Re: [PATCH] hugetlb: select PREEMPT_COUNT if HUGETLB_PAGE for in_atomic use

From: Michal Hocko <mhocko@suse.com>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>,
	linux-mm@kvack.org, linux-kernel@vger.kernel.org,
	"Paul E . McKenney" <paulmck@kernel.org>,
	Shakeel Butt <shakeelb@google.com>,
	tglx@linutronix.de, john.ogness@linutronix.de, urezki@gmail.com,
	ast@fb.com, Eric Dumazet <edumazet@google.com>,
	Mina Almasry <almasrymina@google.com>,
	Andrew Morton <akpm@linux-foundation.org>
Subject: Re: [PATCH] hugetlb: select PREEMPT_COUNT if HUGETLB_PAGE for in_atomic use
Date: Thu, 11 Mar 2021 13:49:17 +0100	[thread overview]
Message-ID: <YEoRzSw4xoEY6SPv@dhcp22.suse.cz> (raw)
In-Reply-To: <YEoA08n60+jzsnAl@hirez.programming.kicks-ass.net>

On Thu 11-03-21 12:36:51, Peter Zijlstra wrote:
[...]
> Also, Linus hates constructs like this:
> 
>   https://lkml.kernel.org/r/CAHk-=wht7kAeyR5xEW2ORj7m0hibVxZ3t+2ie8vNHLQfdbN2_g@mail.gmail.com

Btw. I would really appreciate if somebody would explain why it is
_fundamentally broken_ to check for an atomic context and chose a
different handling in a code path, like put_page, which is out of hands
of the called context? This can be called from a wide variety of
contexts. There is no way to pass a context information to the called
function. I do recognize that this is not an act of beauty but why
fundamentally broken?

The put_page context can certainly work towards robustness and operate
on the most restrictive context grounds (I really hope nobody will ever
come up with an idea that put_page can be called from nmi context). This
can make the code more complex and less optimal in normal case (e.g.
hugetlb is almost never freed from an atomic context - one has to be
really creative to achieve that). So where do we draw a line?

-- 
Michal Hocko
SUSE Labs