Re: [RFC PATCH bpf-next v2 4/6] bpf: Add common attr support for map_create

[Leon Hwang <leon.hwang@linux.dev>]

On 2025/9/19 07:29, Eduard Zingerman wrote:
> On Fri, 2025-09-12 at 00:33 +0800, Leon Hwang wrote:
>
> [...]
>
>> @@ -1355,6 +1356,18 @@ static int map_create(union bpf_attr *attr, bool kernel)
>>  	if (err)
>>  		return -EINVAL;
>>
>> +	if (common_attrs->log_buf) {
>> +		log = kvzalloc(sizeof(*log), GFP_KERNEL);
>> +		if (!log)
>> +			return -ENOMEM;
>> +		err = bpf_vlog_init(log, BPF_LOG_FIXED, u64_to_user_ptr(common_attrs->log_buf),
>> +				    common_attrs->log_size, NULL);
>
> Maybe use common_attrs->log_level instead of BPF_LOG_FIXED?
> Just for consistent behavior with program and btf load operations.
>

It doesn’t really make sense to let users set log_level in this case,
because logging in map_create is too simple for different log levels to
have any meaningful effect.

>> +		if (err) {
>> +			kvfree(log);
>> +			return err;
>> +		}
>> +	}
>> +
>>  	/* check BPF_F_TOKEN_FD flag, remember if it's set, and then clear it
>>  	 * to avoid per-map type checks tripping on unknown flag
>>  	 */
>
> [...]
>
>> @@ -1565,6 +1605,10 @@ static int map_create(union bpf_attr *attr, bool kernel)
>>  	bpf_map_free(map);
>>  put_token:
>>  	bpf_token_put(token);
>> +	if (err && log)
>> +		(void) bpf_vlog_finalize(log, &log_true_size);
>> +	if (log)
>> +		kvfree(log);
>>  	return err;
>>  }
>
> +1 to Andrii's suggestion to report log size back,
> just for consistency reasons.

Right.

The log_true_size will be reported back to users.

Thanks,
Leon