On 02/03/12 08:44, Paulo Zanoni wrote: > 2012/2/3 Chris Wilson<chris at chris-wilson.co.uk>: >> I don't this should an installable tool for use by end-users. Direct >> register access from userspace to registers that the driver owns? I'm >> wary of installing even root-only tools that can screw us over in >> unpredictable ways. >> -Chris > > My idea is to write a patch that adds kernel peroperties to solve > this, but while people still don't have kernel 3.4, the tool would be > useful. > > If adding this tool to i-g-t is not a good idea, that's fine too: I > still plan to write the kernel patches, so people wanting overscan > compensation will have to use the kernel patch instead of the tool. I didn't read that email from Chris as don't add it. I read it as, don't have it installed if they do make install or something. So I think he should clarify himself. If this benefits end users now - I'm all for it. However it should somehow enforce that it is running as root. The most reusable way is to use the register_access_init functions designed for forcewake. Ben
