linux-kernel - Re: [PATCH] netlink: fix potential race issue in netlink_native_seq

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date: Wed, 17 Jan 2024 11:28:24 +0100
From: Eric Dumazet <edumazet@...gle.com>
To: nai lin <ayano2023th@...il.com>
Cc: netdev@...r.kernel.org, "David S. Miller" <davem@...emloft.net>, 
	Jakub Kicinski <kuba@...nel.org>, Paolo Abeni <pabeni@...hat.com>, Simon Horman <horms@...nel.org>, 
	"Liam R. Howlett" <Liam.Howlett@...cle.com>, Anjali Kulkarni <anjali.k.kulkarni@...cle.com>, 
	Li RongQing <lirongqing@...du.com>, David Howells <dhowells@...hat.com>, 
	Kuniyuki Iwashima <kuniyu@...zon.com>, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] netlink: fix potential race issue in netlink_native_seq_show()

On Wed, Jan 17, 2024 at 9:38 AM nai lin <ayano2023th@...il.com> wrote:
>
> Access to the nlk group should be protected by netlink_lock_table() like
> commit <f773608026ee> ("netlink: access nlk groups safely in netlink bind
> and getname"), otherwise there will be potential race conditions.
>
> Signed-off-by: nai lin <ayano2023th@...il.com>

OK, I think you forgot to include this tag I suggested earlier to you.

Fixes: e341694e3eb5 ("netlink: Convert netlink_lookup() to use RCU
protected hash table")

> ---
>  net/netlink/af_netlink.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/net/netlink/af_netlink.c b/net/netlink/af_netlink.c
> index 4ed8ffd58ff3..61ad81fb80f5 100644
> --- a/net/netlink/af_netlink.c
> +++ b/net/netlink/af_netlink.c
> @@ -2693,6 +2693,7 @@ static int netlink_native_seq_show(struct seq_file *seq, void *v)
>                 struct sock *s = v;
>                 struct netlink_sock *nlk = nlk_sk(s);
>
> +               netlink_lock_table();

netlink_lock_table() is heavy weight, appropriate for contexts where
we might sleep.

We could instead use a helper to acquire the lock for a very small period.

diff --git a/net/netlink/af_netlink.c b/net/netlink/af_netlink.c
index 4ed8ffd58ff375f3fa9f262e6f3b4d1a1aaf2731..c50ca0f5adfb9691e6df37b4ac518b95c2d7908f
100644
--- a/net/netlink/af_netlink.c
+++ b/net/netlink/af_netlink.c
@@ -1136,6 +1136,18 @@ static int netlink_connect(struct socket *sock,
struct sockaddr *addr,
        return err;
 }

+static u32 netlink_get_groups_mask(const struct netlink_sock *nlk)
+{
+       unsigned long flags;
+       u32 res;
+
+       read_lock_irqsave(&nl_table_lock, flags);
+       res = nlk->groups ? nlk->groups[0] : 0;
+       read_unlock_irqrestore(&nl_table_lock, flags);
+
+       return res;
+}
+
 static int netlink_getname(struct socket *sock, struct sockaddr *addr,
                           int peer)
 {
@@ -1153,9 +1165,7 @@ static int netlink_getname(struct socket *sock,
struct sockaddr *addr,
        } else {
                /* Paired with WRITE_ONCE() in netlink_insert() */
                nladdr->nl_pid = READ_ONCE(nlk->portid);
-               netlink_lock_table();
-               nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0;
-               netlink_unlock_table();
+               nladdr->nl_groups = netlink_get_groups_mask(nlk);
        }
        return sizeof(*nladdr);
 }
@@ -2697,7 +2707,7 @@ static int netlink_native_seq_show(struct
seq_file *seq, void *v)
                           s,
                           s->sk_protocol,
                           nlk->portid,
-                          nlk->groups ? (u32)nlk->groups[0] : 0,
+                          netlink_get_groups_mask(nlk),
                           sk_rmem_alloc_get(s),
                           sk_wmem_alloc_get(s),
                           READ_ONCE(nlk->cb_running),