Re: [patch 07/12] rseq: Implement syscall entry work for time slice extensions

[Thomas Gleixner <tglx@linutronix.de>]

On Wed, Sep 10 2025 at 10:52, K. Prateek Nayak wrote:
> On 9/9/2025 4:30 AM, Thomas Gleixner wrote:
>> +static inline void rseq_slice_set_need_resched(struct task_struct *curr)
>> +{
>> +	/*
>> +	 * The interrupt guard is required to prevent inconsistent state in
>> +	 * this case:
>> +	 *
>> +	 * set_tsk_need_resched()
>> +	 * --> Interrupt
>> +	 *       wakeup()
>> +	 *        set_tsk_need_resched()
>> +	 *	  set_preempt_need_resched()
>> +	 *     schedule_on_return()
>> +	 *        clear_tsk_need_resched()
>> +	 *	  clear_preempt_need_resched()
>> +	 * set_preempt_need_resched()		<- Inconsistent state
>> +	 *
>> +	 * This is safe vs. a remote set of TIF_NEED_RESCHED because that
>> +	 * only sets the already set bit and does not create inconsistent
>> +	 * state.
>> +	 */
>> +	scoped_guard(irq)
>> +		set_need_resched_current();
>
> nit. any specific reason for using a scoped_guard() instead of just a
> guard() here (and in rseq_cancel_slice_extension_timer()) other than to
> prominently highlight what is being guarded?

Yes, the intention was to highlight it and scoped_guard() really
does. From a code generation perspective it's the same outcome.

Thanks,

        tglx