[ovs-dev] [PATCH v3 1/9] dpif-netdev: associate flow with a mark id

[Chandran, Sugesh]

Regards
_Sugesh


> -----Original Message-----
> From: Yuanhan Liu [mailto:yliu at fridaylinux.org]
> Sent: Tuesday, September 26, 2017 6:37 AM
> To: dev at openvswitch.org
> Cc: Finn Christensen <fc at napatech.com>; Darrell Ball <dball at vmware.com>;
> Chandran, Sugesh <sugesh.chandran at intel.com>; Simon Horman
> <simon.horman at netronome.com>; Yuanhan Liu <yliu at fridaylinux.org>
> Subject: [PATCH v3 1/9] dpif-netdev: associate flow with a mark id
> 
> This patch associate a flow with a mark id (a uint32_t number) by a direct array.
> The array is initiated with 1024 slots at the begninning, and it will be enlarged
> when it's not big enough.  The mark id is allocated with the help of id-pool.
> 
> It re-uses the flow API (more precisely, the ->flow_put method) to setup the hw
> flow. The flow_put implementation then is supposed to create a flow with
> MARK action for current stage, which is, only partial offload is supported.
> 
> Co-authored-by: Finn Christensen <fc at napatech.com>
> Signed-off-by: Yuanhan Liu <yliu at fridaylinux.org>
> Signed-off-by: Finn Christensen <fc at napatech.com>
> ---
>  lib/dpif-netdev.c | 140
> ++++++++++++++++++++++++++++++++++++++++++++++++++++++
>  lib/netdev.h      |   6 +++
>  2 files changed, 146 insertions(+)
> 
> diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c index 071ec14..f417b25 100644
> --- a/lib/dpif-netdev.c
> +++ b/lib/dpif-netdev.c
> @@ -453,6 +453,9 @@ struct dp_netdev_flow {
>      struct ovs_refcount ref_cnt;
> 
>      bool dead;
> +    bool has_mark;               /* A flag to tell whether this flow has a
> +                                    valid mark asscoiated with it. */
> +    uint32_t mark;               /* Unique flow mark assiged to a flow */
> 
>      /* Statistics. */
>      struct dp_netdev_flow_stats stats;
> @@ -1833,6 +1836,82 @@ dp_netdev_pmd_find_dpcls(struct
> dp_netdev_pmd_thread *pmd,
>      return cls;
>  }
> 
> +/*
> + * An array to map from the flow mark (the index) to dp netdev flow.
> + * Initial size is set to 1024, it will be resized when it's not
> + * big enough.
> + */
> +struct flow_mark_map {
> +    size_t max;
> +    struct ovs_mutex mutex;
> +    struct id_pool *mark_pool;
> +    struct dp_netdev_flow **flows;
> +};
[Sugesh] Is it good to make this map is for per port/device.?
The reason being, any access to this map should be mutex protected.
I feel there will be contention if multiple PMD threads trying to program the flows, even
though the flows are in different devices. It can also happen with revalidator threads on flow delete.

> +
> +static struct flow_mark_map flow_mark_map = {
> +    .max = 1024,
> +    .mutex = OVS_MUTEX_INITIALIZER,
> +};
> +

[snip]
> --
> 2.7.4