From mboxrd@z Thu Jan 1 00:00:00 1970 From: Daniel Vetter Subject: Re: [PATCH] drm/i915: Fix initial pipe underrun state tracking Date: Mon, 24 Mar 2014 10:28:23 +0100 Message-ID: <20140324092823.GB26878@phenom.ffwll.local> References: <1395615701-3914-1-git-send-email-daniel.vetter@ffwll.ch> <87ior4vwd8.fsf@intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Return-path: Received: from mail-ee0-f48.google.com (mail-ee0-f48.google.com [74.125.83.48]) by gabe.freedesktop.org (Postfix) with ESMTP id E25826E917 for ; Mon, 24 Mar 2014 02:28:27 -0700 (PDT) Received: by mail-ee0-f48.google.com with SMTP id b57so4167431eek.7 for ; Mon, 24 Mar 2014 02:28:27 -0700 (PDT) Content-Disposition: inline In-Reply-To: <87ior4vwd8.fsf@intel.com> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: intel-gfx-bounces@lists.freedesktop.org Sender: "Intel-gfx" To: Jani Nikula Cc: Daniel Vetter , Intel Graphics Development List-Id: intel-gfx@lists.freedesktop.org On Mon, Mar 24, 2014 at 10:22:59AM +0200, Jani Nikula wrote: > On Mon, 24 Mar 2014, Daniel Vetter wrote: > > Since > > > > commit 5c673b60a9b3b23486f4eda75c72e91d31d26a2b > > Author: Daniel Vetter > > Date: Fri Mar 7 20:34:46 2014 +0100 > > > > drm/i915: Don't enable display error interrupts from the start > > > > we don't enable underrun interrupts any more at takeover time. > > Unfortunately I've forgotten to also adjust the sw-side tracking. > > > > Since the code assumes that disabled pipes have underrun reporting > > enabled set the disable flag on all pipes which are active at takeover > > time. Without this underrun reporting wasn't enabled correctly on the > > first modeset. Note that for fastboot this is another piece of state > > that needs to be fixed up by enabling the underrung reporting after > > watermarks have beend fixed up. > > > > On ivb/hsw an additional effect of this regression was that also all > > cpu crc reporting stopped working since the master error interrupt it > > shared across all pipes and sources. > > > > Cc: Ville Syrj=E4l=E4 > > Cc: Jani Nikula > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=3D76150 > > Signed-off-by: Daniel Vetter > > --- > > drivers/gpu/drm/i915/intel_display.c | 8 ++++++++ > > 1 file changed, 8 insertions(+) > > > > diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i91= 5/intel_display.c > > index 7e4ea8d4e388..9b24ae4fb7bd 100644 > > --- a/drivers/gpu/drm/i915/intel_display.c > > +++ b/drivers/gpu/drm/i915/intel_display.c > > @@ -11502,6 +11502,14 @@ static void intel_sanitize_crtc(struct intel_c= rtc *crtc) > > encoder->base.crtc =3D NULL; > > } > > } > > + if (crtc->active) { > > + /* > > + * We start out with underrun reporting disabled to avoid races. > > + * For correct bookkeeping mark this on active crtcs. > > + */ > = > Why only on active crtcs? The state remains in inactive crtcs. Won't the > problem appear when such an crtc is actived? Setting this for inactive crtcs breaks the code, which I've learned the hard way ;-) The issue is that on ivb/hsw we can _only_ enable the err interrupt if all crtcs allow it. But if some disabled crtc disallows it that condition is never true until userspace has lit up all for crtcs at least once. I've tried to capture this in the commit message, but seem to have failed. Can you please suggest a rewording to clarify things? > Quoth a comment in struct intel_crtc: > = > /* Access to these should be protected by dev_priv->irq_lock. */ > = > Either you need to hold the lock or explain in a comment why it's not > necessary. At worst the interrupt handler runs concurrently (on platforms where we don't yet hide fifo underruns from the start), also setting this to false in the case of an underrun. What about: "No protection against concurrent access is required - at wors a fifo underrun happens which also sets this to false." Thanks, Daniel -- = Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch