On 14/02/17 12:52 PM, Tom St Denis wrote: > On 14/02/17 12:16 PM, Tom St Denis wrote: >> On 14/02/17 12:01 PM, Samuel Pitoiset wrote: >>> No worries. I thought that check was enough. Anyway, writing code >>> without the hardware should be avoided. :) >>> >>> Can you try the thing suggested by Alex? Because I will need to fix up >>> the DRM ioctl codepath as well. >> >> Sure, I'm building a module with a amdgpu_dpm check near the top of the >> debugfs read function. > > It seems returning an error code causes it to do "bad things" on read > > [root at kaveri linux]# git diff > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c > b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c > index 80821b4..d63c443 100644 > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c > @@ -3198,6 +3198,9 @@ static ssize_t amdgpu_debugfs_sensor_read(struct > file *f, char __user *buf, > if (size & 3 || *pos & 0x3) > return -EINVAL; > > + if (amdgpu_dpm == 0) > + return -EINVAL; > + > /* convert offset to sensor number */ > idx = *pos >> 2; > > That's what I applied on top of Samuel's three patches. > > With DPM enabled it works fine obviously, with it disabled the entire > system hangs. I think probably because the VFS spins trying to read but > never finishes... Nevermind, the issue is umr reading UVD registers when dpm=0 on KV. Without sampling UVD the error works properly. So add those three lines to your patch (and similar in the ioctl) you should be fine. Cheers, Tom
