[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <ZQBxPmVmBJ7J+b4B@casper.infradead.org>
Date: Tue, 12 Sep 2023 15:10:06 +0100
From: Matthew Wilcox <willy@...radead.org>
To: "Liam R. Howlett" <Liam.Howlett@...cle.com>,
"Paul E. McKenney" <paulmck@...nel.org>,
Geert Uytterhoeven <geert@...ux-m68k.org>,
Andrew Morton <akpm@...ux-foundation.org>,
maple-tree@...ts.infradead.org, linux-mm@...ck.org,
linux-kernel@...r.kernel.org, stable@...r.kernel.org,
linux-renesas-soc@...r.kernel.org,
Shanker Donthineni <sdonthineni@...dia.com>
Subject: Re: [PATCH v2 1/2] maple_tree: Disable mas_wr_append() when other
readers are possible
On Mon, Sep 11, 2023 at 07:54:52PM -0400, Liam R. Howlett wrote:
> Taking the mutex lock in kernel/irq/manage.c __setup_irq() is calling a
> cond_resched().
>
> >From what Michael said [1] in this thread, since something has already
> set TIF_NEED_RESCHED, it will eventually enable interrupts on us.
>
> I've traced this to running call_rcu() in kernel/rcu/tiny.c and
> is_idle_task(current) is true, which means rcu runs:
> /* force scheduling for rcu_qs() */
> resched_cpu(0);
>
> the task is set idle in sched_init() -> init_idle() and never changed,
> afaict.
Should calling init_idle() be deferred until after interrupts are
all set up?
Powered by blists - more mailing lists