Date: Mon, 8 Aug 2016 19:07:10 -0700 From: Adrian Chadd <adrian@freebsd.org> To: "freebsd-wireless@freebsd.org" <freebsd-wireless@freebsd.org> Subject: help tracking down carambola 2 ethernet link stability Message-ID: <CAJ-Vmo=8EwEufb8D-JLkWoX4K2pq-hgb2aO2be8bFQrYaNwGPQ@mail.gmail.com>
next in thread | raw e-mail | index | archive | help
hiya, FreeBSD-HEAD on the carambola 2 has an interesting effect - the switch port temporarily flaps any time the wifi interface is brought up or down. I've narrowed it to the "tuning_caps" field in the EEPROM. It's overriding some PLL capacitor tuning values, causing the carambola 2 to be more frequency-correct (ie, it solved all of the "why can it only see some but not other devices on 2GHz" issues) but it seems to have introduced this weird behaviour. If I disable that code, then the link doesn't flap at all. If you use this in AP mode then it shows up as the ethernet link flapping whenever the chip is reset (eg stuck beacon.) in STA mode it shows up as the ethernet link flapping each time the channel is changed. The ath9k driver in Linux and openwrt (both carambola 2 and the upstream openwrt) have some patches to make "USB more stable" by disabling some PHY lock bits, but it doesn't make a difference here. I'm wondering if there are other reset and/or PHY PLL bits that we just do differently from the Atheros HAL that ath9k "does better" on this platform. So, I'm asking for some help in tracking this stuff down. I'd really like to see this completely stable on the carambola 2 as it's a cheaply available AR9331 part and it'd be a shame to not support it, but I'm out of spare clock cycles to chase it all down. Thanks, -adrian
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAJ-Vmo=8EwEufb8D-JLkWoX4K2pq-hgb2aO2be8bFQrYaNwGPQ>