[Hyper-V] SR-IOV functionality regression in 16.04 HWE proposed

[Chris Valean]

Hello,

In 1650059 we requested the rebase to the 4.9 kernel LIS commits.

Also, from comments #7 and #8 we confirmed that the SR-IOV functionality was working as expected.
This was in the test kernel build 4.8.0-33.35.

Now, while testing the HWE proposed kernel for Xenial - version 4.8.0-35, we seem to have a regression for SR-IOV.
Issues start from the point where the Intel driver is not loaded automatically, then even with manually inserting it, the SR-IOV VF interfaces are not detected.

We need to determine if between the test kernel and the proposed one if there have been any other related LIS / SR-IOV commits, or what related ones could have caused this.

Thank you!

[Joshua R. Poulson (jrp)]

For now we will test 4.4.0-63 (lts-xenial) which has the 4.9 rebase of Hyper-V, but we should re-examine the HWE rolling release when it catches up with the other requests.

[Joseph Salisbury (jsalisbury)]

Does the bug also exist in 4.4.0-63.84 or does it just exist in 4.8.0-35?

Also, 4.8.0-38 is now in Yakkety -proposed. There are a few SR-IOV commits between -35 and -38. Can you see if -38 still exhibits the bug?

If -38 still has the bug, we can bisect between 4.8.0-34 and 4.8.0-35. That is if -34 did not have the bug. If it does, we would have to test some earlier kernels to know the starting version for the bisect.

[Chris Valean (cvalean)]

Hi Joe,

The bug exists only in the hwe proposed kernel 4.8.0-33.35.
SR-IOV does indeed work on the latest yakkety proposed - 4.8.0.38.49.

On 16.04.1 proposed 4.4.0.63.67 (lts-xenial) we also successfully tested SR-IOV.

[Adrian Suhov (asuhov)]

Hi,

I tested HWE 4.8.0-53-generic kernel and initial issue is solved

[Joseph Salisbury (jsalisbury)]

@chris, can you confirm this bug is now resolved?

[Chris Valean (cvalean)]

Adrian's info are correct, we can close this issue.