Re: [PATCH 1/1] IB/rxe: avoid double kfree_skb

2018-04-27 Thread Doug Ledford
On Wed, 2018-04-25 at 14:56 +0800, Yanjun Zhu wrote: > Hi, all > > rxe_send [rdma_rxe] > ip_local_out > __ip_local_out > ip_output > ip_finish_output > ip_finish_output2 > dev_queue_xmit >

Re: [PATCH 1/1] IB/rxe: avoid double kfree_skb

2018-04-25 Thread Yanjun Zhu
Hi, all rxe_send [rdma_rxe]     ip_local_out         __ip_local_out         ip_output             ip_finish_output                 ip_finish_output2                     dev_queue_xmit                         __dev_queue_xmit                             dev_hard_start_xmit                        

Re: [PATCH 1/1] IB/rxe: avoid double kfree_skb

2018-04-24 Thread Yanjun Zhu
Hi, all rxe_send     ip_local_out         __ip_local_out             nf_hook_slow In the above call process, nf_hook_slow drops and frees skb, then -EPERM is returned when iptables rules(iptables -I OUTPUT -p udp --dport 4791 -j DROP) is set. If skb->users is not changed in softroce,

Re: [PATCH 1/1] IB/rxe: avoid double kfree_skb

2018-04-19 Thread Yanjun Zhu
On 2018/4/20 10:19, Doug Ledford wrote: On Thu, 2018-04-19 at 10:01 -0400, Zhu Yanjun wrote: When skb is dropped by iptables rules, the skb is freed at the same time -EPERM is returned. So in softroce, it is not necessary to free skb again. Or else, crash will occur. The steps to reproduce:

Re: [PATCH 1/1] IB/rxe: avoid double kfree_skb

2018-04-19 Thread Doug Ledford
On Thu, 2018-04-19 at 10:01 -0400, Zhu Yanjun wrote: > When skb is dropped by iptables rules, the skb is freed at the same time > -EPERM is returned. So in softroce, it is not necessary to free skb again. > Or else, crash will occur. > > The steps to reproduce: > > server