Re: [PATCH] epoll: fix use-after-free in eventpoll_release_file

[Jason Baron]

Hi,

Indeed, eventpoll_release_file() is on the 'write' side of the rcu
here (the read-side being in reverse_path_check_proc()). It's a race,
in that 'struct epitem' has to be freed by rcu before the next loop
iteration such that epi->fllink.next pointer is no longer valid. So,
I suspect that makes this more difficult to hit, but obviously it
happens.

Thanks for finding/fixing this!

Acked-by: Jason Baron <jbaron@xxxxxxxxxx>


On 06/16/2014 10:58 PM, Konstantin Khlebnikov wrote:
> This fixes use-after-free of epi->fllink.next inside list loop macro.
> This loop actually releases elements in the body. List is rcu-protected
> but here we cannot hold rcu_read_lock because we need to lock mutex inside.
> 
> Obvious solution is to use list_for_each_entry_safe(). RCU-ness isn't essential
> because nobody can change this list under us, it's final fput for this file.
> 
> Bug is here since ae10b2b4eb01bedc91d29d5c5bb9e416fd806c40
> ("epoll: optimize EPOLL_CTL_DEL using rcu")
> 
> Signed-off-by: Konstantin Khlebnikov <koct9i@xxxxxxxxx>
> Reported-by: Cyrill Gorcunov <gorcunov@xxxxxxxxxx>
> Cc: Stable <stable@xxxxxxxxxxxxxxx> # 3.13+
> Cc: Sasha Levin <sasha.levin@xxxxxxxxxx>
> Cc: Jason Baron <jbaron@xxxxxxxxxx>
> ---
>  fs/eventpoll.c |    4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/eventpoll.c b/fs/eventpoll.c
> index b73e062..b10b48c 100644
> --- a/fs/eventpoll.c
> +++ b/fs/eventpoll.c
> @@ -910,7 +910,7 @@ static const struct file_operations eventpoll_fops = {
>  void eventpoll_release_file(struct file *file)
>  {
>  	struct eventpoll *ep;
> -	struct epitem *epi;
> +	struct epitem *epi, *next;
>  
>  	/*
>  	 * We don't want to get "file->f_lock" because it is not
> @@ -926,7 +926,7 @@ void eventpoll_release_file(struct file *file)
>  	 * Besides, ep_remove() acquires the lock, so we can't hold it here.
>  	 */
>  	mutex_lock(&epmutex);
> -	list_for_each_entry_rcu(epi, &file->f_ep_links, fllink) {
> +	list_for_each_entry_safe(epi, next, &file->f_ep_links, fllink) {
>  		ep = epi->ep;
>  		mutex_lock_nested(&ep->mtx, 0);
>  		ep_remove(ep, epi);
> 

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/