]> www.infradead.org Git - users/hch/misc.git/commitdiff
net: do not delay dst_entries_add() in dst_release()
authorEric Dumazet <edumazet@google.com>
Tue, 8 Oct 2024 14:31:10 +0000 (14:31 +0000)
committerPaolo Abeni <pabeni@redhat.com>
Thu, 10 Oct 2024 09:28:17 +0000 (11:28 +0200)
dst_entries_add() uses per-cpu data that might be freed at netns
dismantle from ip6_route_net_exit() calling dst_entries_destroy()

Before ip6_route_net_exit() can be called, we release all
the dsts associated with this netns, via calls to dst_release(),
which waits an rcu grace period before calling dst_destroy()

dst_entries_add() use in dst_destroy() is racy, because
dst_entries_destroy() could have been called already.

Decrementing the number of dsts must happen sooner.

Notes:

1) in CONFIG_XFRM case, dst_destroy() can call
   dst_release_immediate(child), this might also cause UAF
   if the child does not have DST_NOCOUNT set.
   IPSEC maintainers might take a look and see how to address this.

2) There is also discussion about removing this count of dst,
   which might happen in future kernels.

Fixes: f88649721268 ("ipv4: fix dst race in sk_dst_get()")
Closes: https://lore.kernel.org/lkml/CANn89iLCCGsP7SFn9HKpvnKu96Td4KD08xf7aGtiYgZnkjaL=w@mail.gmail.com/T/
Reported-by: Naresh Kamboju <naresh.kamboju@linaro.org>
Tested-by: Linux Kernel Functional Testing <lkft@linaro.org>
Tested-by: Naresh Kamboju <naresh.kamboju@linaro.org>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Xin Long <lucien.xin@gmail.com>
Cc: Steffen Klassert <steffen.klassert@secunet.com>
Reviewed-by: Xin Long <lucien.xin@gmail.com>
Link: https://patch.msgid.link/20241008143110.1064899-1-edumazet@google.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
net/core/dst.c

index 95f533844f17f119c09f335ccf9bf09515dd3606..9552a90d4772dce49b5fe94d2f1d8da6979d9908 100644 (file)
@@ -109,9 +109,6 @@ static void dst_destroy(struct dst_entry *dst)
                child = xdst->child;
        }
 #endif
-       if (!(dst->flags & DST_NOCOUNT))
-               dst_entries_add(dst->ops, -1);
-
        if (dst->ops->destroy)
                dst->ops->destroy(dst);
        netdev_put(dst->dev, &dst->dev_tracker);
@@ -159,17 +156,27 @@ void dst_dev_put(struct dst_entry *dst)
 }
 EXPORT_SYMBOL(dst_dev_put);
 
+static void dst_count_dec(struct dst_entry *dst)
+{
+       if (!(dst->flags & DST_NOCOUNT))
+               dst_entries_add(dst->ops, -1);
+}
+
 void dst_release(struct dst_entry *dst)
 {
-       if (dst && rcuref_put(&dst->__rcuref))
+       if (dst && rcuref_put(&dst->__rcuref)) {
+               dst_count_dec(dst);
                call_rcu_hurry(&dst->rcu_head, dst_destroy_rcu);
+       }
 }
 EXPORT_SYMBOL(dst_release);
 
 void dst_release_immediate(struct dst_entry *dst)
 {
-       if (dst && rcuref_put(&dst->__rcuref))
+       if (dst && rcuref_put(&dst->__rcuref)) {
+               dst_count_dec(dst);
                dst_destroy(dst);
+       }
 }
 EXPORT_SYMBOL(dst_release_immediate);