Re: [PATCH v2 sched_ext/for-7.3 11/36] sched_ext: Add CID sharding

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] When primary threads are offline, `calc_shard_layout` undercounts the number of unique cores in the LLC, leading to excessive shard fragmentation.
--

commit dc6e5b38d1262d839593a2f89619eff217ae8a4f
Author: Tejun Heo <tj@kernel.org>

sched_ext: Add CID sharding

This commit introduces CID sharding to partition the CID space into
topology-respecting domains for scalable sub-scheduler operations. Each
shard is a contiguous CID range within a last level cache, capped at a
configured size, with cores distributed evenly across them.

> diff --git a/kernel/sched/ext/cid.c b/kernel/sched/ext/cid.c
> index b0ec4d76f8552..f3c54f4223b43 100644
> --- a/kernel/sched/ext/cid.c
> +++ b/kernel/sched/ext/cid.c

[ ... ]

> +static void calc_shard_layout(const struct cpumask *llc_cpus, u32 shard_size,
> +			      u32 *cores_per_shard_p, u32 *nr_large_p)
> +{
> +	u32 nr_cores = 0, nr_cpus = 0, nr_shards;
> +	int cpu;
> +
> +	for_each_cpu(cpu, llc_cpus) {
> +		nr_cpus++;
> +		if (cpumask_first(topology_sibling_cpumask(cpu)) == cpu)
> +			nr_cores++;

[Severity: Medium]
Does this loop undercount nr_cores if the primary SMT threads are offline?

If cpumask_first(topology_sibling_cpumask(cpu)) returns an offline primary
thread, it will not match the current online cpu being evaluated in llc_cpus.

This can result in nr_cores evaluating to 0, which would then cause
*cores_per_shard_p to be 0. Later in scx_cid_init(), this would force every
core into a separate shard, creating excessive shard fragmentation.

> +	}
> +
> +	nr_shards = max_t(u32, 1, DIV_ROUND_UP(nr_cpus, shard_size));
> +	nr_shards = max_t(u32, nr_shards,
> +			  DIV_ROUND_UP(nr_cpus, SCX_CID_SHARD_MAX_CPUS));
> +
> +	*cores_per_shard_p = nr_cores / nr_shards;
> +	*nr_large_p = nr_cores % nr_shards;
> +}

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260706014058.439853-1-tj@kernel.org?part=11