genetlink: fix family dump race

commit 58ad436fcf49810aa006016107f494c9ac9013db upstream.

When dumping generic netlink families, only the first dump call
is locked with genl_lock(), which protects the list of families,
and thus subsequent calls can access the data without locking,
racing against family addition/removal. This can cause a crash.
Fix it - the locking needs to be conditional because the first
time around it's already locked.

A similar bug was reported to me on an old kernel (3.4.47) but
the exact scenario that happened there is no longer possible,
on those kernels the first round wasn't locked either. Looking
at the current code I found the race described above, which had
also existed on the old kernel.

Reported-by: Andrei Otcheretianski <andrei.otcheretianski@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

1 files changed, 7 insertions, 0 deletions

diff --git a/net/netlink/genetlink.c b/net/netlink/genetlink.c
index 1076fe16b122..ba6e55d1ca44 100644
--- a/net/netlink/genetlink.c
+++ b/net/netlink/genetlink.c
@@ -789,6 +789,10 @@ static int ctrl_dumpfamily(struct sk_buff *skb, struct netlink_callback *cb)
         struct net *net = sock_net(skb->sk);
         int chains_to_skip = cb->args[0];
         int fams_to_skip = cb->args[1];
+        bool need_locking = chains_to_skip || fams_to_skip;
+
+        if (need_locking)
+                genl_lock();
 
         for (i = chains_to_skip; i < GENL_FAM_TAB_SIZE; i++) {
                 n = 0;
@@ -810,6 +814,9 @@ errout:
         cb->args[0] = i;
         cb->args[1] = n;
 
+        if (need_locking)
+                genl_unlock();
+
         return skb->len;
 }