Re: [PATCH 4.18 74/88] drm/atomic: Use drm_drv_uses_atomic_modeset() for debugfs creation

[Greg Kroah-Hartman]

On Thu, Sep 27, 2018 at 03:26:37PM +0200, Holger Hoffstätte wrote:
> On 09/27/18 14:37, Greg Kroah-Hartman wrote:
> > On Thu, Sep 27, 2018 at 12:43:33PM +0200, Holger Hoffstätte wrote:
> > > On 09/27/18 11:03, Greg Kroah-Hartman wrote:
> > > > 4.18-stable review patch.  If anyone has any objections, please let me know.
> > > > 
> > > > ------------------
> > > > 
> > > > From: Lyude Paul <lyude@xxxxxxxxxx>
> > > > 
> > > > commit 3c499ea0c662e2f38aafbd4f516b08aab8cfa0e5 upstream.
> > > > 
> > > > As pointed out by Daniel Vetter, we should be usinng
> > > > drm_drv_uses_atomic_modeset() for determining whether or not we want to
> > > > make the debugfs nodes for atomic instead of checking DRIVER_ATOMIC, as
> > > > the former isn't an accurate representation of whether or not the driver
> > > > is actually using atomic modesetting internally (even though it might
> > > > not be exposing atomic capabilities).
> > > > 
> > > > Signed-off-by: Lyude Paul <lyude@xxxxxxxxxx>
> > > > Cc: Daniel Vetter <daniel.vetter@xxxxxxxx>
> > > > Cc: stable@xxxxxxxxxxxxxxx
> > > > Reviewed-by: Sean Paul <seanpaul@xxxxxxxxxxxx>
> > > > Link: https://patchwork.freedesktop.org/patch/msgid/20180917173733.21293-1-lyude@xxxxxxxxxx
> > > > Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> > > 
> > > This patch breaks switching the console to high resolution during boot on my
> > > workstation with a Radeon card; it worked fine with 4.18.10 and reverting it
> > > fixes the problem:
> > 
> > Is 4.19-rc5 also a problem?
> > 
> 
> No, 4.19-rc5 with the same config works fine and properly switches the
> console during boot.
> 
> Interestingly another machine with i915 chip seemed to work fine with this
> patch included (rebooted that one first), so it might well be related to
> different motherboard/chipset or the Radeon card (an admittedly old, but
> otherwise completely functional fanless r600).
> 
> I'll try to find more clues, but for now that's all I got.

Ok, I'll go delete this, but this implies a much deeper problem with the
code here.  No logic should ever change based on a debugfs file creation
failing or succeeding.  The error checks here are all not needed at all.
I'll work on a patch to clean it up for future kernels...

thanks,

greg k-h