Re: [RFC PATCH v3 2/5] selftests/bpf: add iocost cost model test

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider:
- [Medium] Error path for cgroup storage allocation failure skips IOCOST_COST_F_MERGE adjustment.
- [Low] Multi-line comments do not follow the BPF subsystem strict guidelines.
- [Low] IOC_PAGE_SHIFT is undeclared in the BPF program, causing a compilation failure.
--

commit 1e892fcbbd8ad235ee3890cec52bb54f2b43ddb9
Author: Tao Cui <cuitao@kylinos.cn>

selftests/bpf: add iocost cost model test

This commit adds an example cost model test implementing the builtin linear
HDD formula at double cost. It registers the model via struct_ops and binds
it to a device using io.cost.model, verifying the readback and testing write
errors during restoration.

> diff --git a/tools/testing/selftests/bpf/prog_tests/iocost_model.c b/tools/testing/selftests/bpf/prog_tests/iocost_model.c
> new file mode 100644
> index 0000000000000..bf9f245e29a52
> --- /dev/null
> +++ b/tools/testing/selftests/bpf/prog_tests/iocost_model.c
[ ... ]
> +static int bind_model(const char *dev, const char *name)
> +{
> +	char buf[300], got[64];
> +	int err;
> +
> +	snprintf(buf, sizeof(buf), "%s model=%s\n", dev, name);
> +	err = write_cost_model(buf);
> +	if (err)
> +		return err > 0 ? -err : -EIO;
> +	err = readback_model(dev, got, sizeof(got));
> +	if (!err && !strcmp(got, name)) {
> +		/* restore the builtin linear model on every path so the
> +		 * device is not left bound to a model the caller may
> +		 * unregister right after; the write is checked too */

[Severity: Low]
This isn't a bug, but does this violate the BPF subsystem comment style
guidelines? The BPF subsystem strict guidelines require multi-line comments
to have the opening /* on its own line.

> +		snprintf(buf, sizeof(buf), "%s model=linear\n", dev);
> +		err = write_cost_model(buf);
> +		return err > 0 ? -err : (err ? -EIO : 0);
> +	}
> +
> +	/* bind failed or does not read back: still restore; prefer
> +	 * the original failure, but surface the restore error too */

[Severity: Low]
This isn't a bug, but this comment also places the opening /* on the same
line as the text, which doesn't match the required BPF multi-line comment
style.

> +	snprintf(buf, sizeof(buf), "%s model=linear\n", dev);
> +	{
> +		int rerr = write_cost_model(buf);
> +
> +		return err ? (err > 0 ? -err : -EIO)
> +			   : (rerr > 0 ? -rerr : -EIO);
> +	}
> +}
[ ... ]
> diff --git a/tools/testing/selftests/bpf/progs/iocost_model.c b/tools/testing/selftests/bpf/progs/iocost_model.c
> new file mode 100644
> index 0000000000000..81928f897297c
> --- /dev/null
> +++ b/tools/testing/selftests/bpf/progs/iocost_model.c
[ ... ]
> +#define IOC_PAGE_SIZE		4096
> +#define IOC_SECT_TO_PAGE_SHIFT	3	/* 512B sectors to 4kB pages */
> +#define LCOEF_RANDIO_PAGES	4096	/* 16MB seek threshold */
> +#define IOCOST_COST_F_MERGE	(1ULL << 0)	/* not in BTF: a plain macro */
> +#define IOCOST_REQ_OP_MASK	0xff		/* REQ_OP_MASK, not in BTF */

[Severity: Low]
Is IOC_PAGE_SHIFT missing from these manual constant definitions?

> +
> +/*
> + * DIV64_U64_ROUND_UP / DIV_ROUND_UP_ULL equivalents, folded at
> + * compile time
> + */
[ ... ]
> +SEC("struct_ops")
> +u64 BPF_PROG(iocost_2x_calc_cost, u64 opf, u64 nbytes, u64 sector,
> +	     struct blkcg *blkcg, u64 model_flags)
> +{
> +	u64 pages, seek_pages = 0, base, coef_page, randio, cost;
> +
> +	/* builtin truncates: max(sectors >> IOC_SECT_TO_PAGE_SHIFT, 1) */
> +	pages = nbytes >> IOC_PAGE_SHIFT;

[Severity: Low]
Does this code compile? IOC_PAGE_SHIFT is defined in an anonymous enum within
block/blk-iocost.c and is not exported to vmlinux.h. Attempting to use it
here directly seems like it would cause a build error.

> +	if (!pages)
> +		pages = 1;
> +
> +	if ((opf & IOCOST_REQ_OP_MASK) == REQ_OP_READ) {
[ ... ]
> +	{
> +		__u64 *cursor, cur;
> +		int priced = (opf & IOCOST_REQ_OP_MASK) == REQ_OP_READ ||
> +			     (opf & IOCOST_REQ_OP_MASK) == REQ_OP_WRITE;
> +
> +		cursor = bpf_cgrp_storage_get(&cursor_store,
> +					      blkcg->css.cgroup, NULL,
> +					      BPF_LOCAL_STORAGE_GET_F_CREATE);
> +		if (!cursor)
> +			return 2 * (base + pages * coef_page);

[Severity: Medium]
Does this early return miss the IOCOST_COST_F_MERGE adjustment?

If bpf_cgrp_storage_get() returns NULL, we return immediately here, which
skips the subsequent logic that zeros out base when model_flags has
IOCOST_COST_F_MERGE set. Could this result in merged operations being
incorrectly charged a base cost under failure conditions?

> +		cur = *cursor;
> +		if (cur && priced) {
> +			seek_pages = sector > cur ? sector - cur
> +						   : cur - sector;
> +			seek_pages >>= IOC_SECT_TO_PAGE_SHIFT;
> +			if (seek_pages > LCOEF_RANDIO_PAGES)
> +				base = randio;
> +		}
> +		if (priced && nbytes &&
> +		    !(model_flags & IOCOST_COST_F_MERGE))
> +			*cursor = sector + (nbytes >> 9);
> +	}
> +
> +	if (model_flags & IOCOST_COST_F_MERGE)
> +		base = 0;
> +
> +	cost = 2 * (base + pages * coef_page);
> +	return cost;
> +}

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260914073356.791518-1-cui.tao@linux.dev?part=2