[Devel] [PATCH 1/6] host uid and gid parameters

[Glauber Costa]

>> @@ -1996,6 +2006,24 @@ static int parse(envid_t veid, vps_param
>> *vps_p, char *val, int id)
>>       case PARAM_IPTABLES:
>>           ret = parse_iptables(&vps_p->res.env, val);
>>           break;
>> +
>> +    case PARAM_LOCAL_UID:
>> +        ret = parse_ul(val, &vps_p->res.misc.local_uid);
> 
> 1 Are we OK with the fact that the biggest number returned from
> parse_ul() is LONG_MAX and not ULONG_MAX?
> 

Humm, LONG_MAX is different in 32 bits, right? So yes, it might be a
problem =(

I was basically in a 64-bit mindset. We don't care, because the kernel
type is actually half of that.

What's the best tool for this job, then ?

> 2 We use conf_parse_ulong() and pointer to unsigned long in similar
> places in order to detect situations such as unset value or the value
> set twice in config. If you don't want that, I guess it's fine to use
> parse_ul() directly...
> 

I have no particular preference, parse_ul seemed to already do all I
wanted...

>> +        if (ret != 0)
>> +            break;
>> +
>> +        if (vps_p->res.misc.local_uid == 0)
>> +            vps_p->res.misc.local_uid = VZ_DEFAULT_UID;
> 
> So here you deliberately disable possibility to use 0 as an offset,
> making 0 mean "use default". RIght?
> 
Note that those values are ignored when user namespaces are not available.

When they are, I don't really think running containers as root should be
even allowed. So yes, this is by design.