On 2/13/25 04:35, Alexei Starovoitov wrote:
> From: Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx>
>
> In !PREEMPT_RT local_lock_irqsave() disables interrupts to protect
> critical section, but it doesn't prevent NMI, so the fully reentrant
> code cannot use local_lock_irqsave() for exclusive access.
>
> Introduce localtry_lock_t and localtry_lock_irqsave() that
> disables interrupts and sets acquired=1, so localtry_lock_irqsave()
> from NMI attempting to acquire the same lock will return false.
>
> In PREEMPT_RT local_lock_irqsave() maps to preemptible spin_lock().
> Map localtry_lock_irqsave() to preemptible spin_trylock().
> When in hard IRQ or NMI return false right away, since
> spin_trylock() is not safe due to PI issues.
>
> Note there is no need to use local_inc for acquired variable,
> since it's a percpu variable with strict nesting scopes.
>
> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx>
> Signed-off-by: Alexei Starovoitov <ast@xxxxxxxxxx>
...
>
> +/* localtry_lock_t variants */
> +
> +#define __localtry_lock_init(lock) \
> +do { \
> + __local_lock_init(&(lock)->llock); \
> + WRITE_ONCE(&(lock)->acquired, 0); \
This needs to be WRITE_ONCE((lock)->acquired, 0);
I'm adopting this implementation for my next slab sheaves RFC. But I'll want
localtry_trylock() without _irqsave too, so I've added it locally. Posting
below with the init fix and making the PREEMPT_RT comment clear. Feel free
to fold everything, it would make it easier for me. Or just the fixes, if
you don't want code you don't use yourself.
----8<----
