[Devel] [PATCH rh7 2/2] ixgbe: don't check firmware errors

[Kirill Tkhai]

On 10.01.2020 18:56, Konstantin Khorenko wrote:
> On 01/10/2020 06:45 PM, Kirill Tkhai wrote:
>> On 10.01.2020 18:31, Konstantin Khorenko wrote:
>>> There is a check in new ixgbe version driver in RHEL7.7
>>> which verify the firmware error register value for correctness.
>>>
>>> Apparently seems the check is not fully correct itself
>>> causing kernel to spoil logs (every second).
>>>
>>> The driver in RHEL7.6 does not have that check.
>>> The driver from vendor site does not have that check.
>>> https://downloadmirror.intel.com/14687/eng/ixgbe-5.6.5.tar.gz
>>>
>>> So let's drop the check as well in Virtuozzo kernels.
>>>
>>> https://bugs.centos.org/view.php?id=16495
>>> https://forum.proxmox.com/threads/pve-6-0-7-ixgbe-firmware-errors.58592/
>>> https://jira.sw.ru/browse/PSBM-100722
>>>
>>> Signed-off-by: Konstantin Khorenko <khorenko at virtuozzo.com>
>>> ---
>>>  drivers/net/ethernet/intel/ixgbe/ixgbe_main.c | 9 ---------
>>>  1 file changed, 9 deletions(-)
>>>
>>> diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
>>> index eef479f762f93..b39a1b477b246 100644
>>> --- a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
>>> +++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
>>> @@ -7913,15 +7913,6 @@ static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
>>>  static bool ixgbe_check_fw_error(struct ixgbe_adapter *adapter)
>>>  {
>>>  	struct ixgbe_hw *hw = &adapter->hw;
>>> -	u32 fwsm;
>>> -
>>> -	/* read fwsm.ext_err_ind register and log errors */
>>> -	fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
>>> -
>>> -	if (fwsm & IXGBE_FWSM_EXT_ERR_IND_MASK ||
>>> -	    !(fwsm & IXGBE_FWSM_FW_VAL_BIT))
>>> -		e_dev_warn("Warning firmware error detected FWSM: 0x%08X\n",
>>> -			   fwsm);
>>
>> Maybe we warn here only once?
> 
> i thought about it but did not find any reason why do we need this warning (even once).
> 1) we know it is triggered both on our internal hardware and there are complains from others in the internet
> 2) there were no this check and warning previously
> 3) there is no such check in the vendor driver version
> 4) network works fine with drivers without this warning (both old and new)
> 
> i don't see how this warning can be useful for us,
> moreother it will threaten user while they can do nothing with it.
> 
> So, i've chosen to just remove it completely.

Sounds reasonable.

Acked-by: Kirill Tkhai <ktkhai at virtuozzo.com>