Re: [PATCH net] sctp: not copy sctp_sock pd_lobby in sctp_copy_descendant

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Mon, Mar 18, 2019 at 07:58:29PM +0800, Xin Long wrote:
> Now sctp_copy_descendant() copies pd_lobby from old sctp scok to new
> sctp sock. If sctp_sock_migrate() returns error, it will panic when
> releasing new sock and trying to purge pd_lobby due to the incorrect
> pointers in pd_lobby.
> 
>   [  120.485116] kasan: CONFIG_KASAN_INLINE enabled
>   [  120.486270] kasan: GPF could be caused by NULL-ptr deref or user
>   [  120.509901] Call Trace:
>   [  120.510443]  sctp_ulpevent_free+0x1e8/0x490 [sctp]
>   [  120.511438]  sctp_queue_purge_ulpevents+0x97/0xe0 [sctp]
>   [  120.512535]  sctp_close+0x13a/0x700 [sctp]
>   [  120.517483]  inet_release+0xdc/0x1c0
>   [  120.518215]  __sock_release+0x1d2/0x2a0
>   [  120.519025]  sctp_do_peeloff+0x30f/0x3c0 [sctp]
> 
> We fix it by not copying sctp_sock pd_lobby in sctp_copy_descendan(),
> and skb_queue_head_init() can also be removed in sctp_sock_migrate().
> 
> Reported-by: syzbot+85e0b422ff140b03672a@xxxxxxxxxxxxxxxxxxxxxxxxx
> Fixes: 89664c623617 ("sctp: sctp_sock_migrate() returns error if sctp_bind_addr_dup() fails")
> Signed-off-by: Xin Long <lucien.xin@xxxxxxxxx>
> ---
>  net/sctp/socket.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/net/sctp/socket.c b/net/sctp/socket.c
> index 6140471..65b5386 100644
> --- a/net/sctp/socket.c
> +++ b/net/sctp/socket.c
> @@ -9169,7 +9169,7 @@ static inline void sctp_copy_descendant(struct sock *sk_to,
>  {
>  	int ancestor_size = sizeof(struct inet_sock) +
>  			    sizeof(struct sctp_sock) -
> -			    offsetof(struct sctp_sock, auto_asconf_list);
> +			    offsetof(struct sctp_sock, pd_lobby);
>  
>  	if (sk_from->sk_family == PF_INET6)
>  		ancestor_size += sizeof(struct ipv6_pinfo);
> @@ -9253,7 +9253,6 @@ static int sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
>  	 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
>  	 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
>  	 */
> -	skb_queue_head_init(&newsp->pd_lobby);
>  	atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
>  
>  	if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
> -- 
> 2.1.0
> 
> 
Acked-by: Neil Horman <nhorman@xxxxxxxxxxxxx>



[Index of Archives]     [Linux Networking Development]     [Linux OMAP]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux