[dpdk-dev] [PATCH] igb_uio: issue FLR during open and release of device file

[Ferruh Yigit]

On 7/7/2017 11:03 AM, Shijith Thotton wrote:
> On Thu, Jul 06, 2017 at 08:27:17PM +0300, Gregory Etelson wrote:
>> I could not reproduce server crash with http://dpdk.org/dev/patchwork/patch/25267/ [1]
>> However, pci_try_reset_function() API used in that patch is not defined in RedHat-6.x Linux-2.6.32 kernels 
>> Therefore I work with http://dpdk.org/dev/patchwork/patch/25061/  patch [2].
>> [2] was successfully tested with IXGBE & I40e VFs on RH 6.x, RH 7.x Ubuntu 14.04 and SLES-11.4 
>>  
>> Regards,
>> Gregory
>>
>> On Thursday, 6 July 2017 19:41:40 IDT Ferruh Yigit wrote:
>>> On 6/12/2017 10:38 AM, Shijith Thotton wrote:
>>>> Set UIO info device file operations open and release. Call pci reset
>>>> function inside open and release to clear device state at start and end.
>>>> Copied this behaviour from vfio_pci kernel module code. With this patch,
>>>> it is not mandatory to issue FLR by PMD's during init and close.
>>>>
>>>> Bus master enable and disable are added in open and release respectively
>>>> to take care of device DMA.
>>>>
>>>> Signed-off-by: Shijith Thotton <shijith.thotton at caviumnetworks.com>
>>>
>>> This patch, and Gregory's patch [1] are very similar and main target is
>>> to leave device in a more proper state when DPDK application quits
>>> unexpectedly.
>>>
>>> Difference between two are, this one implements both .open and .release
>>> ops, and sets / clears bus master accordingly.
>>>
>>> Although main concern is .reset, I am OK to follow vfio_pci approach
>>> here, and clearing bus master on .reset can prevent unwanted DMA access.
>>>
>>> So, I am for this patch and I am testing it for a few days without a
>>> problem.
>>>
>>> But Gregory reported a crash with older version of this patch, without
>>> more detail, we should clear that first. With Gregory's Tested-by, I am
>>> OK with this patch.
>>>
>>>
>>> Gregory,
>>>
>>> Are you using your version, what are the results? And would you mind
>>> testing this patch?
>>>
>>> Thanks,
>>> ferruh
>>>
>>>
>>> [1]
>>> http://dpdk.org/dev/patchwork/patch/25061/
>>>
>>>
> 
> Hi Gregory,
> 
> Please try the following change:
> s/pci_try_reset_function/pci_reset_function/
> 
> pci_try_reset_function is same as pci_reset_function, except it returns -EAGAIN
> if unable to lock the device[1].
> 
> If everyone agrees, I can submit v2 with this change.

pci_try_reset_function() not being available in older kernel versions
seems a problem and blocking Gregory.

To move forward, I would suggest sending the v2, and we can continue
discussion based on it.

Thanks,
ferruh

> 
> 1. http://elixir.free-electrons.com/linux/latest/source/drivers/pci/pci.c#L4293
> 
> Thanks,
> Shijith
>