Re: [bug] splat at boot

[Alexei Starovoitov]

On Thu, Dec 14, 2023 at 5:45 PM Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
>
> On Thu, 14 Dec 2023 17:25:46 -0800
> Alexei Starovoitov <alexei.starovoitov@xxxxxxxxx> wrote:
>
> > Hi All,
> >
> > just noticed a boot splat that probably was there for lone time:
> >
> > [    1.118691] ftrace: allocating 50546 entries in 198 pages
> > [    1.129690] ftrace: allocated 198 pages with 4 groups
> > [    1.130156]
> > [    1.130158] =============================
> > [    1.130159] [ BUG: Invalid wait context ]
> > [    1.130161] 6.7.0-rc3-00837-g403f3e8fda60 #5272 Not tainted
> > [    1.130163] -----------------------------
> > [    1.130165] swapper/0 is trying to lock:
> > [    1.130166] ffff88823fffb1d8 (&zone->lock){....}-{3:3}, at:
> > __rmqueue_pcplist+0xe80/0x1100
> > [    1.130181] other info that might help us debug this:
> > [    1.130182] context-{5:5}
>
> Can you trigger this with CONFIG_PROVE_RAW_LOCK_NESTING disabled?
>
> If not, then I wouldn't worry about it for now, but this will need to be
> addressed when PREEMPT_RT is included.
>
> Basically, a spin_lock() in PREEMPT_RT is converted into a mutex, and most
> interrupt handlers and all softirqs are turned into threads. But there's
> still cases where spin_lock() can not be used. One is for interrupt
> handlers that will not turn into a thread (like the timer interrupt), and
> for when a raw_spin_lock is held. You can't have:
>
>   raw_spin_lock(rawlock);
>   spin_lock(spinlock);
>
> order.

Thanks for explaining.
It's fine without PROVE_RAW_LOCK_NESTING.