Dear all, I have recently switched from backfire to trunk. This forced a change from madwifi to ath5k for atheros platforms. I also compile and test firmware on ar71xx, ramips and brcm47xx.
I notice that on ar71xx and ramips the actual router/board is identified in /proc/cpuinfo. Luci in trunk is using this file almost exclusively to determine the router information (refer to http://luci.subsignal.org/trac/browser/luci/trunk/libs/sys/luasrc/sys.lua#170) Unfortunately on the atheros platform and brcm47xx the field 'machine' on /proc/cpuinfo is not set. The system then uses system type, which is not defining the router, but give a generic processor name such as BCM47XX (see https://dev.openwrt.org/attachment/ticket/7604/openwrt-cpuinfo) or AR2315 (http://www.zoobab.com/engenius-ecb3500). Interestingly when madwifi was used by atheros, the information about the board was not taken from /proc/cpuinfo, but from /proc/sys/dev/$device/dev_name. Ath5k does not generate that file, but obviously if madwifi had a way of getting to the information without /proc/cpuinfo, so OpenWrt must be able to as well, right? What I am certain of is that the devices I tested with a serial such as Fon+, Nano2, Bullet2, Pico2 and Pico2HP are all aware of their identity during the Redboot boot process; they all say something like 'Board: Ubiquiti NanoStation 2' so the information what the board actually is is certainly known to the device and it should be possible to bring it to the fore to userland and ultimately to Luci. The current situation with atheros and ath5k is certainly a regression fro madwifi as far as device identification is concerned. Any suggestions how that can be achieved? Cheers Hanno _______________________________________________ openwrt-devel mailing list openwrt-devel@lists.openwrt.org https://lists.openwrt.org/mailman/listinfo/openwrt-devel
