diff mbox series

sunrpc: bug on rpc_task_set_client when no client is present.

Message ID 20211018123812.71482-1-trbecker@gmail.com (mailing list archive)
State New, archived
Headers show
Series sunrpc: bug on rpc_task_set_client when no client is present. | expand

Commit Message

Thiago Rafael Becker Oct. 18, 2021, 12:38 p.m. UTC
If we pass a NULL client to rpc_task_set_client and no client is
attached to the task, then the kernel will crash later. Antecipate the
crash by checking if a client is available for the task.

Signed-off-by: Thiago Rafael Becker <trbecker@gmail.com>
---
 net/sunrpc/clnt.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Comments

Trond Myklebust Oct. 20, 2021, 7:04 p.m. UTC | #1
On Mon, 2021-10-18 at 09:38 -0300, Thiago Rafael Becker wrote:
> If we pass a NULL client to rpc_task_set_client and no client is
> attached to the task, then the kernel will crash later. Antecipate
> the
> crash by checking if a client is available for the task.
> 
> Signed-off-by: Thiago Rafael Becker <trbecker@gmail.com>
> ---
>  net/sunrpc/clnt.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
> index f056ff931444..ccbc9a9715da 100644
> --- a/net/sunrpc/clnt.c
> +++ b/net/sunrpc/clnt.c
> @@ -1076,7 +1076,7 @@ void rpc_task_set_transport(struct rpc_task
> *task, struct rpc_clnt *clnt)
>  static
>  void rpc_task_set_client(struct rpc_task *task, struct rpc_clnt
> *clnt)
>  {
> -
> +       BUG_ON(clnt == NULL && task->tk_client == NULL);
>         if (clnt != NULL) {
>                 rpc_task_set_transport(task, clnt);
>                 task->tk_client = clnt;


I'm not seeing the point of this BUG_ON(). Why not just change this
code to not check for clnt == NULL, and let the thing Oops when it
tries to dereference clnt?
Thiago Rafael Becker Oct. 20, 2021, 8:20 p.m. UTC | #2
On Wed, Oct 20, 2021 at 07:04:35PM +0000, Trond Myklebust wrote:
Hello,

> I'm not seeing the point of this BUG_ON(). Why not just change this
> code to not check for clnt == NULL, and let the thing Oops when it
> tries to dereference clnt?

This was changed in 58f9612c6ea85, prior to that, this was not tested. I'm
not sure why this test exists, the only reason I can imagine is to keep
the previous task's rpc_client in case the new client is NULL. Decided
to go conservative on this, and BUG_ON() when no client is available.

Inside the linux source, I don't see how this may happen unless the code
has a bug, so I think it's possible to remove this test.

-- Thiago
 
> -- 
> Trond Myklebust
> Linux NFS client maintainer, Hammerspace
> trond.myklebust@hammerspace.com
> 
>
diff mbox series

Patch

diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index f056ff931444..ccbc9a9715da 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -1076,7 +1076,7 @@  void rpc_task_set_transport(struct rpc_task *task, struct rpc_clnt *clnt)
 static
 void rpc_task_set_client(struct rpc_task *task, struct rpc_clnt *clnt)
 {
-
+	BUG_ON(clnt == NULL && task->tk_client == NULL);
 	if (clnt != NULL) {
 		rpc_task_set_transport(task, clnt);
 		task->tk_client = clnt;