On Sun, Jul 15, 2018 at 3:16 AM André Przywara <andre.przywara@xxxxxxx> wrote: > > On 07/14/2018 03:18 PM, Qiang Yu wrote: > > Hi, > > > On Sat, Jul 14, 2018 at 8:07 PM André Przywara <andre.przywara@xxxxxxx> wrote: > >> > >> On 07/14/2018 02:14 AM, Qiang Yu wrote: > >> > >> Hi, > >> > >>> Thanks for your info. What a surprise that exist such a SoC. That > >>> means I have to > >>> judge if it's a 64bit CPU in other way than just check ARM64 config. > >> > >> Yeah, you should do anyways. Actually you should try to avoid those > >> explicit checks in the first place. Drivers shouldn't need to care about > >> the "bit size" of the CPU and Linux provides many ways to automatically > >> cope with that, with types likes phys_addr_t for instance. > >> Quickly grep-ing I find "need_dma32" in lima_ttm.c:lima_ttm_init(), is > >> that the only place you need to check? > > Yes, mali GPU can only use physical mem within 32bit address space. > > Ah, thanks, I was wondering about that. > > > So I check ARM64 and ARM_LPAE for possible more than 32bit > > address space to pass in “need_dma32”. > > Mmh, but this is not how I understand this parameter. To me it looks > like it is a property of the device (GPU and bus), but not the CPU. > So if Mali 4xx can only do 32-bit DMA, then this parameter should > *always* be true, regardless of whether the CPU can address more than > 4GB. If the system is restricted to 32-bit anyways, it should not hurt > to have it set to true, even though it is not needed in this case. > > So you can drop this check altogether and just always pass "true". This parameter is for TTM to alloc page with GFP_DMA32 flag, I'd love to always set it if GFP_DMA32 won't cause any trouble on 32bit CPUs. Regards, Qiang > > Cheers, > Andre. > > >>> On Sat, Jun 16, 2018 at 1:23 AM Andre Przywara <andre.przywara@xxxxxxx> wrote: > >>>> > >>>> On 05/23/2018 17:16, Marek Vasut wrote: > >>>>> On 05/18/2018 11:28 AM, Qiang Yu wrote: > >>>>>> From: Lima Project Developers <dri-devel@xxxxxxxxxxxxxxxxxxxxx> > >>>>>> > >>>>>> Signed-off-by: Qiang Yu <yuq825@xxxxxxxxx> > >>>>>> Signed-off-by: Neil Armstrong <narmstrong@xxxxxxxxxxxx> > >>>>>> Signed-off-by: Simon Shields <simon@xxxxxxxxxxxxx> > >>>>>> Signed-off-by: Heiko Stuebner <heiko@xxxxxxxxx> > >>>>>> --- > >>>>>> drivers/gpu/drm/Kconfig | 2 ++ > >>>>>> drivers/gpu/drm/Makefile | 1 + > >>>>>> drivers/gpu/drm/lima/Kconfig | 9 +++++++++ > >>>>>> drivers/gpu/drm/lima/Makefile | 19 +++++++++++++++++++ > >>>>>> 4 files changed, 31 insertions(+) > >>>>>> create mode 100644 drivers/gpu/drm/lima/Kconfig > >>>>>> create mode 100644 drivers/gpu/drm/lima/Makefile > >>>>>> > >>>>>> diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig > >>>>>> index deeefa7a1773..f00d529ee034 100644 > >>>>>> --- a/drivers/gpu/drm/Kconfig > >>>>>> +++ b/drivers/gpu/drm/Kconfig > >>>>>> @@ -289,6 +289,8 @@ source "drivers/gpu/drm/pl111/Kconfig" > >>>>>> > >>>>>> source "drivers/gpu/drm/tve200/Kconfig" > >>>>>> > >>>>>> +source "drivers/gpu/drm/lima/Kconfig" > >>>>>> + > >>>>>> # Keep legacy drivers last > >>>>>> > >>>>>> menuconfig DRM_LEGACY > >>>>>> diff --git a/drivers/gpu/drm/Makefile b/drivers/gpu/drm/Makefile > >>>>>> index 50093ff4479b..aba686e41d6b 100644 > >>>>>> --- a/drivers/gpu/drm/Makefile > >>>>>> +++ b/drivers/gpu/drm/Makefile > >>>>>> @@ -103,3 +103,4 @@ obj-$(CONFIG_DRM_MXSFB) += mxsfb/ > >>>>>> obj-$(CONFIG_DRM_TINYDRM) += tinydrm/ > >>>>>> obj-$(CONFIG_DRM_PL111) += pl111/ > >>>>>> obj-$(CONFIG_DRM_TVE200) += tve200/ > >>>>>> +obj-$(CONFIG_DRM_LIMA) += lima/ > >>>>>> diff --git a/drivers/gpu/drm/lima/Kconfig b/drivers/gpu/drm/lima/Kconfig > >>>>>> new file mode 100644 > >>>>>> index 000000000000..4ce9ac2e8204 > >>>>>> --- /dev/null > >>>>>> +++ b/drivers/gpu/drm/lima/Kconfig > >>>>>> @@ -0,0 +1,9 @@ > >>>>>> + > >>>>>> +config DRM_LIMA > >>>>>> + tristate "LIMA (DRM support for ARM Mali 400/450 GPU)" > >>>>>> + depends on DRM > >>>>>> + depends on ARCH_SUNXI || ARCH_ROCKCHIP || ARCH_EXYNOS || ARCH_MESON > >>>>> > >>>>> You can add ARCH_ZYNQMP here too , it has Mali 400 MP2. > >>>> > >>>> Well, as Qiang Yu already figured, it seems much smarter to not enumerate > >>>> every possible platform here. > >>>> More than that, the Kconfig depends should be strictly technical. There is > >>>> nothing in this driver which is ARM specific, in fact I managed to compile > >>>> it for x86-64 as well (with some small fix in a random header file). > >>>> In fact there are x86-64 based SoCs pairing Intel Atom cores with a Mali GPUs: > >>>> https://en.wikipedia.org/wiki/Rockchip#Tablet_processors_with_integrated_modem > >>>> > >>>> So you can get rid of this whole line at all, meaning you don't even need > >>>> the "depends on ARM || ARM64 || COMPILE_TEST" you have in your gitlab repo. > >>>> > >>>> Cheers, > >>>> Andre. > >>>> > >>>>> > >>>>> -- > >>>>> Best regards, > >>>>> Marek Vasut > >>>>> _______________________________________________ > >>>>> dri-devel mailing list > >>>>> dri-devel@xxxxxxxxxxxxxxxxxxxxx > >>>>> https://lists.freedesktop.org/mailman/listinfo/dri-devel > >> > _______________________________________________ dri-devel mailing list dri-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/dri-devel
