net: aquantia: fix rx checksum offload bits
authorDmitry Bogdanov <dmitry.bogdanov@aquantia.com>
Tue, 27 Nov 2018 14:51:17 +0000 (14:51 +0000)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 21 Dec 2018 13:15:16 +0000 (14:15 +0100)
[ Upstream commit 37c4b91f955fdd5f4ad771956b97d35f1321098e ]

The last set of csum offload fixes had a leak:

Checksum enabled status bits from rx descriptor were incorrectly
interpreted. Consequently all the other valid logic worked on zero bits.
That caused rx checksum offloads never to trigger.

Tested by dumping rx descriptors and validating resulting csum_level.

Reported-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Dmitry Bogdanov <dmitry.bogdanov@aquantia.com>
Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Fixes: ad703c2b9127f ("net: aquantia: invalid checksumm offload implementation")
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_b0.c

index 88705de..56363ff 100644 (file)
@@ -667,7 +667,7 @@ static int hw_atl_b0_hw_ring_rx_receive(struct aq_hw_s *self,
 
                rx_stat = (0x0000003CU & rxd_wb->status) >> 2;
 
-               is_rx_check_sum_enabled = (rxd_wb->type) & (0x3U << 19);
+               is_rx_check_sum_enabled = (rxd_wb->type >> 19) & 0x3U;
 
                pkt_type = 0xFFU & (rxd_wb->type >> 4);