On Fri, Dec 13, 2013 at 8:00 PM, Sujith Manoharan <sujith@xxxxxxxxxxx> wrote: > Dave Taht wrote: >> OK, I couldn't help myself but boot up that release. Wet paint! It >> successfully brought up >> the 5ghz radio, but did not manage to assign an ip address to it >> (netifd bug?) and failed on the 2ghz radio utterly. >> >> trying to restart it manually fails to bring up the 5ghz radio as well. >> Here's an strace of that. >> >> http://snapon.lab.bufferbloat.net/~d/hostapd.strace.txt >> >> I don't see it beacon, either. >> >> Now, I don't have a grip on what started happening two releases back >> (I was out of town) but I figure it is perhaps more relevant than >> chasing the DMA tx thing. And ENOTIME for me on this til sunday. I >> will revert this patch and bisect backwards. > > I am not sure how the patch would break things. > > Booting OpenWrt trunk (with the patch) on an AP96 reference board seems > to work fine: http://pastebin.com/3rPSfuad > > Sujith We appear to have a deeper problem. I reverted your patch, moved my build back to 3.10.21 to match openwrt, felix reverted some stuff in netifd... ... and on first boot, somehow, all the wireless interfaces come up. After a reboot, most don't. doing openwrt's wifi up is also interesting with all sorts of failures trying to bring up different interfaces. So I suspect netifd has a race still. Afer we get that sorted, we can move onto trying to poke into the DMA tx error, or since it seems by hammering on it I can get an interface up I can move forward a bit, in parallel. -- Dave Täht Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
