syzbot is reporting uninit-value at aes_encrypt(), for block cipher assumes that bytes to encrypt/decrypt is multiple of block size for that cipher but tls_alloc_encrypted_msg() is not initializing padding bytes when required_size is not multiple of block cipher's block size. In order to make sure that padding bytes are automatically initialized, enable __GFP_ZERO flag when setsockopt(SOL_TCP, TCP_ULP, "tls") is called. Reported-by: syzbot <syzbot+828dfc12440b4f6f305d@xxxxxxxxxxxxxxxxxxxxxxxxx> Closes: https://syzkaller.appspot.com/bug?extid=828dfc12440b4f6f305d Signed-off-by: Tetsuo Handa <penguin-kernel@xxxxxxxxxxxxxxxxxxx> --- According to C reproducer, this problem happens when bpf_exec_tx_verdict() is called with lower 4 bits of required_size being 0001 and does not happen when being 0100. Thus, I assumed that this problem is caused by lack of initializing padding bytes. But I couldn't figure out why KMSAN reports this problem when bpf_exec_tx_verdict() is called with lower 4 bits of required_size being 0001 for the second time and does not report this problem when bpf_exec_tx_verdict() is called with lower 4 bits of required_size being 0001 for the first time. More deeper problem exists? KMSAN reporting this problem when accessing u64 relevant? net/tls/tls_main.c | 1 + 1 file changed, 1 insertion(+) diff --git a/net/tls/tls_main.c b/net/tls/tls_main.c index f2e7302a4d96..cd5366966864 100644 --- a/net/tls/tls_main.c +++ b/net/tls/tls_main.c @@ -1025,6 +1025,7 @@ static int tls_init(struct sock *sk) struct tls_context *ctx; int rc = 0; + sk->sk_allocation |= __GFP_ZERO; tls_build_proto(sk); #ifdef CONFIG_TLS_TOE -- 2.34.1
