Messages in this thread | | | From | "Khimenko Victor" <> | Date | Thu, 30 Apr 1998 23:17:04 +0400 (MSD) | Subject | Re: Varlinks - how not to break semantics |
| |
RW> Hi there. RW> RW> >>>> a) Link is processed as varlink if and only if it beggins with ///. RW> >>>> [normally, you have no reason to prepend ///, so it should be safe] RW> RW> >>> Happens easily if you append paths, and violates POSIX (the only RW> >>> permitted escapes from the namespace are // and /../). RW> RW> >> Ook, so go for //, or go for /proc/varlink. /proc/varlink is pretty RW> >> long to type, but we already have more strange things in /proc RW> >> filesystem :-). RW> RW> > Looks like /proc/varlink prefix is preferrable. This way you'll not break RW> > anything for already existing links (links to now not existing /proc/varlink RW> > will be affected but this could break only program specially written to be RW> > broken with varlinks :-) Of course /proc/varlink is long to type, but this RW> > is better than other suggestions... RW> RW> If the length is a problem, I'm sure a quick... RW> RW> Q> ln -s /proc/varlink /vl RW> RW> ...would soon cure that... RW> I am not know about how this is could be done, but in simplest possible realisation this will not work. The idea is that symlink must start from exactly following 14 bytes: 2F 70 72 6F 63 2F 76 61 72 6C 69 6E 6B 2F and this prefix will be stripped out before varlink expansion... With this approach you could not find directory /varlink in /proc but /proc/varlink//.${uid}/tmp will be expanded to /.500/tmp, /.501/tmp, etc...
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu
| |