Leak in selinux stuff?

Leak in selinux stuff?

[Po Lu]

fileio.c is littered with code like such:

          /* See https://debbugs.gnu.org/11245 for ENOTSUP.  */
	  if (fail && errno != ENOTSUP)
	    report_file_error ("Doing lsetfilecon", absname);

	  context_free (parsed_con);
	  freecon (con);

where `parsed_con' and `con' are allocated by libselinux functions by
the likes of `lgetfilecon'.  Around such code, no measures are
apparently taken for deallocating those objects should a non local exit
occur.  Am I missing anything, or should this be fixed?

Re: Leak in selinux stuff?

[Eli Zaretskii]

> From: Po Lu <luangruo@yahoo.com>
> Date: Sat, 19 Aug 2023 11:37:06 +0800
> 
> fileio.c is littered with code like such:
> 
>           /* See https://debbugs.gnu.org/11245 for ENOTSUP.  */
> 	  if (fail && errno != ENOTSUP)
> 	    report_file_error ("Doing lsetfilecon", absname);
> 
> 	  context_free (parsed_con);
> 	  freecon (con);
> 
> where `parsed_con' and `con' are allocated by libselinux functions by
> the likes of `lgetfilecon'.  Around such code, no measures are
> apparently taken for deallocating those objects should a non local exit
> occur.  Am I missing anything, or should this be fixed?

IMO, they should either free the resources before signaling an error,
or use record_unwind_protect to free the allocated resources, yes.