Re: [patch net-next] team: do not use -ENOENT

[David Miller <davem@davemloft.net>]

From: Jiri Pirko <jiri@resnulli.us>
Date: Thu, 17 Jan 2013 21:52:12 +0100

> Thu, Jan 17, 2013 at 09:42:40PM CET, davem@davemloft.net wrote:
>>From: Jiri Pirko <jiri@resnulli.us>
>>Date: Thu, 17 Jan 2013 21:33:47 +0100
>>
>>>>> @@ -2320,7 +2320,7 @@ static int team_nl_cmd_options_set(struct sk_buff *skb, struct genl_info *info)
>>>>>  			list_add(&opt_inst->tmp_list, &opt_inst_list);
>>>>>  		}
>>>>>  		if (!opt_found) {
>>>>> -			err = -ENOENT;
>>>>> +			err = -EINVAL;
>>>>>  			goto team_put;
>>>>>  		}
>>>>>  	}
>>
>>
>>> I'm not really sure. But in this case, I do not think that is a problem.
>>> 
>>> 1) I'm most probably the only one (libteam) who is using this api and
>>> libteam does not mind about what err code is returned in these cases.
>>> 
>>> 2) In this case, it is only about different number. And one number or
>>> another, it does not imply userspace to behave differently. In other words,
>>> userspace should not take different actions in case for example -ENOENT
>>> or -ENODEV is returned.
>>
>>I agree with this analysis.
>>
>>But for the team_nl_cmd_options_set() case, I would strongly advise
>>that you use some error code more descriptive than -EINVAL.  In fact
>>the existing -ENOENT I feel is better, because it tells the caller
>>what kind of problem there was.
>>
>>Even if you don't like the fact that -ENOENT is oriented towards file
>>existence, it does convey a ton more information than -EINVAL does.
> 
> I understand your feeling, because I have the same one :)
> But looking all over the code and on possible err codes as well, I did
> not find any suitable err code to indicate some object was not found.
> And since I recently saw email from Linus about the fact that -ENOENT
> should be used only in relation to files, -EINVAL the "default:" in my
> "switch()".

Look in the packet scheduler API for how much we use -ENOENT in this
kind of situation where the requested object to operate on could not
be found.

I think it is entirely appropriate to use -ENOENT, if not completely
consistent with the rest of the networking.