Make sure we use predictable DMA Channel Select configuration. Otherwise bad thins may happen. So fart this fix is not related to any know issue and was noticed by investigating other bugs. Signed-off-by: Oleksij Rempel <o.rempel@xxxxxxxxxxxxxx> --- drivers/net/designware_eqos.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/drivers/net/designware_eqos.c b/drivers/net/designware_eqos.c index 5e5c9ebe68..845f9f51ef 100644 --- a/drivers/net/designware_eqos.c +++ b/drivers/net/designware_eqos.c @@ -351,6 +351,11 @@ int eqos_set_ethaddr(struct eth_device *edev, const unsigned char *mac) memcpy(eqos->macaddr, mac, ETH_ALEN); + /* mac_hi is only partially overwritten by the following code. Part of + * this variable is DCS (DMA Channel Select). If this variable is not + * zeroed, we may get some random DMA RX channel. + */ + mac_hi = 0; /* Update the MAC address */ memcpy(&mac_hi, &mac[4], 2); memcpy(&mac_lo, &mac[0], 4); -- 2.39.2
