2014-06-24 13:30 GMT-07:00 Daniel Gimpelevich <dan...@gimpelevich.san-francisco.ca.us>: > On Tue, 2014-06-24 at 12:38 -0700, Florian Fainelli wrote: >> I think AR231x has none of those, it's an End of Life platform, the >> code base has been mostly static and well know for a while, so I would >> argue that Device Tree should not be made a requirement here as it >> will just delay Sergey's upstreaming effort even more. >> > Very valuable input. Still, there is no way for software to determine > which AR231x board it's running on, and they all have different uses of > GPIO, plus the AR2317 watchdog operates completely differently from the > AR2315 one. What solution do you propose? Some earlier discussion: > http://patchwork.openwrt.org/patch/4351/
For GPIOs, since the way they are used most likely varies on a per-board basis, we could probably come up with the same mechanism as used on ath79 where we end-up patching the kernel command-line to insert a MIPS machine id for instance. For the watchdog driver, if we have access to a revision register we can read at runtime, then we could use a separate platform driver name (e.g: ar2315-wdt vs ar2317-wdt) that would lead to either two separate drivers to get registered, or have different code-paths being used in the same ar231x driver. In case we do not have that revision register, we can leverage solution 1) for GPIOs. -- Florian _______________________________________________ openwrt-devel mailing list openwrt-devel@lists.openwrt.org https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel