Re: [PATCH iwl-next v1] ixgbe: fix eeprom_id staleness and non-fatal EMPR reload failure

[Simon Horman <horms@kernel.org>]

On Fri, Mar 20, 2026 at 06:14:54AM +0100, Aleksandr Loktionov wrote:
> Three related bugs around FW version reporting after EMPR reset on E610:
> 
> 1. ixgbe_refresh_fw_version() silently discarded the error return from
>    ixgbe_get_flash_data(), so a failed NVM re-read left adapter->eeprom_id
>    with whatever stale data it held before the reset.  Propagate the error
>    and set eeprom_id to "unknown" on failure so that ethtool -i and
>    devlink dev info never show a version that no longer reflects reality.
> 
> 2. ixgbe_devlink_reload_empr_finish() returned 0 without ever refreshing
>    the FW version after the EMPR completed.  Add the refresh call, but
>    treat it as best-effort: a failure to re-read flash does not mean the
>    EMPR itself failed.  Log a netdev_warn() and return 0 so devlink
>    reports the correct reload outcome.
> 
> 3. ixgbe_reinit_locked() never refreshed the FW version for E610.
>    Because E610 has no FW event that notifies peer PFs when an EMPR
>    triggered by another PF's devlink reload completes, any PF that
>    subsequently goes through reinit would keep stale data in hw->flash
>    and adapter->eeprom_id.  Add the same best-effort refresh here,
>    gated on ixgbe_mac_e610, with a netdev_warn() on failure.
> 
> Signed-off-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>

Reviewed-by: Simon Horman <horms@kernel.org>

...

> diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_ethtool.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_ethtool.c
> index 56aabaa..40d593b 100644
> --- a/drivers/net/ethernet/intel/ixgbe/ixgbe_ethtool.c
> +++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_ethtool.c
> @@ -1155,12 +1155,30 @@ static int ixgbe_set_eeprom(struct net_device *netdev,
>  	return ret_val;
>  }
>  
> -void ixgbe_refresh_fw_version(struct ixgbe_adapter *adapter)
> +/**
> + * ixgbe_refresh_fw_version - re-read flash data and update eeprom_id cache
> + * @adapter: board private structure
> + *
> + * Re-reads the NVM/flash and refreshes the cached adapter->eeprom_id string.
> + * On failure the cache is set to "unknown" so that ethtool -i never shows a
> + * stale version string after a failed reset.
> + *
> + * Return: 0 on success, negative error code on failure.
> + */
> +int ixgbe_refresh_fw_version(struct ixgbe_adapter *adapter)
>  {
>  	struct ixgbe_hw *hw = &adapter->hw;
> +	int err;
> +
> +	err = ixgbe_get_flash_data(hw);
> +	if (err) {
> +		strscpy(adapter->eeprom_id, "unknown",
> +			sizeof(adapter->eeprom_id));

nit: I think you can omit the size argument to strscpy
     because eeprom_id is an array.

> +		return err;
> +	}
>  
> -	ixgbe_get_flash_data(hw);
>  	ixgbe_set_fw_version_e610(adapter);
> +	return 0;
>  }
>  
>  static void ixgbe_get_drvinfo(struct net_device *netdev,

...