Re: [PATCH 2/7] drm/vc4: Fix vblank handling

From: Boris Brezillon <boris.brezillon@free-electrons.com>
To: Eric Anholt <eric@anholt.net>
Cc: Mark Rutland <mark.rutland@arm.com>, Lee Jones <lee@kernel.org>,
	Liviu Dudau <Liviu.Dudau@arm.com>,
	dri-devel@lists.freedesktop.org,
	Gerd Hoffmann <kraxel@redhat.com>,
	Florian Fainelli <f.fainelli@gmail.com>,
	Eben Upton <eben@raspberrypi.org>,
	bcm-kernel-feedback-list@broadcom.com,
	devicetree@vger.kernel.org, Pawel Moll <pawel.moll@arm.com>,
	Stephen Warren <swarren@wwwdotorg.org>,
	Ray Jui <rjui@broadcom.com>, Rob Herring <robh+dt@kernel.org>,
	linux-rpi-kernel@lists.infradead.org, "Hollingworth,
	Gordon" <gordon@raspberrypi.org>,
	"Cobley, Dom" <dom@raspberrypi.org>,
	Ian Campbell <ijc+devicetree@hellion.org.uk>,
	Scott Branden <sbranden@broadcom.com>,
	Kumar Gala <galak@codeaurora.org>,
	Shawn Guo <shawnguo@kernel.org>
Subject: Re: [PATCH 2/7] drm/vc4: Fix vblank handling
Date: Fri, 16 Jun 2017 09:47:58 +0200	[thread overview]
Message-ID: <20170616094758.66b7ea29@bbrezillon> (raw)
In-Reply-To: <87fuf0u8ul.fsf@eliezer.anholt.net>

On Thu, 15 Jun 2017 16:30:58 -0700
Eric Anholt <eric@anholt.net> wrote:

> Boris Brezillon <boris.brezillon@free-electrons.com> writes:
> 
> > There are two problems related to VBLANK handling in the current CRTC
> > driver:
> >
> > * VBLANK events are missed when the CRTC is being disabled because the
> >   driver does not wait till the end of the frame before stopping the
> >   HVS and PV blocks. In this case, we should explicitly issue a VBLANK
> >   event if there's one waiting
> > * when we are enabling a CRTC, drm_crtc_vblank_get() is called before
> >   drm_crtc_vblank_on(), which is not supposed to happen (hence the
> >   WARN_ON() in the code). To solve the problem, we delay the 'update
> >   display list' operation after the CRTC is actually enabled.  
> 
> Does drm_crtc_vblank_get() actually fail when drm_crtc_vblank_on()
> hasn't been called?  The code is a bit twisty, but it looked like we
> would track the get refcount and then turn on the interrupt once
> drm_crtc_vblank_on() was called.

I thought I had faced the WARN_ON() backtrace while developing the TXP
block, but I can't reproduce it (even after reverting these changes),
so it was probably caused by something else.

> 
> The first change of the two seems straightforward and good, though (and
> possible to separate, right?).

I'll prepare a new patch containing only the first change.

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel