On Wed, Nov 5, 2014 at 9:21 AM, David Drysdale <drysdale@xxxxxxxxxx> wrote: > On Tue, Nov 4, 2014 at 9:40 AM, David Drysdale <drysdale@xxxxxxxxxx> wrote: >> On Mon, Nov 3, 2014 at 5:22 PM, Eric W.Biederman <ebiederm@xxxxxxxxxxxx> wrote: >>> On November 3, 2014 7:42:58 AM PST, Andy Lutomirski <luto@xxxxxxxxxxxxxx> wrote: >>>>On Mon, Nov 3, 2014 at 7:20 AM, Al Viro <viro@xxxxxxxxxxxxxxxxxx> >>>>wrote: >>>>> On Mon, Nov 03, 2014 at 11:48:23AM +0000, David Drysdale wrote: >>>>>> Add a new O_BENEATH flag for openat(2) which restricts the >>>>>> provided path, rejecting (with -EACCES) paths that are not beneath >>>>>> the provided dfd. In particular, reject: >>>>>> - paths that contain .. components >>>>>> - paths that begin with / >>>>>> - symlinks that have paths as above. >>>>> >>>>> Yecch... The degree of usefulness aside (and I'm not convinced that >>>>it >>>>> is non-zero), >>>> >>>>This is extremely useful in conjunction with seccomp. >>>> >>>>> WTF pass one bit out of nameidata->flags in a separate argument? >>>>> Through the mutual recursion, no less... And then you are not even >>>>attempting >>>>> to detect symlinks that are not followed by interpretation of _any_ >>>>pathname. >>>> >>>>How many symlinks like that are there? Is there anything except >>>>nd_jump_link users? All of those are in /proc. Arguably O_BENEATH >>>>should prevent traversal of all of those links. >>> >>> Not commenting on the sanity of this one way or another, and I haven't read the patch. There is an absolutely trivial implementation of this. >>> >>> After the path is resolved, walk backwards along d_parent and the mount tree, and see if you come to the file or directory dfd refers to. >>> >>> That can handle magic proc symlinks, and does not need to disallow .. or / explicitly so it should be much simpler code. >>> >>> My gut says that if Al says blech when looking at your code it is too complex to give you a security guarantee. >>> >>> Eric >> >> Well, the 'yecch' was deserved for the unnecessary duplication of the >> flags. Without that, the patch looks much simpler -- I'll send out a v2 >> with those changes for discussion, and think about your alternative >> implementation suggestion (thanks!) separately. > > One concern with the "walk upwards and see if you get back where you > started" approach -- it will allow use of a symlink that lives outside the > original directory, but which points back inside it. That's going to be > slightly surprising behaviour for users, and I worry that there's the > potential for unexpected information leakage from it. > > (BTW, size-wise my initial naive implementation of the walk-upward > approach is only marginally smaller than the v2 patch.) It has another problem. Since we still haven't fixed the eternal /proc/PID/fd-doesn't-respect-file-mode issue, you can have a read-only fd somewhere and reopen it read-write using O_BENEATH on a different fd. For example: fd 3 points to /sandbox /sandbox/blocked has mode 0700 and isn't owned by us /sandbox/blocked/foo has mode 0666 fd 4 points to /sandbox/blocked/foo, O_RDONLY openat(3, "/proc/self/fd/4", O_RDWR | O_BENEATH) will get a read-write descriptor pointing at /sandbox/blocked/foo, which should have been impossible. Also, I really don't like the information leak. The result of asking a server for "/home/victim/compromising-directory/../../www/index.html" should not reveal whether compromising-directory exists. --Andy -- To unsubscribe from this list: send the line "unsubscribe linux-api" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
