Re: [PATCH/RFC rocker-net-next 2/6] net: flow: Handle error when putting a field while putting a flow

[John Fastabend <john.fastabend@gmail.com>]

On 01/04/2015 10:50 PM, Simon Horman wrote:
> Signed-off-by: Simon Horman <simon.horman@netronome.com>
> ---
>   net/core/flow_table.c | 9 +++++++--
>   1 file changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/net/core/flow_table.c b/net/core/flow_table.c
> index 2af831e..753ebe0 100644
> --- a/net/core/flow_table.c
> +++ b/net/core/flow_table.c
> @@ -981,8 +981,9 @@ done:
>
>   int net_flow_put_flow(struct sk_buff *skb, struct net_flow_flow *flow)
>   {
> -	struct nlattr *flows, *matches;
> +	struct nlattr *flows;
>   	struct nlattr *actions = NULL; /* must be null to unwind */
> +	struct nlattr *matches = NULL; /* must be null to unwind */

Actually we don't need to initialize to NULL now. That was some crazy
unwind scheme I had in place initially.

Now we only ever do a nla_nest_cancel on nested attributes that have
been initialized with nla_nest_start(). So I can simplify this to

	struct nlattr *flows, *matches, *actions;

>   	int err, j, i = 0;
>
>   	flows = nla_nest_start(skb, NET_FLOW_FLOW);
> @@ -1005,7 +1006,11 @@ int net_flow_put_flow(struct sk_buff *skb, struct net_flow_flow *flow)
>   			if (!f->header)
>   				continue;
>
> -			nla_put(skb, NET_FLOW_FIELD_REF, sizeof(*f), f);
> +			err = nla_put(skb, NET_FLOW_FIELD_REF, sizeof(*f), f);

Great thanks. I missed this one.

> +			if (err) {
> +				nla_nest_cancel(skb, matches);
> +				goto flows_put_failure;
> +			}
>   		}
>   		nla_nest_end(skb, matches);
>   	}
>

I'll fold this into the series and resubmit thanks.

.John

-- 
John Fastabend         Intel Corporation