On Sun, 2017-02-26 at 09:34 -0800, Eric Dumazet wrote:

I do not believe this bug is mlx4 specific.

Anything doing the following while hard irq were not masked :

local_bh_disable();
napi_reschedule(&priv->rx_cq[ring]->napi);
local_bh_enable();

Like in mlx4_en_recover_from_oom()

Can trigger the issue really.

Unfortunately I do not see how core layer can handle this.
Only the driver hard irq could possibly know that it could not grab
NAPI_STATE_SCHED

Actually we could use an additional bit for that, that the driver would
set even if NAPI_STATE_SCHED could not be grabbed.

Let me try something.