Re: [PATCH v3 1/4] ethdev: increase port_id range

[Adrien Mazarguil <adrien.mazarguil@6wind.com>]

Hi Zhiyong,

On Sat, Sep 09, 2017 at 10:47:24PM +0800, Zhiyong Yang wrote:
> Extend port_id definition from uint8_t to uint16_t in lib and drivers
> data structures, specifically rte_eth_dev_data.
> Modify the APIs, drivers and app using port_id at the same time.
> 
> Fix some checkpatch issues from the original code and remove some
> unnecessary cast operations.
> 
> Signed-off-by: Zhiyong Yang <zhiyong.yang@intel.com>

Thanks for the mlx4/mlx5 update, I have some comments, see below.

By the way, since this commit breaks ABI compatibility for pretty much all
public functions in librte_ether and a few in other places, remember to
update rte_ethdev_version.map and other affected .map files accordingly.
I think this change has to be part of this series.

<snip>
> diff --git a/drivers/net/mlx4/mlx4.h b/drivers/net/mlx4/mlx4.h
> index c0ade4f1a..fe911c3a7 100644
> --- a/drivers/net/mlx4/mlx4.h
> +++ b/drivers/net/mlx4/mlx4.h
> @@ -334,7 +334,7 @@ struct priv {
>  	} vlan_filter[MLX4_MAX_VLAN_IDS]; /* VLAN filters table. */
>  	/* Device properties. */
>  	uint16_t mtu; /* Configured MTU. */
> -	uint8_t port; /* Physical port number. */
> +	uint16_t port; /* Physical port number. */
>  	unsigned int started:1; /* Device started, flows enabled. */
>  	unsigned int promisc:1; /* Device in promiscuous mode. */
>  	unsigned int allmulti:1; /* Device receives all multicast packets. */

This is a physical port number internal to the device, not DPDK's. Please
drop this change.

What would need change is struct rxq's port_id field (already large enough)
if you want to enforce uint16_t everywhere.

> diff --git a/drivers/net/mlx5/mlx5.h b/drivers/net/mlx5/mlx5.h
> index 43c538419..54a2e8a54 100644
> --- a/drivers/net/mlx5/mlx5.h
> +++ b/drivers/net/mlx5/mlx5.h
> @@ -114,7 +114,7 @@ struct priv {
>  	unsigned int vlan_filter_n; /* Number of configured VLAN filters. */
>  	/* Device properties. */
>  	uint16_t mtu; /* Configured MTU. */
> -	uint8_t port; /* Physical port number. */
> +	uint16_t port; /* Physical port number. */

Same comment here.

>  	unsigned int started:1; /* Device started, flows enabled. */
>  	unsigned int promisc_req:1; /* Promiscuous mode requested. */
>  	unsigned int allmulti_req:1; /* All multicast mode requested. */
> diff --git a/drivers/net/mlx5/mlx5_ethdev.c b/drivers/net/mlx5/mlx5_ethdev.c
> index b0eb3cdfc..58e7cf571 100644
> --- a/drivers/net/mlx5/mlx5_ethdev.c
> +++ b/drivers/net/mlx5/mlx5_ethdev.c
> @@ -75,7 +75,7 @@ struct ethtool_link_settings {
>  	uint32_t cmd;
>  	uint32_t speed;
>  	uint8_t duplex;
> -	uint8_t port;
> +	uint16_t port;

This is the ethtool interface from the Linux kernel, again this field has to
remain 8 bits.

That's the kind of mistakes a dedicated type might have prevented by the
way, I'm still not convinced that not adding one was the right decision.

>  	uint8_t phy_address;
>  	uint8_t autoneg;
>  	uint8_t mdio_support;
> diff --git a/drivers/net/mlx5/mlx5_rxtx.h b/drivers/net/mlx5/mlx5_rxtx.h
> index 7de1d1086..c47e54fe0 100644
> --- a/drivers/net/mlx5/mlx5_rxtx.h
> +++ b/drivers/net/mlx5/mlx5_rxtx.h
> @@ -112,7 +112,6 @@ struct rxq {
>  	unsigned int sges_n:2; /* Log 2 of SGEs (max buffers per packet). */
>  	unsigned int cqe_n:4; /* Log 2 of CQ elements. */
>  	unsigned int elts_n:4; /* Log 2 of Mbufs. */
> -	unsigned int port_id:8;
>  	unsigned int rss_hash:1; /* RSS hash result is enabled. */
>  	unsigned int mark:1; /* Marked flow available on the queue. */
>  	unsigned int pending_err:1; /* CQE error needs to be handled. */
> @@ -120,6 +119,7 @@ struct rxq {
>  	unsigned int :6; /* Remaining bits. */
>  	volatile uint32_t *rq_db;
>  	volatile uint32_t *cq_db;
> +	uint16_t port_id;
>  	uint16_t rq_ci;
>  	uint16_t rq_pi;
>  	uint16_t cq_ci;

This change might have a performance impact. All important fields are found
at the beginning of this structure for that reason. Since there's not enough
room in the above bit-field (8 + 6), I guess there's no other choice for
now, however you should at least update the "Remaining bits" comment.

<snip>

-- 
Adrien Mazarguil
6WIND