Re: [4.4, 027/193] binder: check for binder_thread allocation failure in binder_poll()

[Guenter Roeck]

On Fri, Feb 23, 2018 at 07:24:20PM +0100, gregkh@xxxxxxxxxxxxxxxxxxx wrote:
> 4.4-stable review patch.  If anyone has any objections, please let me know.
> 
> ------------------
> 
> From: Eric Biggers <ebiggers@xxxxxxxxxx>
> 
> commit f88982679f54f75daa5b8eff3da72508f1e7422f upstream.
> 
> If the kzalloc() in binder_get_thread() fails, binder_poll()
> dereferences the resulting NULL pointer.
> 
> Fix it by returning POLLERR if the memory allocation failed.
> 
> This bug was found by syzkaller using fault injection.
> 
> Reported-by: syzbot <syzkaller@xxxxxxxxxxxxxxxx>
> Fixes: 457b9a6f09f0 ("Staging: android: add binder driver")
> Cc: stable@xxxxxxxxxxxxxxx
> Signed-off-by: Eric Biggers <ebiggers@xxxxxxxxxx>
> Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> ---
>  drivers/android/binder.c |    2 ++
>  1 file changed, 2 insertions(+)
> 
> --- a/drivers/android/binder.c
> +++ b/drivers/android/binder.c
> @@ -2622,6 +2622,8 @@ static unsigned int binder_poll(struct f
>  	binder_lock(__func__);
>  
>  	thread = binder_get_thread(proc);
> +	if (!thread)
> +		return POLLERR;
>  
Noticed while merging into chromeos-4.4:

This will cause trouble in v4.4.y. Notice the call to "binder_lock(__func__)"
above. This call has been removed upstream, but not in v4.4.y. As a result,
the lock won't be released, which will result in subsequent hangups
if/when the function is called again.

v4.9.y has the same problem. v4.14.y+ are fine.

Greg - can you fix this up yourself or do you want me to send fixup
patches ? It might take a few days for me to get to it.

Guenter

>  	wait_for_proc_work = thread->transaction_stack == NULL &&
>  		list_empty(&thread->todo) && thread->return_error == BR_OK;