Since NAPI works by shutting down event interrupts when theres
work and turning them on when theres none, the net driver must
make sure that interrupts are disabled when it reschedules polling.
By calling napi_reschedule, the driver switches to polling mode,
therefor there should be no interrupt interference.
Any received packets will be handled in nps_enet_poll by polling the HW
indication of received packet until all packets are handled.
Re: [PATCH 1/1] net: nps_enet: Disable interrupts before napi reschedule
On Thu, 2016-05-26 at 15:00 +-0300, Elad Kanfi wrote:
+AD4- From: Elad Kanfi +ADw-eladkan+AEA-mellanox.com+AD4-
+AD4- Since NAPI works by shutting down event interrupts when theres
+AD4- work and turning them on when theres none, the net driver must
+AD4- make sure that interrupts are disabled when it reschedules polling.
+AD4- By calling napi+AF8-reschedule, the driver switches to polling mode,
+AD4- therefor there should be no interrupt interference.
+AD4- Any received packets will be handled in nps+AF8-enet+AF8-poll by polling the HW
+AD4- indication of received packet until all packets are handled.
+AD4- Signed-off-by: Elad Kanfi +ADw-eladkan+AEA-mellanox.com+AD4-
+AD4- Acked-by: Noam Camus +ADw-noamca+AEA-mellanox.com+AD4-
+AD4- +AKA-drivers/net/ethernet/ezchip/nps+AF8-enet.c +AHwAoACgAKAAoA-4 +-+-+--
+AD4- +AKA-1 files changed, 3 insertions(+-), 1 deletions(-)
+AD4- diff --git a/drivers/net/ethernet/ezchip/nps+AF8-enet.c b/drivers/net/ethernet/ezchip/nps+AF8-enet.c
+AD4- index 085f912..06f0317 100644
+AD4- --- a/drivers/net/ethernet/ezchip/nps+AF8-enet.c
+AD4- +-+-+- b/drivers/net/ethernet/ezchip/nps+AF8-enet.c
+AD4- +AEAAQA- -205,8 +-205,10 +AEAAQA- static int nps+AF8-enet+AF8-poll(struct napi+AF8-struct +ACo-napi, int budget)
+AD4- +AKA- +AKAAKg- re-adding ourselves to the poll list.
+AD4- +AKA- +AKAAKg-/
+AD4- - if (priv-+AD4-tx+AF8-skb +ACYAJg- +ACE-tx+AF8-ctrl+AF8-ct)
+AD4- +- if (priv-+AD4-tx+AF8-skb +ACYAJg- +ACE-tx+AF8-ctrl+AF8-ct) +AHs-
+AD4- +- nps+AF8-enet+AF8-reg+AF8-set(priv, NPS+AF8-ENET+AF8-REG+AF8-BUF+AF8-INT+AF8-ENABLE, 0)+ADs-
+AD4- +AKA- napi+AF8-reschedule(napi)+ADs-
+AD4- +- +AH0-
+AD4- +AKA- +AH0-
+AD4- +AKA- return work+AF8-done+ADs-
We just bumped into the same problem (data exchange hangs on the very first +ACI-ping+ACI-)
with released Linux v4.6 and linux-next on our nSIM OSCI virtual platform.
I believe it was commit+AKA-05c00d82f4d1 (+ACI-net: nps+AF8-enet: bug fix - handle lost tx interrupts+ACI-)
that introduced the problem. At least reverting it I got networking working.
And indeed that patch fixes mentioned issue.
In other words...
P.S. Given my observation is correct please add following to your commit
message if you ever do a respin:
Fixes: 05c00d82f4d1 (+ACI-net: nps+AF8-enet: bug fix - handle lost tx interrupts+ACI-)