Quoting Chris Wilson (2018-05-03 18:35:48) > Quoting Paulo Zanoni (2018-05-03 16:24:47) > > Em Qui, 2018-05-03 às 12:59 +0300, Joonas Lahtinen escreveu: > > > Please split the patch here and add a respective Fixes: tag to when > > > base Icelake support was introduced. Lacking this portion when > > > running ICL will > > > cause random memory corruption so it's important to have this landed > > > early. > > > > Icelake is still hidden behind the alpha_support flag, so no point in > > backporting for i915.ko specifically. If we're talking about > > backporting just because of the memory reservation without graphics, > > how far back should we go? > > The danger for the reserved portion is that it is used by the bios/uefi > scanout and vgacon. So if the gfx reserved portion is absent from the > 820 memmap (buggy bios or whatever), then it may be reused by the system > even though it's been written to by the console (and unrelated to > i915.ko). That's where the danger comes from, it's happened before so we > fear it may happen again. Yep, This same discussion was had for Cannonlake; alpha_support protection is not enough to justify system level memory corruption which has nothing to do with i915 module but all to do with enforcing e820 memory maps not to include with the BIOS framebuffer as regular memory. So this step really should be a separate patch, and should be one of the first patches to upstream. Regards, Joonas _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
