The kernel may crash when deleting a genetlink family if there are still
listeners for that family:

Oops: Kernel access of bad area, sig: 11 [#1]
  ...
  NIP [c000000000c080bc] netlink_update_socket_mc+0x3c/0xc0
  LR [c000000000c0f764] __netlink_clear_multicast_users+0x74/0xc0
  Call Trace:
__netlink_clear_multicast_users+0x74/0xc0
genl_unregister_family+0xd4/0x2d0

Change the unsafe loop on the list to a safe one, because inside the
loop there is an element removal from this list.

Fixes: b8273570f802 ("genetlink: fix netns vs. netlink table locking (2)")\
Cc: stable@vger.kernel.org
Signed-off-by: Anastasia Kovaleva <a.kovaleva@yadro.com>
Reviewed-by: Dmitry Bogdanov <d.bogdanov@yadro.com>
---
 include/net/sock.h       | 2 ++
 net/netlink/af_netlink.c | 3 ++-
 2 files changed, 4 insertions(+), 1 deletion(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index c58ca8dd561b..eec77a18602a 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -894,6 +894,8 @@ static inline void sk_add_bind_node(struct sock *sk,
 	hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
 #define sk_for_each_bound(__sk, list) \
 	hlist_for_each_entry(__sk, list, sk_bind_node)
+#define sk_for_each_bound_safe(__sk, tmp, list) \
+        hlist_for_each_entry_safe(__sk, tmp, list, sk_bind_node)
 
 /**
  * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset
diff --git a/net/netlink/af_netlink.c b/net/netlink/af_netlink.c
index 0b7a89db3ab7..0a9287fadb47 100644
--- a/net/netlink/af_netlink.c
+++ b/net/netlink/af_netlink.c
@@ -2136,8 +2136,9 @@ void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group)
 {
 	struct sock *sk;
 	struct netlink_table *tbl = &nl_table[ksk->sk_protocol];
+	struct hlist_node *tmp;
 
-	sk_for_each_bound(sk, &tbl->mc_list)
+	sk_for_each_bound_safe(sk, tmp, &tbl->mc_list)
 		netlink_update_socket_mc(nlk_sk(sk), group, 0);
 }
 
-- 
2.40.1

On Tue, 1 Oct 2024 14:58:28 +0300 Anastasia Kovaleva wrote:
> The kernel may crash when deleting a genetlink family if there are still
> listeners for that family:

Could you add a selftest? Should be fairly easy using YNL, ncdevmem is
the only user so far.

> Oops: Kernel access of bad area, sig: 11 [#1]
>   ...
>   NIP [c000000000c080bc] netlink_update_socket_mc+0x3c/0xc0
>   LR [c000000000c0f764] __netlink_clear_multicast_users+0x74/0xc0
>   Call Trace:
> __netlink_clear_multicast_users+0x74/0xc0
> genl_unregister_family+0xd4/0x2d0
> 
> Change the unsafe loop on the list to a safe one, because inside the
> loop there is an element removal from this list.
> 
> Fixes: b8273570f802 ("genetlink: fix netns vs. netlink table locking (2)")\

nit: trailing \ at the end of the line here

On Wed, 2 Oct 2024 06:02:40 -0700 Jakub Kicinski wrote:
> Could you add a selftest? Should be fairly easy using YNL, ncdevmem is
> the only user so far.

Thank you for your review.

Originally I've stumbled upon this bug while developing a functionality in the
iscsit driver to inform the userspace about a failed iscsi session
authentication. I am not familiar with the netdev test tools; with YNL and
ncdevmem particularly. The change seems trivial enough. It fixes an obvious bug
with the deletion of an element of a list that you are currently iterating
over. I assume that it wasn’t reproduced often because there are few users of
the genetlink and usually there is no listeners of a family while removing a
driver that uses it.

If it is really necessary, it would take me a couple of days to make a test. If
not, I can send a v3 patch that just removes a trailing \.

On Wed, 2 Oct 2024 17:53:18 +0000 Anastasia Kovaleva wrote:
> If it is really necessary, it would take me a couple of days to make a test. 

The only tricky part, I think, is finding a family which can be easily
unloaded for the test. I don't see any good candidates, so fine.