Re: [RFC PATCH v2 4/5] selftests/bpf: add multi-stream sequentiality example model

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Concurrent accesses and modifications to shared cgroup storage without synchronization.
--

commit 00d96497e1688fa01fa5b76f7ec3a10048823d74
Author: Tao Cui <cuitao@kylinos.cn>

selftests/bpf: add multi-stream sequentiality example model

This commit introduces a new example cost model for iocost using eBPF,
replacing the single-cursor sequentiality heuristic with a per-cgroup table
of stream slots. The stream state is maintained in a CGRP_STORAGE map keyed
by the cgroup.

> diff --git a/tools/testing/selftests/bpf/progs/iocost_ms.c b/tools/testing/selftests/bpf/progs/iocost_ms.c
> new file mode 100644
> index 0000000000000..3cc57c03d3c88
> --- /dev/null
> +++ b/tools/testing/selftests/bpf/progs/iocost_ms.c

[ ... ]

> +	now = bpf_ktime_get_ns();
> +	for (i = 0; i < NSLOTS; i++) {
> +		if (s->expected[i] == sector && s->stamp[i]) {
> +			found = i;
> +			break;
> +		}
> +	}
> +	if (found != 0xFFFFFFFF) {
> +		/* sequential: keep the seq base from the op branch */
> +		s->expected[found] = sector + advance;
> +		s->stamp[found] = now;

[Severity: Medium]
Are these lockless updates susceptible to data races? Multiple CPUs processing
IO for the same cgroup could concurrently execute iocost_ms_calc_cost() and
modify the shared cgroup storage.

Could this concurrency result in lost updates or interleaved reads and writes,
leading to inaccuracies in the sequential IO heuristic?

> +	} else {
> +		base = randio;
> +		for (i = 1; i < NSLOTS; i++) {
> +			if (s->stamp[i] < s->stamp[victim])
> +				victim = i;
> +		}
> +		s->expected[victim] = sector + advance;
> +		s->stamp[victim] = now;

[Severity: Medium]
Similarly, when claiming and updating a new stream slot, could multiple
CPUs select the same victim concurrently?

While lockless updates are often an intentional performance tradeoff in IO
hotpaths and this is an example model, could the resulting data races on the
expected and stamp arrays cause unintended behavior when tracking interleaved
streams?

> +	}
> +
> +	pages = RU(nbytes, IOC_PAGE_SIZE);

[ ... ]

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260910125817.223354-1-cui.tao@linux.dev?part=4