confusion was
> around task_work and TIF_NOTIFY_SIGNAL (which I have now decoupled
> making the semantics clearer).
Hmm. I love removing tracehook.c, but this looks like it hasn't been
in linux-next.
The header file changes messes with other changes, and we have
kernel/sched/fair.c:2884:9: error: implicit declaration of function
‘init_task_work’; did you mean ‘init_irq_work’?
[-Werror=implicit-function-declaration]
2884 | init_task_work(&p->numa_work, task_numa_work);
| ^~~~~~~~~~~~~~
as a result (also a few other things in that same file).
Now, this is trivial to fix - just add an include for
<linux/task_work.h> from that file - and that's the right thing to do
anyway.
But I'm a bit unhappy that this was either not tested in linux-next,
or if it was, I wasn't notified about the semantic in the pull
request.
So I've pulled this, and fixed up things in my merge, but I'm a bit
worried that there might be other situations like this where some
header file is no longer included and it was included implicitly
before through that disgusting tracehook.h header..
I *hope* it was just the scheduler header file updates that ended up
having this effect, and nothing else is affected.
Let's see if the test robots start complaining about non-x86
architecture-specific stuff that I don't build test.
Linus
