Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:18 as permitted sender) client-ip=2620:137:e000::1:18;
MIME-Version: 1.0
References: <164786042536.122591.4459156564791679956.tglx@xen13>
In-Reply-To: <164786042536.122591.4459156564791679956.tglx@xen13>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Mon, 21 Mar 2022 12:47:01 -0700
Message-ID: <CAHk-=wgfSKuzHggdngkXxSDvRp3FOMwbbvPJqag99NiRekbJfg@mail.gmail.com>
Subject: Re: [GIT pull] core/core for v5.18-rc1
To:     Thomas Gleixner <tglx@linutronix.de>,
        Oleg Nesterov <oleg@redhat.com>
Cc:     Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        "the arch/x86 maintainers" <x86@kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Mon, Mar 21, 2022 at 4:02 AM Thomas Gleixner <tglx@linutronix.de> wrote:
>
>   - Provide a mechanism to delay raising signals from atomic context on
>     PREEMPT_RT enabled kernels as sighand::lock cannot be acquired.  Store
>     the information in the task struct and raise it in the exit path.

This isn't super-pretty, but doing it conditionally makes it even uglier.

I've pulled this, but my reaction to it all is
"force_sig_info_to_task() isn't that critical, and
CONFIG_RT_DELAYED_SIGNALS might as well have been unconditional".

In fact, is there any reason to make it conditional even at run-time, ie this:

> +       if (!in_atomic())
> +               return false;

looks a bit unnecessary too. Why not just always do that thing and
avoid having two code-paths?

I really think we don't want more complexity in this unusual "force
signal on exception" case, and TIF_NOTIFY_RESUME doesn't seem like
it's the wrong place to do it.

Hmm?

Anyway, I've pulled it in this form, I just don't like the special
cases that seem a bit pointless. They not only make the code bigger
and more complex, they actively mean that test coverage is much worse
too.

             Linus