linux-kernel - Re: [PATCH] afs: Fix poential UAF in afs_make

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] [day] [month] [year] [list]

Date:   Wed, 11 Jan 2023 15:13:01 +0000
From:   David Howells <dhowells@...hat.com>
To:     Zheng Wang <zyytlz.wz@....com>
Cc:     dhowells@...hat.com, marc.dionne@...istor.com,
        linux-afs@...ts.infradead.org, linux-kernel@...r.kernel.org,
        hackerzheng666@...il.com, alex000young@...il.com,
        security@...nel.org
Subject: Re: [PATCH] afs: Fix poential UAF in afs_make_call

Zheng Wang <zyytlz.wz@....com> wrote:

> There is a function call : afs_fs_get_capabilities calls afs_make_call, in
> afs_make_call, if error occurs in rxrpc_kernel_send_data, it will call
> afs_put_call twice, which will free the call. And the access of
> call->state will trigger a use-after-free bug.

Are you sure of that?  Did you actually trigger a UAF?

> -		if (cancel_work_sync(&call->async_work))
> -			afs_put_call(call);
> +		cancel_work_sync(&call->async_work);

The problem with this change is that it will leak a ref if the work item was
queued but gets cancelled before being run, given to it here:
  
  static void afs_wake_up_async_call(struct sock *sk, struct rxrpc_call *rxcall,
  				   unsigned long call_user_ID)
  {
  	struct afs_call *call = (struct afs_call *)call_user_ID;
  	int r;

  	trace_afs_notify_call(rxcall, call);
  	call->need_attention = true;

  	if (__refcount_inc_not_zero(&call->ref, &r)) {  <----
  		trace_afs_call(call->debug_id, afs_call_trace_wake, r + 1,
  			       atomic_read(&call->net->nr_outstanding_calls),
  			       __builtin_return_address(0));

  		if (!queue_work(afs_async_calls, &call->async_work))
  			afs_put_call(call);
  	}
  }

I *think* that cancel_work_sync() returns false if the work item is executing,
but hasn't been requeued.

David