[Intel-gfx] [PATCH] drm/i915: fix suspend/resume for GENs w/o runtime PM support

[Imre Deak]

On Tue, 2014-08-26 at 09:45 +0200, Daniel Vetter wrote:
> On Tue, Aug 26, 2014 at 09:44:42AM +0200, Daniel Vetter wrote:
> > On Mon, Aug 18, 2014 at 01:20:06PM +0300, Imre Deak wrote:
> > > Before sharing common parts between the system and runtime s/r
> > > handlers we WARNed if the runtime s/r handlers were called on GENs that
> > > didn't support RPM. But this WARN is not correct if the same handler is
> > > called from the system s/r path, since that can happen on any platform.
> > > This also broke system s/r on old platforms.
> > > 
> > > The issue was introduced in
> > > 
> > > commit 016970beb05da6285c2f3ed2bee1c676cb75972e
> > > Author: Sagar Kamble <sagar.a.kamble at intel.com>
> > > Date:   Wed Aug 13 23:07:06 2014 +0530
> > > 
> > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=82751
> > > Signed-off-by: Imre Deak <imre.deak at intel.com>
> > 
> > Adding boolean arguments to control warnings always feels a bit too much
> > like just shutting up the warnings. Can't we instead wrap the relevant
> > calls into HAS_RUNTIME_PM checks?

I could instead remove the WARN from intel_suspend_complete/resume, and
do an early return from intel_runtime_suspend/resume for
!HAS_RUNTIME_PM(). Atm we only WARN there.

> > Imo that would also lead to clearer code
> > by making the intention clear - with this you essentially have to git
> > blame to figure out why we sometimes disable the warning.
> 
> Also the patch subject is a bit misleading - we only shut up a wrong
> warning, it's not a code fix.

We return -ENODEV for old GENs which breaks system suspend for them.

--Imre
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 490 bytes
Desc: This is a digitally signed message part
URL: <http://lists.freedesktop.org/archives/intel-gfx/attachments/20140826/89c4c174/attachment.sig>