NACK/Cmnt: [PATCH 0/1] [SRU][X/B] i40e PF reset due to incorrect MDD event

[Tim Gardner]

On 3/9/21 2:12 PM, Heitor Alves de Siqueira wrote:
> Hi all,
> 
> I've updated the corresponding LP bug with more detailed testing for these
> patches. In summary, I cooked up a set of kprobes that force the firmware to
> raise MDDs, which allowed some more interesting tests that exercise the new code
> path. I'll send a v2 for these with a new test procedure which is why I'm
> NACKing the current patches for now.
> 
> Tim, I've also searched for the changes you mentioned in upstream master,
> linux-next and net-next but didn't find any differences from the behavior
> introduced by this patch. I tried looking for any 'fixes:' tags and went through
> the code in i40e_handle_mdd_event(), but it seems to be the same since v5.2
> (when this patch was introduced). The only related change that doesn't seem to
> be in our kernels is a7da7f16267b, which reduces the allowed number of MDDs on
> the VFs to 1 (previously 3). That patch doesn't seem related to the PF resets,
> so maybe I missed any changes in related functions? Do you have the commit ids
> for the 5.12-rc changes you mentioned, so that we can evaluate whether we should
> pull in any additional patches together with this one?
> 

Nice work! Perhaps I was only seeing possible regressions with no basis 
in reality. After reading your bug report I'm a lot more comfortable 
with the side effects of this patch now that you are able to exercise 
the new behaviors.

rtg
-----------
Tim Gardner
Canonical, Inc