On 07/29/2018 09:59 AM, Josh Poimboeuf wrote:

On Fri, Jul 27, 2018 at 10:43:02PM +0000, Jeremy Cline wrote:

quoted

'family' can be a user-controlled value, so sanitize it after the bounds
check to avoid speculative out-of-bounds access.

Cc: Josh Poimboeuf <redacted>
Cc: stable@vger.kernel.org
Signed-off-by: Jeremy Cline <redacted>
---
 net/socket.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/net/socket.c b/net/socket.c
index f15d5cbb3ba4..608e29ae6baf 100644
--- a/net/socket.c
+++ b/net/socket.c

@@ -2672,7 +2672,8 @@ EXPORT_SYMBOL(sock_unregister);
 
 bool sock_is_registered(int family)
 {
-	return family < NPROTO && rcu_access_pointer(net_families[family]);
+	return family < NPROTO &&
+		rcu_access_pointer(net_families[array_index_nospec(family, NPROTO)]);
 }
 
 static int __init sock_init(void)

This is another one where I think it would be better to do the nospec
clamp higher up the call chain.  The untrusted 'family' value comes from
__sock_diag_cmd():

__sock_diag_cmd
  sock_load_diag_module
    sock_is_registered

That function has a bounds check, and also uses the value in some other
array accesses:

	if (req->sdiag_family >= AF_MAX)
		return -EINVAL;

	if (sock_diag_handlers[req->sdiag_family] == NULL)
		sock_load_diag_module(req->sdiag_family, 0);

	mutex_lock(&sock_diag_table_mutex);
	hndl = sock_diag_handlers[req->sdiag_family];
	...

So I think clamping 'req->sdiag_family' right after the bounds check
would be the way to go.

Indeed, the clamp there would cover this clamp. I had a scheme that I
quickly fix all the gadgets in functions with local comparisons, but
clearly that's going to result in call chains with multiple clamps.

I can fix this in a follow-up with a clamp here, or respin this patch
set, whatever is easier for David.

Thanks for the review!