From mboxrd@z Thu Jan 1 00:00:00 1970 From: Stephen Warren Subject: Re: [RFC v2 6/8] gpu: drm: tegra: Remove redundant host1x Date: Fri, 21 Dec 2012 14:19:23 -0700 Message-ID: <50D4D25B.7030506@wwwdotorg.org> References: <50BF345A.8050201@nvidia.com> <20121205120429.GA29943@avionic-0098.adnet.avionic-design.de> <50C5CAB5.3040000@nvidia.com> <20121212160829.GA30278@avionic-0098.adnet.avionic-design.de> <50C99677.6090306@nvidia.com> <20121213085750.GA14740@avionic-0098.adnet.avionic-design.de> <50CA175F.60002@wwwdotorg.org> <50CAC2AC.1010704@nvidia.com> <50CB5205.1030303@wwwdotorg.org> <50CB850F.9090704@nvidia.com> <20121216121603.GA31780@avionic-0098.adnet.avionic-design.de> <50D2D792.1050401@nvidia.com> <50D34775.5010606@wwwdotorg.org> <50D42486.7080901@nvidia.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <50D42486.7080901-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org> Sender: linux-tegra-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: Arto Merilainen Cc: Terje Bergstrom , Thierry Reding , "linux-tegra-u79uwXL29TY76Z2rM5mHXA@public.gmane.org" , "dri-devel-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org" , "linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org" List-Id: dri-devel@lists.freedesktop.org On 12/21/2012 01:57 AM, Arto Merilainen wrote: > On 12/20/2012 07:14 PM, Stephen Warren wrote: >> >> What's wrong with just having each device ask the host1x (its parent) >> for a pointer to the (dummy) tegradrm device. That seems extremely >> > > We are talking about creating a dummy device because: > - we need to give something for drm_platform_init(), > - drm related data should be stored somewhere, Yes to those too, I believe. > - some data must be passed to all driver probe() functions. This is not > hw-related data, but just some lists that are used to ensure that all > drivers have been initialised before calling drm_platform_init(). I haven't really thought through /when/ host1x would create the dummy device. I guess if tegradrm really must initialize after all the devices that it uses (rather than using something like deferred probe) then the above may be true. > All these issues are purely tegra-drm related and solving them elsewhere > (in host1x) would be just wrong! host1x would not even use the dummy > device for anything so creating it there seems bizarre. I see the situation more like: * There's host1x hardware. * There's a low-level driver just for host1x itself; the host1x driver. * There's a high-level driver for the entire host1x complex of devices. That is tegradrm. There may be more high-level drivers in the future (e.g. v4l2 camera driver if it needs to aggregate a bunch of host1x sub-devices liek tegradrm does). * The presence of the host1x DT node logically implies that both the two drivers in the previous two points should be instantiated. * Linux instantiates a single device per DT node. * So, it's host1x's responsibility to instantiate the other device(s). I think it's reasonable for the host1x driver to know exactly what features the host1x HW complex supports; raw host1x access being one, and the higher level concept of a tegradrm being another. So, it's reasonable for host1x to trigger the instantiation of tegradrm. * If DRM core didn't stomp on the device object's drvdata (or whichever field it is), I would recommend not creating a dummy device at all, but rather having the host1x driver directly implement multiple driver interfaces. There are many examples like this already in the kernel, e.g. combined GPIO+IRQ driver, combined GPIO+IRQ+pinctrl driver, etc.