Red Hat Bugzilla – Bug 5314
/usr/libexec/pt_chown should be suid root
Last modified: 2008-05-01 11:37:51 EDT
/usr/libexec/pt_chown should be suid root and it isn't, this causes mc's subshell and gnome-terminal to fail when run by non root. Probably also causes other problems. This is with the latest glibc from the snapshots dir: glibc-2.1.2-11
Can't reproduce this here; glibc-2.1.2-11, latest gnome-core & mc.
Hmm, could you do a "ls -l /usr/libexec/pt_chown" and check if it is suid root? It should be, but in the rpm it isn't. Did you test this as an ordinary user? As root it works fine. If it is suid root, could you also do "rpm --verify glibc" ? I made it suid root myself and rpm --verify glibc complains that the mode is not ok. "rpm -ql --dump glibc|grep pt_chown" gives: /usr/libexec/pt_chown 34643 937818817 3f730db73c6f4ff5d4f20b37a7fe7976 0100755 root root 0 0 19171 X Which should be: /usr/libexec/pt_chown 34643 937818817 3f730db73c6f4ff5d4f20b37a7fe7976 0104755 root root 0 0 19171 X rpm -q glibc gives: glibc-2.1.2-11
Just read you mail that it isn't suid root for you, you could have mentioned that over here too ;) Okay, you're partially right mc seems to have been fixed to not use pt_chown anymore. But gnome-terminal still gives: [hans@shalem hans]$ gnome-terminal pt_chown: needs to be installed setuid `root' Error: unable to fork: Success [hans@shalem hans]$
correction it is a tricky bug: Change /usr/libexec/pt_chown to 755 If I open one gnome-terminal all is well second all is well, third fails What happens is that pt_chown is only called for pty's which haven't been used by the owner before so If I first open a few gnome-terminals with /usr/libexec/pt_chown suid root. And then try it without /usr/libexec/pt_chown suid root then I can open as many terminals as I already had open, it goes wrong as soon as I open as many as I had open + 1. If I'm as far that gnome-terminal starts failing so does mc's subshell.
Do you have the /dev/pts filesystem mounted?
oops, I didn't, with it it works fine sorry for the hassle.
That's OK.
After some further checking I found out that the problem is devfs related, this bug does happen on devfs kernel, since devfs doesn't support the pts filesystem, so maybe this should still be fixed, maybe ?