Re: [PATCH bpf-next v2 14/23] bpf: Reject stack arguments if tail call reachable

On 5/9/26 3:42 AM, [email protected] wrote:
> Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
> - [High] Order-dependent DFS logic and incomplete fallback checks allow bypassing the stack arguments restriction and cause false rejections.
> --
>
> commit d6cf15a73bfea82a8d40934257b79e361027a9b1
> Author: Yonghong Song <[email protected]>
>
> bpf: Reject stack arguments if tail call reachable
>
> Tail calls are deprecated and will be replaced by indirect calls
> in the future. Reject programs that combine tail calls with stack
> arguments rather than adding complexity for a deprecated feature.
>
>> diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
>> index b6d07f7a52cdf..95fe21d05c100 100644
>> --- a/kernel/bpf/verifier.c
>> +++ b/kernel/bpf/verifier.c
> [ ... ]
>> @@ -5270,14 +5275,23 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,
>>   	 * this info will be utilized by JIT so that we will be preserving the
>>   	 * tail call counter throughout bpf2bpf calls combined with tailcalls
>>   	 */
>> -	if (tail_call_reachable)
>> +	if (tail_call_reachable) {
>>   		for (tmp = idx; tmp >= 0; tmp = dinfo[tmp].caller) {
>>   			if (subprog[tmp].is_exception_cb) {
>>   				verbose(env, "cannot tail call within exception cb\n");
>>   				return -EINVAL;
>>   			}
>> +			if (subprog_has_stack_args(&subprog[tmp])) {
>> +				verbose(env, "tail_calls are not allowed in programs with stack args\n");
>> +				return -EINVAL;
>> +			}
> Does this create an order dependence where the verifier might falsely reject
> valid programs?
>
> If the main program calls a subprogram with a tail call, the
> tail_call_reachable flag becomes true. If the main program subsequently calls
> an independent subprogram that uses stack arguments, the flag can leak into
> that branch, causing the program to be rejected. Would reversing the call
> order cause the same program to pass?

The requirement here is that if any subprog has both tail_call_reachable
and > 5 arguments, we should reject this bpf program. So ordering issue
does not really matter here.

>
>>   			subprog[tmp].tail_call_reachable = true;
>>   		}
>> +	} else if (!idx && subprog[0].has_tail_call && subprog_has_stack_args(&subprog[0])) {
>> +		verbose(env, "tail_calls are not allowed in programs with stack args\n");
>> +		return -EINVAL;
>> +	}
> Does this fallback check miss nested stack arguments?
>
> If the main program directly contains a tail call, the tail_call_reachable
> variable remains false. If it then calls a wrapper function that passes >5
> arguments to another subprogram, won't the nested stack arguments be missed
> because the fallback only checks subprog[0]?
>
> Can this also bypass restrictions for callback roots?
>
> Since the fallback hardcodes !idx and subprog[0], if an exception callback
> (which is verified with idx > 0) contains a direct tail call and uses outgoing
> stack arguments, it bypasses the main check loop because tail_call_reachable
> is false. Will it also bypass this fallback block because !idx is false?
>
> Could we decouple this restriction from the call graph traversal and globally
> check env->subprog_info for the presence of both features in the program
> instead?

Again, see the above cretiria.