On Mon, Jan 20, 2020 at 09:17:35AM +0000, Hans-Christian Egtvedt (hegtvedt) wrote:
Hello,
I am seeing a regression in the macvlan kernel driver after Linux stable release 4.4.209, bisecting identifies commit https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=l...
Noticed it too last week (on net-next), but Eric already fixed it: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...
I assume the patch will find its way to 4.4.y soon now that it is in mainline.
There seems to be a history behind this, and I do not have the full overview of the intention behind the change.
What I see on my target, Aarch64 CPU, is that this patch moves the eth pointer in macvlan_broadcast() function some bytes. This will cause everything within the ethhdr struct to be wrong AFAICT.
An example: Original code "eth = eth_hdr(skb)" eth = ffffffc007a1b002 New code "eth = skb_eth_hdr(skb)" eth = ffffffc007a1b010
Let me know if I can assist in any way.
-- Best regards, Hans-Christian Noren Egtvedt