On Thu, Jul 12, 2007 at 04:49:35PM +0200, Stefan Reinauer wrote: > > > + mkdirp((const char *) dirname(path), 0755); > > > > I don't think lar should mkdir -p implicitly. > > Oh it does not do this implicitly. It only does it if the archive > explicitly contains a directory.
It needs to be there, but I think the sanity checking could be good. On Thu, Jul 12, 2007 at 04:52:12PM +0200, Stefan Reinauer wrote: > > But I think a bit of sanity would be nice here since a lar could > > otherwise be used to overwrite arbitrary system files. > > You compile LinuxBIOS as root? No, but I may run lar as root because I need to tweak something right before running flashrom as root. Yes, bad practice. No, I'm not the only one. > > I'll make a patch for mkdirp() that ensures the directory to be > > created is actually below the current directory > > Rather check the path before mkdir()ing. Exactly the idea. > I am pretty sure the mkdir efforts can easily be tricked by a > couple of symlinks in the path, realpath() handles that. To do full path resolution is not so simple though, so I cut some corners. It may be too much effort to be at all worthwhile and we'll instead let root shoot foot. > so I wonder how much use there is in trying to make this "secure" > since it never runs as root anyways, and in a very controlled > environment. LB build mostly yes, lar not so sure. //Peter -- linuxbios mailing list [email protected] http://www.linuxbios.org/mailman/listinfo/linuxbios
