netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH 2.5.69] Bug in sys_accept() module ref counts
@ 2003-05-06 19:28 Sridhar Samudrala
  2003-05-07  0:14 ` Arnaldo Carvalho de Melo
  0 siblings, 1 reply; 3+ messages in thread
From: Sridhar Samudrala @ 2003-05-06 19:28 UTC (permalink / raw)
  To: davem, acme; +Cc: netdev


I think there is a bug in the recent changes to sys_accept() to implement 
module ref counts.

module_put() gets called twice on error. Once via the explicit module_put and
the second via sock_release(). Also i think we should do a __module_get() with 
newsock's owner(although same as the original listening sock).

The following patch against 2.5.69 should fix the problem.

Thanks
Sridhar
-------------------------------------------------------------------------------
diff -Nru a/net/socket.c b/net/socket.c
--- a/net/socket.c	Tue May  6 12:14:35 2003
+++ b/net/socket.c	Tue May  6 12:14:35 2003
@@ -1280,26 +1280,26 @@
 	 * We don't need try_module_get here, as the listening socket (sock)
 	 * has the protocol module (sock->ops->owner) held.
 	 */
-	__module_get(sock->ops->owner);
+	__module_get(newsock->ops->owner);
 
 	err = sock->ops->accept(sock, newsock, sock->file->f_flags);
 	if (err < 0)
-		goto out_module_put;
+		goto out_release;
 
 	if (upeer_sockaddr) {
 		if(newsock->ops->getname(newsock, (struct sockaddr *)address, &len, 2)<0) {
 			err = -ECONNABORTED;
-			goto out_module_put;
+			goto out_release;
 		}
 		err = move_addr_to_user(address, len, upeer_sockaddr, upeer_addrlen);
 		if (err < 0)
-			goto out_module_put;
+			goto out_release;
 	}
 
 	/* File flags are not inherited via accept() unlike another OSes. */
 
 	if ((err = sock_map_fd(newsock)) < 0)
-		goto out_module_put;
+		goto out_release;
 
 	security_socket_post_accept(sock, newsock);
 
@@ -1307,8 +1307,6 @@
 	sockfd_put(sock);
 out:
 	return err;
-out_module_put:
-	module_put(sock->ops->owner);
 out_release:
 	sock_release(newsock);
 	goto out_put;

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH 2.5.69] Bug in sys_accept() module ref counts
  2003-05-06 19:28 [PATCH 2.5.69] Bug in sys_accept() module ref counts Sridhar Samudrala
@ 2003-05-07  0:14 ` Arnaldo Carvalho de Melo
  2003-05-07  0:43   ` Arnaldo Carvalho de Melo
  0 siblings, 1 reply; 3+ messages in thread
From: Arnaldo Carvalho de Melo @ 2003-05-07  0:14 UTC (permalink / raw)
  To: Sridhar Samudrala; +Cc: davem, netdev

Em Tue, May 06, 2003 at 12:28:41PM -0700, Sridhar Samudrala escreveu:
> 
> I think there is a bug in the recent changes to sys_accept() to implement 
> module ref counts.

Yes, well spotted, small comment below, I'll be sending this patch
to DaveM, thanks a lot!
 
> module_put() gets called twice on error. Once via the explicit module_put and
> the second via sock_release(). Also i think we should do a __module_get() with 
> newsock's owner(although same as the original listening sock).
> 
> The following patch against 2.5.69 should fix the problem.
> 
> Thanks
> Sridhar
> -------------------------------------------------------------------------------
> diff -Nru a/net/socket.c b/net/socket.c
> --- a/net/socket.c	Tue May  6 12:14:35 2003
> +++ b/net/socket.c	Tue May  6 12:14:35 2003
> @@ -1280,26 +1280,26 @@
>  	 * We don't need try_module_get here, as the listening socket (sock)
>  	 * has the protocol module (sock->ops->owner) held.
>  	 */
> -	__module_get(sock->ops->owner);
> +	__module_get(newsock->ops->owner);

This one is OK, but the two operations are the same, so the effect, as well,
is the same, but for correctness, better have it with newsock.
  
>  	err = sock->ops->accept(sock, newsock, sock->file->f_flags);
>  	if (err < 0)
> -		goto out_module_put;
> +		goto out_release;
>  
>  	if (upeer_sockaddr) {
>  		if(newsock->ops->getname(newsock, (struct sockaddr *)address, &len, 2)<0) {
>  			err = -ECONNABORTED;
> -			goto out_module_put;
> +			goto out_release;
>  		}
>  		err = move_addr_to_user(address, len, upeer_sockaddr, upeer_addrlen);
>  		if (err < 0)
> -			goto out_module_put;
> +			goto out_release;
>  	}
>  
>  	/* File flags are not inherited via accept() unlike another OSes. */
>  
>  	if ((err = sock_map_fd(newsock)) < 0)
> -		goto out_module_put;
> +		goto out_release;
>  
>  	security_socket_post_accept(sock, newsock);
>  
> @@ -1307,8 +1307,6 @@
>  	sockfd_put(sock);
>  out:
>  	return err;
> -out_module_put:
> -	module_put(sock->ops->owner);
>  out_release:
>  	sock_release(newsock);
>  	goto out_put;
> 

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH 2.5.69] Bug in sys_accept() module ref counts
  2003-05-07  0:14 ` Arnaldo Carvalho de Melo
@ 2003-05-07  0:43   ` Arnaldo Carvalho de Melo
  0 siblings, 0 replies; 3+ messages in thread
From: Arnaldo Carvalho de Melo @ 2003-05-07  0:43 UTC (permalink / raw)
  To: Sridhar Samudrala; +Cc: davem, netdev

Em Tue, May 06, 2003 at 09:14:19PM -0300, Arnaldo C. Melo escreveu:
> Em Tue, May 06, 2003 at 12:28:41PM -0700, Sridhar Samudrala escreveu:

> > Also i think we should do a __module_get() with newsock's owner(although
> > same as the original listening sock).

Forget about the comment, its what you said above, sorry for answering too fast
8)

- Arnaldo

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2003-05-07  0:43 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2003-05-06 19:28 [PATCH 2.5.69] Bug in sys_accept() module ref counts Sridhar Samudrala
2003-05-07  0:14 ` Arnaldo Carvalho de Melo
2003-05-07  0:43   ` Arnaldo Carvalho de Melo

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).