On Tue, 23 Oct 2018, Rodrigo Vivi <rodrigo.vivi@xxxxxxxxx> wrote: > But I thought while doing this I could consolidade it along with all > the other has_feature cases. > > I believe we should either have everything as info.has_feature or everything > as has_feature(). > > for instance if we end up ever having 2 platforms of same gen where > one has_sagv and the other doesn't we would have that in the platform > definition while making us to define another codename and add it here > or even worse if we don't have a codename available like CNL_WITH_PORT_F :/ If that is to be in device info, it doesn't require an extra codename, it requires an extra device info with the flag. This ties to the goal of making dev_priv->info a pointer to the static const data in i915_pci.c i.e. making ->info truly const. There's three categories of info: 1) immutable device properties 2) properties set once during probe, immutable afterwards 3) runtime Currently we more or less happily conflate these, along with some module parameters too. The mkwrite_device_info() use has profilerated much wider than it was ever intended; we need to nuke that. We also have HAS_FOO() and IS_FOO() macros that do checks on pci id or gen or platform or a combination of them. It's a mess, and it's not getting better without conscious effort. BR, Jani. -- Jani Nikula, Intel Open Source Graphics Center _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
