Re: [PATCH] pasta.c: modify hostname when detaching new namespace

[Stefano Brivio <sbrivio@redhat.com>]

On Mon, 29 Jul 2024 19:26:28 +0530
Danish Prakash <contact@danishpraka.sh> wrote:

> Hi Stefano,
> 
> I was caught up with some work last month, hence the delay in responses
> since my last email.
> 
> On 5/24/24 11:09 PM, Stefano Brivio wrote:
> >> +		if (sethostname(hostname, strlen(hostname)))  
> > 
> > So, I mentioned before that you don't really need to set a NULL
> > terminating byte for sethostname() itself, because it takes a length.
> > 
> > But strlen() needs it. If gethostname() truncated the hostname,
> > according to POSIX, it's unspecified whether we'll have a NULL byte at
> > the end of 'hostname', and strlen() would read out-of-bounds, past the
> > end of 'hostname'.
> > 
> > That's not an issue with glibc, but if POSIX says it's not guaranteed,
> > we shouldn't take anything for granted.
> > 
> > I would suggest that you simply add a NULL byte at HOST_NAME_MAX,
> > unconditionally, that should cover the normal case as well as the
> > ENAMETOOLONG case. I haven't tested this by the way.  
> 
> Did you mean explicitly setting the NULL byte to `hostname`?
> 
>         hostname[HOST_NAME_MAX] = '\0';

Yes, exactly.

> Doing that after gethostname() and before sethostname() yields the
> desired result. I tested a few cases, seems to be fine.

Yes, I don't expect any difference with glibc, but better safe than
sorry (with other C libraries). Thanks.

-- 
Stefano