bpf: fix double-fdput in replace_map_fd_with_map_ptr()
authorJann Horn <jannh@google.com>
Tue, 26 Apr 2016 20:26:26 +0000 (22:26 +0200)
committerDavid S. Miller <davem@davemloft.net>
Tue, 26 Apr 2016 21:37:21 +0000 (17:37 -0400)
When bpf(BPF_PROG_LOAD, ...) was invoked with a BPF program whose bytecode
references a non-map file descriptor as a map file descriptor, the error
handling code called fdput() twice instead of once (in __bpf_map_get() and
in replace_map_fd_with_map_ptr()). If the file descriptor table of the
current task is shared, this causes f_count to be decremented too much,
allowing the struct file to be freed while it is still in use
(use-after-free). This can be exploited to gain root privileges by an
unprivileged user.

This bug was introduced in
commit 0246e64d9a5f ("bpf: handle pseudo BPF_LD_IMM64 insn"), but is only
exploitable since
commit 1be7f75d1668 ("bpf: enable non-root eBPF programs") because
previously, CAP_SYS_ADMIN was required to reach the vulnerable code.

(posted publicly according to request by maintainer)

Signed-off-by: Jann Horn <jannh@google.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
kernel/bpf/verifier.c

index 618ef77c302ab628cd1a2c517568d4e0a10aa93e..db2574e7b8b008395bb8fee53baf513fdc56c392 100644 (file)
@@ -2030,7 +2030,6 @@ static int replace_map_fd_with_map_ptr(struct verifier_env *env)
                        if (IS_ERR(map)) {
                                verbose("fd %d is not pointing to valid bpf_map\n",
                                        insn->imm);
-                               fdput(f);
                                return PTR_ERR(map);
                        }