The freeing and re-allocation of algorithm are protected by cpool_mutex,
so it doesn't fix an actual use-after-free, but avoids a deserved
refcount_warn_saturate() warning.

A trivial fix for the racy behavior.

Fixes: 8c73b26315aa ("net/tcp: Prepare tcp_md5sig_pool for TCP-AO")
Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Dmitry Safonov <dima@arista.com>
---
 net/ipv4/tcp_sigpool.c | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/net/ipv4/tcp_sigpool.c b/net/ipv4/tcp_sigpool.c
index 55b310a722c7..8512cb09ebc0 100644
--- a/net/ipv4/tcp_sigpool.c
+++ b/net/ipv4/tcp_sigpool.c
@@ -162,9 +162,8 @@ int tcp_sigpool_alloc_ahash(const char *alg, size_t scratch_size)
 		if (strcmp(cpool[i].alg, alg))
 			continue;
 
-		if (kref_read(&cpool[i].kref) > 0)
-			kref_get(&cpool[i].kref);
-		else
+		/* pairs with tcp_sigpool_release() */
+		if (!kref_get_unless_zero(&cpool[i].kref))
 			kref_init(&cpool[i].kref);
 		ret = i;
 		goto out;

---
base-commit: 1a44b0073b9235521280e19d963b6dfef7888f18
change-id: 20231222-tcp-ao-kref_get_unless_zero-fe7105781ba4

Best regards,
-- 
Dmitry Safonov <dima@arista.com>

On Fri, Dec 22, 2023 at 2:14 AM Dmitry Safonov <dima@arista.com> wrote:
>
> The freeing and re-allocation of algorithm are protected by cpool_mutex,
> so it doesn't fix an actual use-after-free, but avoids a deserved
> refcount_warn_saturate() warning.
>
> A trivial fix for the racy behavior.
>
> Fixes: 8c73b26315aa ("net/tcp: Prepare tcp_md5sig_pool for TCP-AO")
> Suggested-by: Eric Dumazet <edumazet@google.com>
> Signed-off-by: Dmitry Safonov <dima@arista.com>

Reported-by: syzbot <syzkaller@googlegroups.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>

On Fri, Dec 22, 2023 at 01:13:59AM +0000, Dmitry Safonov wrote:
> The freeing and re-allocation of algorithm are protected by cpool_mutex,
> so it doesn't fix an actual use-after-free, but avoids a deserved
> refcount_warn_saturate() warning.
> 
> A trivial fix for the racy behavior.
> 
> Fixes: 8c73b26315aa ("net/tcp: Prepare tcp_md5sig_pool for TCP-AO")
> Suggested-by: Eric Dumazet <edumazet@google.com>
> Signed-off-by: Dmitry Safonov <dima@arista.com>
> ---
>  net/ipv4/tcp_sigpool.c | 5 ++---
>  1 file changed, 2 insertions(+), 3 deletions(-)
> 
> diff --git a/net/ipv4/tcp_sigpool.c b/net/ipv4/tcp_sigpool.c
> index 55b310a722c7..8512cb09ebc0 100644
> --- a/net/ipv4/tcp_sigpool.c
> +++ b/net/ipv4/tcp_sigpool.c
> @@ -162,9 +162,8 @@ int tcp_sigpool_alloc_ahash(const char *alg, size_t scratch_size)
>  		if (strcmp(cpool[i].alg, alg))
>  			continue;
>  
> -		if (kref_read(&cpool[i].kref) > 0)
> -			kref_get(&cpool[i].kref);
> -		else
> +		/* pairs with tcp_sigpool_release() */
> +		if (!kref_get_unless_zero(&cpool[i].kref))
>  			kref_init(&cpool[i].kref);
>  		ret = i;
>  		goto out;
> 
> ---
> base-commit: 1a44b0073b9235521280e19d963b6dfef7888f18
> change-id: 20231222-tcp-ao-kref_get_unless_zero-fe7105781ba4
> 

No observable regressions when booting the kernel with this patch applied.

Tested-by: Bagas Sanjaya <bagasdotme@gmail.com>

-- 
An old man doll... just what I always wanted! - Clara