Re: [PATCH RFC 3/9] drm/omap: Add ovl_name() and mgr_name() to dispc_ops

From: Tomi Valkeinen <tomi.valkeinen@ti.com>
To: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Cc: dri-devel@lists.freedesktop.org, Jyri Sarha <jsarha@ti.com>
Subject: Re: [PATCH RFC 3/9] drm/omap: Add ovl_name() and mgr_name() to dispc_ops
Date: Wed, 28 Feb 2018 16:31:01 +0200	[thread overview]
Message-ID: <d4445bab-75c4-6372-46eb-3bcbb4d7e2b6@ti.com> (raw)
In-Reply-To: <1733406.ld8gWDTvng@avalon>

On 28/02/18 16:24, Laurent Pinchart wrote:
> Hi Tomi,
> 
> On Wednesday, 28 February 2018 16:05:34 EET Tomi Valkeinen wrote:
>> On 28/02/18 15:23, Laurent Pinchart wrote:
>>> On Wednesday, 28 February 2018 13:37:48 EET Tomi Valkeinen wrote:
>>>> On 27/02/18 16:35, Laurent Pinchart wrote:
>>>>> the whole omap_irq_fifo_underflow() and omap_irq_ocp_error_handler() IRQ
>>>>> handling to the DSS side, as they're not DRM/KMS-related ?
>>>>
>>>> I think we should react to both errors somehow (I'm not sure how,
>>>> disable output probably), and that has to be done on the KMS level. We
>>>> don't do that now, but moving this to DSS side would make error handling
>>>> more difficult to do in the future.
>>>
>>> Ideally I'd demultiplex interrupts on the DSS side and report events to
>>> the KMS side (page flip completion, underflows, ...).
>>
>> That's more or less what Jyri's "drm/omap: Make omapdss API more
>> generic" does, isn't it? Or what is the difference with interrupt and
>> event in your mind? Function calls vs status bits?
> 
> Yes, that's the difference in my mind. I'd keep the status bits on the DSS 
> side. We don't have to implement one callback function for each status bit, we 
> could translate them into abstract event bits that are not specific to a 
> particular DSS version. What I'd like to avoid is omapdrm calling into omapdss 
> to retrieve a name for a bit that is DSS-specific. If we want hardware names 
> in debug messages I think they should be printed on the omapdss side, and if 
> we want to handle status bits on the omapdrm side they shouldn't require 
> hardware names.

Ok, yes, I see your point, and agree.

Here, I think what's done (after the IRQ change patch) is that we get a
an abstracted bit for the underflow. Omapdrm can map that bit to an
omap_plane, and should get the name from omap_plane, instead of asking
it directly from dispc.

 Tomi

-- 
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki.
Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel