This bug report relates to a device using the AR8216 switch chip and a trunk version where the additional phy reset in 3.14 was still enabled.
Seems like there was an issue with autoneg. This is not fully clear to me as after the first phy reset in the kernel phy code we do the hw_init in the driver and set BMCR_RESET | BMCR_ANENABLE. This seems to re-establish autoneg support for other members of the AR8xxx switch chip family but not for AR8216. For AR8216 the BMCR_ANENABLE flag seems to be ignored. However I can't test this as I have no 8216-based device and no datasheet. Maybe to make AR8216 happy we have to enable autoneg explicitely like this before resetting? mdiobus_write(bus, i, MII_BMCR, BMCR_ANENABLE); mdiobus_write(bus, i, MII_BMCR, BMCR_RESET | BMCR_ANENABLE); For the other AR8xxx switches this isn't needed but doesn't disturb. Anybody with a 8216-based device who can test this? However changeset 43410 disabled the additional phy reset. This makes 8216-based devices work properly again so there's no uregent need to check this. Heiner
_______________________________________________ openwrt-devel mailing list openwrt-devel@lists.openwrt.org https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel
