On 2018/2/14 18:36, Yuval Shaia wrote:
}
- skb_copy = skb_clone(skb, GFP_ATOMIC);
- if (skb_copy)
- rxe_add_ref(qp); /* for the new SKB */
Are you sure we don't need this?
From my stress tests and performance tests, it will get better performance
to remove skb_clone.
My concern is only with the above ref count.
Hi, Yuval
Thanks for your code review.
From the function send_atomic_ack, the variable skb is used in the
following 2 places.
"
...
res->type = RXE_ATOMIC_MASK;
res->atomic.skb = skb; <----One is here.
res->first_psn = ack_pkt.psn;
res->last_psn = ack_pkt.psn;
res->cur_psn = ack_pkt.psn;
rc = rxe_xmit_packet(rxe, qp, &ack_pkt, skb); <---The other is
here.
...
"
The 2 will call kfree_skb after skb is used.
void kfree_skb(struct sk_buff *skb)
{
if (!skb_unref(skb))
return;
trace_kfree_skb(skb, __builtin_return_address(0));
__kfree_skb(skb);
}
In the function kfree_skb, user ref is decreased and checked. To make
sure in the 2 places, skb can be used.
I increased user ref.
Best Regards,
Zhu Yanjun
And there is no memory leak. The whole soft RoCE can work well.
So I think removing this function is a good choice.
Zhu Yanjun
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
