Re: [PATCH v3.10.y 0/4] IB/IPoIB fixes for SRIOV / cloud environment

[Or Gerlitz <ogerlitz@mellanox.com>]

On 06/08/2013 14:22, Or Gerlitz wrote:
> On 06/08/2013 12:38, Greg KH wrote:
>> On Tue, Aug 06, 2013 at 11:38:38AM +0300, Or Gerlitz wrote:
>>> Hi Greg,
>>>
>>> This series contains few fixes to the IB core, mlx4 IB driver and 
>>> IPoIB.
>>>
>>> The patches enable working properly with IPoIB devices set over SRIOV
>>> Virtual Functions probed to VMs in cloud environment when the cloud
>>> management system uses non trivial settings of IB PKEYs (Partition-Keys
>>> , sort of Ethernet vlans equivalent).
>> That really sounds like a new feature, and not a bugfix, for these 
>> devices, right?
>
>
>
> Hi Greg,
>
> well... not really
>
> 3eac103 IB/mlx4: Use default pkey when creating tunnel QPs
> ef5ed41 IB/core: Create QP1 using the pkey index which contains the 
> default pkey
>
> are point bug fixes to code in the IB core and mlx4 IB driver which 
> assumed
> that the default IB Partition-Key is located in index 0 of the PKEY 
> table,
> which isn't necessarily the case in virtualization environments.
>
> 3d790a4 IPoIB: Make sure child devices use valid/proper pkeys
>
> is bug fix to the IPoIB driver to make sure we disallow child device
> creation with invalid IB PKEY
>
> and finally
>
> c290414 IPoIB: Fix pkey change flow for virtualization environments
>
> is indeed a bit of heavier but is a fix, the text there saying
>
>> IPoIB's required behaviour w.r.t to the pkey used by the device is 
>> the following:
>>
>> - For "parent" interfaces (e.g ib0, ib1, etc) who are created
>>    automatically as a result of hot-plug events from the IB core, the
>>    driver needs to take whatever pkey vlaue it finds in index 0, and
>>    stick to that index.
>>
>> - For child interfaces (e.g ib0.8001, etc) created by admin directive,
>>    the driver needs to use and stick to the value provided during its
>>    creation.
>>
>
> came to describe the driver design and how it works when the bug 
> doesn't trigger,
> the patch itself fixes the driver to always work along this design
>
>
>> I need an agreement from the IB maintainers that this really is stable
>> material before I can accept it.
>
> sure, fair enough, just wanted to shed more light on the matter.
>

Hi Roland,

Can you comment here, please.

Or.