On Sparx5 and lan969x, a received frame whose source port does not map
to an active port is never consumed from the FDMA RX ring: the NAPI poll
stops on it, and finds the same DB again on every following poll.
Patch 1 fixes Sparx5, where the same skb grows on every poll until
skb_over_panic().
Patch 2 fixes lan969x, where RX stops on all ports.
The issue was reported by Zhang Hewei, and these patches replace his
series [1]. Tested on Sparx5 and lan969x.
[1] https://lore.kernel.org/netdev/tencent_D2451CD0D7B4B812CEBADB56DC722BC33406@qq.com/ (local)
Signed-off-by: Daniel Machon <daniel.machon@microchip.com>
---
Daniel Machon (2):
net: sparx5: fix RX ring stall on frames from inactive ports
net: lan969x: fix RX ring stall on frames from inactive ports
drivers/net/ethernet/microchip/sparx5/lan969x/lan969x_fdma.c | 11 ++++++-----
drivers/net/ethernet/microchip/sparx5/sparx5_fdma.c | 12 +++++++++---
2 files changed, 15 insertions(+), 8 deletions(-)
---
base-commit: aa40afa831ebfe45e40a23721711059f2ac584d4
change-id: 20261116-sparx5-rx-inactive-port-df606d3f9b1a
Best regards,
--
Daniel Machon [off-list ref]