Re: [PATCH] xprtsock: Fix a loop in xs_tcp_setup_socket()

From: Trond Myklebust
Date: Sun Apr 21 2024 - 18:35:21 EST


On Sat, 2024-04-20 at 19:48 +0900, Lex Siegel wrote:
> [You don't often get email from usiegl00@xxxxxxxxx. Learn why this is
> important at https://aka.ms/LearnAboutSenderIdentification ;]
>
> When using a bpf on kernel_connect(), the call can return -EPERM.
> This causes xs_tcp_setup_socket() to loop forever, filling up the
> syslog and causing the kernel to freeze up.
>
> Signed-off-by: Lex Siegel <usiegl00@xxxxxxxxx>
> ---
>  net/sunrpc/xprtsock.c | 2 ++
>  1 file changed, 2 insertions(+)
>
> diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
> index bb9b747d58a1..47b254806a08 100644
> --- a/net/sunrpc/xprtsock.c
> +++ b/net/sunrpc/xprtsock.c
> @@ -2446,6 +2446,8 @@ static void xs_tcp_setup_socket(struct
> work_struct *work)
>                 /* Happens, for instance, if the user specified a
> link
>                  * local IPv6 address without a scope-id.
>                  */
> +       case -EPERM:
> +               /* Happens, for instance, if a bpf is preventing the
> connect */
>         case -ECONNREFUSED:
>         case -ECONNRESET:
>         case -ENETDOWN:
> --
> 2.39.3
>

This looks as if it will have some rather subtle interactions if the
RPC_TASK_SOFTCONN flag is set. Have you tested this kind of scenario?

--
Trond Myklebust
Linux NFS client maintainer, Hammerspace
trond.myklebust@xxxxxxxxxxxxxxx