Re: [PATCH 4/6] drm/omap: Allow allocation of larger buffers

[Archit Taneja <archit@ti.com>]

On Monday 14 April 2014 04:00 PM, Tomi Valkeinen wrote:
> On 14/04/14 13:18, Archit Taneja wrote:
>> On Monday 14 April 2014 02:55 PM, Tomi Valkeinen wrote:
>>> On 11/04/14 10:23, Archit Taneja wrote:
>>>> The drm ioctl DRM_IOCTL_MODE_ADDFB2 doesn't let us allocate buffers
>>>> which are
>>>> greater than what is specified in the driver through dev->mode_config.
>>>>
>>>> Create helpers for DISPC which return the max manager width and
>>>> height supported
>>>> by the device. The maximum width for a framebuffer is set to the
>>>> combined width
>>>> of the all the crtcs, assuming they are arranged horizontally.
>>>>
>>>> Signed-off-by: Archit Taneja <archit@ti.com>
>>>> ---
>>>>    drivers/gpu/drm/omapdrm/omap_drv.c | 10 ++++++----
>>>>    drivers/video/omap2/dss/dispc.c    | 12 ++++++++++++
>>>>    include/video/omapdss.h            |  2 ++
>>>>    3 files changed, 20 insertions(+), 4 deletions(-)
>>>>
>>>> diff --git a/drivers/gpu/drm/omapdrm/omap_drv.c
>>>> b/drivers/gpu/drm/omapdrm/omap_drv.c
>>>> index c8270e4..55ec575 100644
>>>> --- a/drivers/gpu/drm/omapdrm/omap_drv.c
>>>> +++ b/drivers/gpu/drm/omapdrm/omap_drv.c
>>>> @@ -306,11 +306,13 @@ static int omap_modeset_init(struct drm_device
>>>> *dev)
>>>>        dev->mode_config.min_width = 32;
>>>>        dev->mode_config.min_height = 32;
>>>>
>>>> -    /* note: eventually will need some cpu_is_omapXYZ() type stuff here
>>>> -     * to fill in these limits properly on different OMAP generations..
>>>> +    /*
>>>> +     * Note: the maximum width is set to the combined width of all the
>>>> +     * crtcs. We could assume the same for the maximum height too, but
>>>> +     * we generally don't use such a configuration.
>>>>         */
>>>> -    dev->mode_config.max_width = 2048;
>>>> -    dev->mode_config.max_height = 2048;
>>>> +    dev->mode_config.max_width = num_crtcs * dispc_mgr_max_width();
>>>> +    dev->mode_config.max_height = dispc_mgr_max_height();
>>>
>>> This looks very strange.
>>>
>>> If the max size is supposed to be the maximum output size we support,
>>> then multiplying with num_crtcs doesn't make sense.
>>>
>>> If, on the other hand, it tells the possible maximum size of the
>>> framebuffer in the memory, of which only small part is shown (where's
>>> the max size of that "part" defined, then?), then there should be no
>>> limits as the only limit is the size of the memory.
>>
>> These parameters are used to tell the max size of framebuffer we can
>> allocate in memory.
>>
>> I'm not sure why there is a limit in the first place, but if we have a
>> really low minimum(like 2048 pixels right now), we can't have multiple
>> displays showing different regions of the same buffer.
>
> In that case shouldn't the limit be 0 (if that's allowed) or some surely
> high enough value, say, 32767. Why calculate it at all based on the
> dispc's maximum, if it has no relevance?

I understand your point. When I wrote the patch, I thought more like: "I 
have 3 displays arranged horizontally, if I want a buffer large enough 
to fill the 3 displays, what should be it's width/height?"

Archit