[dpdk-dev] [PATCH v3] net/i40e: fix incorrect byte counters

[Ferruh Yigit]

On 9/21/2020 10:59 AM, Junyu Jiang wrote:
> This patch fixed the issue that rx/tx bytes statistics counters
> overflowed on 48 bit limitation by enlarging the limitation.
> 
> Fixes: 4861cde46116 ("i40e: new poll mode driver")
> Cc: stable at dpdk.org
> 
> Signed-off-by: Junyu Jiang <junyux.jiang at intel.com>
> ---
>   doc/guides/nics/i40e.rst       |  7 +++++++
>   drivers/net/i40e/i40e_ethdev.c | 32 ++++++++++++++++++++++++++++++++
>   drivers/net/i40e/i40e_ethdev.h |  9 +++++++++
>   3 files changed, 48 insertions(+)
> 
> diff --git a/doc/guides/nics/i40e.rst b/doc/guides/nics/i40e.rst
> index b7430f6c4..4baa58be6 100644
> --- a/doc/guides/nics/i40e.rst
> +++ b/doc/guides/nics/i40e.rst
> @@ -830,3 +830,10 @@ Tx bytes affected by the link status change
>   
>   For firmware versions prior to 6.01 for X710 series and 3.33 for X722 series, the tx_bytes statistics data is affected by
>   the link down event. Each time the link status changes to down, the tx_bytes decreases 110 bytes.
> +
> +RX/TX statistics may be incorrect when register overflowed
> +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> +
> +The rx_bytes/tx_bytes statistics register is 48 bit length. Although this limitation is enlarged to 64 bit length
> +on the software side, but there is no way to detect if the overflow occurred more than once. So rx_bytes/tx_bytes
> +statistics data is correct when statistics are updated at least once between two overflows.
> diff --git a/drivers/net/i40e/i40e_ethdev.c b/drivers/net/i40e/i40e_ethdev.c
> index 563f21d9d..212338ef0 100644
> --- a/drivers/net/i40e/i40e_ethdev.c
> +++ b/drivers/net/i40e/i40e_ethdev.c
> @@ -3052,6 +3052,19 @@ i40e_dev_link_update(struct rte_eth_dev *dev,
>   	return ret;
>   }
>   
> +static void
> +i40e_stat_update_48_in_64(uint64_t *new_bytes,
> +			  uint64_t *prev_bytes,
> +			  bool offset_loaded)
> +{
> +	if (offset_loaded) {
> +		if (I40E_RXTX_BYTES_L_48_BIT(*prev_bytes) > *new_bytes)
> +			*new_bytes += (uint64_t)1 << I40E_48_BIT_WIDTH;
> +		*new_bytes += I40E_RXTX_BYTES_H_16_BIT(*prev_bytes);
> +	}
> +	*prev_bytes = *new_bytes;
> +}
> +

I was more thinking reading stats and extending in same function, 
instead of extracting the extending part into its own function, like:

static void
i40e_stat_update_48_in_64(struct i40e_hw *hw,
		uint32_t hireg,
		uint32_t loreg,
		bool offset_loaded,
		uint64_t *offset,
		uint64_t *stat,
		uint64_t *prev_bytes) {

	i40e_stat_update_48(...)

	/* logic to convert 'stat' to 64 bits */
}

Does it make sense?