I have the following situation, I have a running GNU screen session where I can't access AFS anymore - my token has expired. I can however access it from a new shell. The difference to this question is that I don't have a Kerberos ticket (well, not for the realm aklog is looking for), so I can't call aklog. I also can't get such a ticket. I have no idea how AFS is set up, but it works.
Now, Kerberos tickets are "stored" in /tmp/krb5cc*, and pointed to by a variable called KRB5CCNAME. If I have this problem with Kerberos and screen/tmux, I can either do kinit, or transplant the newer ticket to the old shell by setting KRB5CCNAME.
I wonder how AFS credentials are pointed to, and if I can similarly transplant them from the outer shell (the one I ssh into, which has AFS access) to the inner shell (the one I get after screen -r, which has no more AFS access). There seems to be no relevant environment variable changed between both shells. strace tokens tells me that it just accesses /proc/fs/openafs/afs_ioctl, which suggests it is tied to the process and using a special kernel feature, which would make it pretty hard. Any ideas how I can get AFS access back in my shell without closing it and opening a new one?