kernel: audit.c: is it a memleak of net in function audit_send_reply

[Paul Moore]

On Mon, Apr 20, 2020 at 3:54 AM 亿一 <teroincn at gmail.com> wrote:
>
> Hi, all:
>
> when reviewing code in function audit_send_reply, I notice that if
> kthread_run return failure, the net reference would not be released
> because reply has been kfree directly.

Thanks for reporting this.

Looking at the code, it's a little worse than that.  If kthread_run()
fails then audit_send_reply() will return early, holding both a
reference to @net as well as leaking @reply.

Let me finish getting through my mail and I'll put together a quick
patch to resolve this (I'm seeing a few other related things we should
fix in audit_send_reply()).

> static void audit_send_reply(struct sk_buff *request_skb, int seq, int
> type, int done,
>      int multi, const void *payload, int size)
> {
>     struct net *net = sock_net(NETLINK_CB(request_skb).sk);
>     struct sk_buff *skb;
>     struct task_struct *tsk;
>     struct audit_reply *reply = kmalloc(sizeof(struct audit_reply),
>                                                            GFP_KERNEL);
>
>     if (!reply)
>         return;
>
>     skb = audit_make_reply(seq, type, done, multi, payload, size);
>     if (!skb)
>         goto out;
>
>     reply->net = get_net(net);     // hold a reference of net here
>     reply->portid = NETLINK_CB(request_skb).portid;
>     reply->skb = skb;
>
>     tsk = kthread_run(audit_send_reply_thread, reply, "audit_send_reply");
>     if (!IS_ERR(tsk))
>         return;
>     kfree_skb(skb);
>
>     out:
>         kfree(reply);      // kfree reply without release the net reference.
> }



-- 
paul moore
www.paul-moore.com