linux-kernel - BUG: User triggerable kernel panic in 4.8 (possibly 4.9)

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <Pine.LNX.4.64.1610201520290.25450@oraqre.bcra-fbhepr.pb.hx>
Date:   Thu, 20 Oct 2016 15:24:37 +0100
From:   Darren Austin <lists@...erdark.org.uk>
To:     linux-kernel@...r.kernel.org
Subject: BUG: User triggerable kernel panic in 4.8 (possibly 4.9)

[ Please CC me with any replies as I will be leaving the list shortly ]

Hi,
  I'm not sure if this is the best place to report an issue i've discovered 
with the kernel and the 'fsc' mount option - please let me know if there is 
some other mailing list or person I should be notifying about this.

The bug appears (at least for me) when using an NFS server, and a client 
which mounts an export from that server with the 'fsc' option (whether or 
not the fscache daemon is running or not).  It also seems easiest to trigger 
using the 'nano' editor, but other commands will trigger it randomly.

I've tested this bug on the Ubuntu 1610 kernel (4.8.0) and with the 4.8.2 
kernel from http://kernel.ubuntu.com/~kernel-ppa/mainline/.  The latter 
purports to be a kernel built from the unmodified kernel source, and simply 
packaged in a .deb, so this isn't a Ubuntu kernel problem.  Unless the 
problem has been corrected in the 4.9 series, I suspect the bug may also 
persist there - i've not had the opportunity to check 4.9.x kernels as yet.

I can repeatidly reproduce this bug on my system, so it's definitely not a 
one off - it causes a kernel panic and complete lock-up every time.

The NFS share I tested with is exported with options: 
  rw,async,insecure,insecure_locks,no_root_squash,anongid=99,anonuid=99,no_subtree_check
(but the export options don't seem to matter to trigger the  bug)
and mounted on the client with options:
  vers=4,hard,intr,acl,rw,fsc
via the Linux automounter (but the issue persists on mounts from fstab or 
when mounted manually).

To reproduce the bug is quite simple...
1) Set up the server export and client mount as detailed above.
2) From the console (or in a terminal; but I only tested this once) in the 
directory where you've mounted the NFS share, run:
  nano testfile.txt
and add write some text to the file.
2) Save the file and exit (Ctl+X is how I did it).
3) When back at the prompt, immediately hit the up arrow on the keyboard (to
load the last typed command into the buffer) and hit enter.
4) Watch as the pretty text of the panic scrolls by :)

With the help of people on the nano-dev mailing list, I figured out that 
it's the 'fsc' option which causes the panic - repeated tests without that 
option active do not trigger it.  However, this is /not/ a nano specific bug 
- it can be triggered by any command used on the mount.  And besides, nano 
shouldn't be able to take down the system :)

If any further information is required, please don't hesitate to reply.

Darren.