Re: [PATCH] seqlock,lockdep: Only check for preemption_disabled in non-rt

["Ahmed S. Darwish" <a.darwish@linutronix.de>]

On Fri, May 07, 2021, Davidlohr Bueso wrote:
> This silences the writer hitting this nonsensical warning on PREEMPT_RT.
>
> Reported-by: Shung-Hsi Yu <shung-hsi.yu@suse.com>
> Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
> ---
>  include/linux/seqlock.h | 2 ++
>  1 file changed, 2 insertions(+)
>
> diff --git a/include/linux/seqlock.h b/include/linux/seqlock.h
> index f61e34fbaaea..c8f9253f1a2f 100644
> --- a/include/linux/seqlock.h
> +++ b/include/linux/seqlock.h
> @@ -268,7 +268,9 @@ static inline bool __seqprop_preemptible(const seqcount_t *s)
>
>  static inline void __seqprop_assert(const seqcount_t *s)
>  {
> +#ifndef CONFIG_PREEMPT_RT
>  	lockdep_assert_preemption_disabled();
> +#endif
>  }
>

Nope, it is more complicated than that.

In general, for RT, seqcount_LOCKNAME_t variants should be used instead
of plain seqcount_t, as they can be safely used while preemption is
enabled on the write side.

For plain seqcount_t (which __seqprop_assert() is about), preemption
must be disabled, even for PREEMPT_RT. So the patch above is invalid.

Now, there are still some call sites in the kernel which needs
conversion obviously. I have a large patch series in queue which convert
a number of remaining networking call sites (the changes are locking
algorithm changes, not just direct substitution).

Good luck,

--
Ahmed S. Darwish
Linutronix GmbH