Re: [PATCH] rcu/tree: Add comment to describe GP done condition in fqs loop

[Joel Fernandes <joel@joelfernandes.org>]

On Thu, Jun 09, 2022 at 12:43:40PM +0530, Neeraj Upadhyay wrote:
> Add a comment to explain why !rcu_preempt_blocked_readers_cgp() condition
> is required on root rnp node, for GP completion check in rcu_gp_fqs_loop().
> 
> Signed-off-by: Neeraj Upadhyay <quic_neeraju@quicinc.com>
> ---
>  kernel/rcu/tree.c | 12 +++++++++++-
>  1 file changed, 11 insertions(+), 1 deletion(-)
> 
> diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> index a93c5f4d7e09..9cd1ba512fdc 100644
> --- a/kernel/rcu/tree.c
> +++ b/kernel/rcu/tree.c
> @@ -2083,7 +2083,17 @@ static noinline_for_stack void rcu_gp_fqs_loop(void)
>  		rcu_gp_torture_wait();
>  		WRITE_ONCE(rcu_state.gp_state, RCU_GP_DOING_FQS);
>  		/* Locking provides needed memory barriers. */
> -		/* If grace period done, leave loop. */
> +		/*
> +		 * If grace period done, leave loop. rcu_preempt_blocked_readers_cgp(rnp)
> +		 * check is required for the case where we only have single node in the
> +		 * rcu_node tree; so, root rcu node is also the leaf node, where readers
> +		 * blocking current gp are queued. For multi-node tree, checking ->qsmask
> +		 * on the root node is sufficient, as root rcu node's ->qsmask is only
> +		 * cleared, when all leaf rcu nodes have propagated their quiescent
> +		 * state to their parent node, which happens only after both ->qsmask
> +		 * and rcu_preempt_blocked_readers_cgp(rnp_leaf) are cleared for those
> +		 * leaf nodes.
> +		 */
>  		if (!READ_ONCE(rnp->qsmask) &&
>  		    !rcu_preempt_blocked_readers_cgp(rnp))
>  			break;

Paul's wording changes are OK with me.

For the patch:
Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>

thanks,

 - Joel