This is the mail archive of the
mailing list for the glibc project.
RE: Runtime change to LD_LIBRARY_PATH
- To: "Ed Connell" <Ed dot Connell at sas dot com>
- Subject: RE: Runtime change to LD_LIBRARY_PATH
- From: "Ken Whaley" <ken at believe dot com>
- Date: Fri, 17 Aug 2001 13:01:20 -0700
- Cc: <libc-alpha at sources dot redhat dot com>
Sounds like a potential security problem if a program is able to
change out the LD_LIBRARY_PATH in mid-execution to affect future
loads without the user's knowledge.
> -----Original Message-----
> From: firstname.lastname@example.org
> [mailto:email@example.com]On Behalf Of Ed Connell
> Sent: Friday, August 17, 2001 11:32 AM
> To: 'firstname.lastname@example.org'
> Subject: Runtime change to LD_LIBRARY_PATH
> I noticed that if you set LD_LIBRARY_PATH after your program
> is running, ld.so does not take that into account when trying
> to resolve future dlopen's. (_dl_init_paths() sets
> env_path_list while ld.so is initializing and it is left
> alone after that.) It would be easy to check LD_LIBRARY_PATH
> for changes each time you want to map a new object. Would
> that be rational change to make? I'll code it up if people
> like the idea.