Re: does numeric only name of jails not allowed (bug?)

On Sat, 9 Aug 2025 19:10:56 +0300
Anthony Pankov <anthony.pankov@yahoo.com> wrote:

> Hello, Kyle.
> 
> >> I think to prevent ambiguity there must be a possibility or requirement
> >> to quote jail name to force interpretation as a string. Using "47777"
> >> instead of 47777 doesn't help for now. 
> 
> > What ambiguity are you referring to here?  In context it's pretty 
> > unambiguous, the name is the thing that appears before an opening brace.  
> 
> For  me 47777 is a hexadecimal number (there is also a32b1 which works
> well).   I  want  it  to be the name of a jail (processed as a string)
> without any relation to jid.

Well, it certainly isn't - it may be in the same way "octal", then. To be
unambiguously hexadecimal it should be 0x47777 of 47777h.

> But 47777 may  be recognized as a decimal number also and, consequently,
> will be used as a jid and a name.

UID and GID are also decimal, yes.

> On 9 августа 2025 г., 16:30:32 you wrote:
> 
> > On 8/9/25 06:01, Anthony Pankov wrote:  
> >> Hello Kyle,
> >>   
> >>> This is an artifact of how numeric names work.  When you use a purely
> >>> numeric name, it's taken as both the jid and the name.  jail(8) has some
> >>> logic up-front that will set one or the other variable based on the
> >>> name, but it omits the other.  
> >> 
> >> I think to prevent ambiguity there must be a possibility or requirement
> >> to quote jail name to force interpretation as a string. Using "47777"
> >> instead of 47777 doesn't help for now. 
> 
> > What ambiguity are you referring to here?  In context it's pretty 
> > unambiguous, the name is the thing that appears before an opening brace.  
> 
> >>> When you use a purely numeric name, it's taken as both the jid  
> >> I've never hear about this feature. But I'm in doubt how to guarantee
> >> jail startup with jid=name=10 along system functioning. If non-numeric
> >> jail occupied jid 10 then jail 10  will not start?  
> 
> > This is not new, jails have operated this way for 15+ years (since 
> > numeric names were first allowed).  Indeed, if it's already occupied 
> > then the jail already exists and it will not start.  
> 
> >> However, thinking of jid uniqueness in a wider area than localhost
> >> definitely leads us to use a bigger numbers. In such a case using
> >> hexadecimal format seems to be beneficial. Then there will be some code
> >> to interpret numeric names in decimal/hexadecimal format. So there will
> >> be some not so small branch of code  for numeric jail name. And, again,
> >> to split  this branch from pure string names I suggest to use name
> >> quotation to enforce string interpretation. 
> 
> >>> I think to prevent ambiguity there must be a possibility or requirement
> >>> to quote jail name to force interpretation as a string. Using "47777"
> >>> instead of 47777 doesn't help for now. 
> >  
> >> What ambiguity are you referring to here?  In context it's pretty 
> >> unambiguous, the name is the thing that appears before an opening brace.  
>  hex does not
> > get resolved to base 10, and you end up with a jail that has a hex name
> > and a jid allocated in the usual way.  I'm not sure what else you're 
> > writing about here, because it's really not that complicated: if it 
> > coerces to a number in base 10 it's the jid, otherwise it's the name. 
> > Jails without a name use the jid as their name.  
> 
> > This review will fix the bug I noted in my previous e-mail:
> > https://reviews.freebsd.org/D51831  
> 
> > Thanks,  
> 
> > Kyle Evans  
> 
> >> 
> >> Friday, August 8, 2025, 8:24:32 PM, you wrote:
> >>   
> >>> On 8/6/25 09:37, Anthony Pankov wrote:  
> >>>> Dear freebsd-hackers.  
> >>>>> It seems there is no sign that numeric only name is prohibited:
> >>>>> "  
> >>>> The jail name.  This is an arbitrary string that identifies a
> >>>>                jail (except it may not contain a ďż˝.ďż˝).
> >>>> "  
> >>>>> But  
> >>>> # head jail-47777.conf
> >>>> ...
> >>>> 47777 {
> >>>>           host.hostname = "${name}";
> >>>> ...  
> >>>>> # jail -c 47777  
> >>>> jail: 47777: host.hostname: variable "name" not found  
> >>>>> If I add letter prefix, say w47777, it worked.
> >>>>> Because of strange error "variable "name" not found" it seems like a
> >>>>> bug. Does it?  
> >>>>  
> >>> This is an artifact of how numeric names work.  When you use a purely
> >>> numeric name, it's taken as both the jid and the name.  jail(8) has some
> >>> logic up-front that will set one or the other variable based on the
> >>> name, but it omits the other.  
> >>   
> >>> I have this almost fixed in 15.0, but trying your example I've found a
> >>> bug in it -- setting the name, we need to use the string_param() of the
> >>> KP_JID param as the `value`; the current use of KP_JID's cfparam for `p`
> >>> to add_param() will adopt the name of KP_JID, which is a peculiarity
> >>> that I wasn't aware of.  
> >>   
> >>> Thanks,  
> >>   
> >>> Kyle Evans  
\

-- 
WBR, @nuclight