On Mon, Nov 30, 2020 at 12:02:15AM -0600, Charlie Burnett wrote: > Doesn’t DRM_INFO output even if DRMDEBUG is enabled? I thought DRM_DEBUG > only output when debug’s enabled while DRM_INFO is pretty much just a > wrapper for printk?
Currently DRM_INFO will call into printk which returns early for KERN_INFO messages if DRMDEBUG is not set, see printk in drm_linux.c. > > Either way, I found the call writing to the GPU register it doesn’t like > earlier this weekend, but haven’t figured out what exactly it’s having > issues with. Have a busy week or two but I’ll try to come back to it then! > Just wanted to send something in case anyone else had a similar issue. Including the error message you saw would be helpful. > > On Sun, Nov 29, 2020 at 11:25 PM Jonathan Gray <[email protected]> wrote: > > > On Sun, Nov 29, 2020 at 10:17:22PM -0600, Charlie Burnett wrote: > > > Howdy all, > > > For reasons that are beyond me, when printf is called in amdgpu_device.c > > to > > > print the CU info, it gives me a psp firmware load failure on a Radeon > > VII > > > (Vega 20) gpu. Switching the printf statement to a DRM_INFO statement as > > > used in the rest of amdgpu seems to fix it though. > > > > Find what is sensitive to the delay. > > > > Hiding this printf under DRMDEBUG makes no sense. You should be able to > > load kernels with and without DRMDEBUG. > > > > > > > > ok? > > > > > > diff --git sys/dev/pci/drm/amd/amdgpu/amdgpu_device.c > > > sys/dev/pci/drm/amd/amdgpu/amdgpu_device.c > > > index 45eff483e86..fba5a7caf23 100644 > > > --- sys/dev/pci/drm/amd/amdgpu/amdgpu_device.c > > > +++ sys/dev/pci/drm/amd/amdgpu/amdgpu_device.c > > > @@ -3210,7 +3210,7 @@ fence_driver_init: > > > default: > > > chip_name = amdgpu_asic_name[adev->asic_type]; > > > } > > > - printf("%s: %s %d CU rev 0x%02x\n", adev->self.dv_xname, > > > + DRM_INFO( "%s: %s %d CU rev 0x%02x\n", adev->self.dv_xname, > > > chip_name, adev->gfx.cu_info.number, adev->rev_id); > > > } > > > #endif > > > > > > > >
