On Thu, 2025-02-27 at 09:35 +0100, Eric Dumazet wrote:

On Thu, Feb 27, 2025 at 9:24 AM Geliang Tang [off-list ref]
wrote:

quoted

From: Geliang Tang <redacted>

Instead of using sock_kmalloc() to allocate a tcp_ao_key "new_key"
and
then immediately duplicate the input "key" to it in
tcp_ao_copy_key(),
the newly added sock_kmemdup() helper can be used to simplify the
code.

Signed-off-by: Geliang Tang <redacted>
---
 net/ipv4/tcp_ao.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/net/ipv4/tcp_ao.c b/net/ipv4/tcp_ao.c
index bbb8d5f0eae7..d21412d469cc 100644
--- a/net/ipv4/tcp_ao.c
+++ b/net/ipv4/tcp_ao.c

@@ -246,12 +246,11 @@ static struct tcp_ao_key

*tcp_ao_copy_key(struct sock *sk,
 {
        struct tcp_ao_key *new_key;

-       new_key = sock_kmalloc(sk, tcp_ao_sizeof_key(key),
+       new_key = sock_kmemdup(sk, key, tcp_ao_sizeof_key(key),
                               GFP_ATOMIC);
        if (!new_key)
                return NULL;

-       *new_key = *key;

Note that this only copies 'sizeof(struct tcp_ao_key)' bytes, which
is
smaller than tcp_ao_sizeof_key(key)

Yes, indeed. sock_kmemdup() shouldn't be used here then. I'll drop this
patch in v2.

Thanks,
-Geliang