Re: [PATCH bpf-next v1 3/5] selftests/bpf: Test referenced kptr arguments of struct_ops programs

[Martin KaFai Lau <martin.lau@xxxxxxxxx>]

On 2/14/25 8:45 AM, Amery Hung wrote:
> diff --git a/tools/testing/selftests/bpf/progs/struct_ops_refcounted_fail__global_subprog.c b/tools/testing/selftests/bpf/progs/struct_ops_refcounted_fail__global_subprog.c
> new file mode 100644
> index 000000000000..ae074aa62852
> --- /dev/null
> +++ b/tools/testing/selftests/bpf/progs/struct_ops_refcounted_fail__global_subprog.c
> @@ -0,0 +1,39 @@
> +#include <vmlinux.h>
> +#include <bpf/bpf_tracing.h>
> +#include "../test_kmods/bpf_testmod.h"
> +#include "bpf_misc.h"
> +
> +char _license[] SEC("license") = "GPL";
> +
> +extern void bpf_task_release(struct task_struct *p) __ksym;
> +
> +__noinline int subprog_release(__u64 *ctx __arg_ctx)
> +{
> +	struct task_struct *task = (struct task_struct *)ctx[1];
> +	int dummy = (int)ctx[0];
> +
> +	bpf_task_release(task);
> +
> +	return dummy + 1;
> +}
> +
> +/* Test that the verifier rejects a program that contains a global
> + * subprogram with referenced kptr arguments
> + */
> +SEC("struct_ops/test_refcounted")
> +__failure __log_level(2)
> +__msg("Validating subprog_release() func#1...")
> +__msg("invalid bpf_context access off=8. Reference may already be released")
> +int refcounted_fail__global_subprog(unsigned long long *ctx)
> +{
> +	struct task_struct *task = (struct task_struct *)ctx[1];
> +
> +	bpf_task_release(task);
> +
> +	return subprog_release(ctx);

One question, swap the subprog_release and bpf_task_release order will still be 
the same failure, right?  Meaning:

	subprog_release(ctx);

	bpf_task_release(task);

	return 0;

which is fine based on the changes in the do_check_common() in patch 2. Just 
want to confirm my understanding.

Other than that,

Acked-by: Martin KaFai Lau <martin.lau@xxxxxxxxxx>