2017년 03월 09일 15:54에 Andrzej Hajda 이(가) 쓴 글:
> On 09.03.2017 04:54, Michel Dänzer wrote:
>> On 08/03/17 11:58 PM, Andrzej Hajda wrote:
>>> Current implementation of event handling assumes that vblank interrupt is
>>> always called at the right time. It is not true, it can be delayed due to
>>> various reasons. As a result different races can happen. The patch fixes
>>> the issue by using hardware frame counter present in DECON to serialize
>>> vblank and commit completion events.
>>>
>>> Signed-off-by: Andrzej Hajda <a.hajda@xxxxxxxxxxx>
>>> ---
>>> v2:
>>> - added internal decon_get_frame_count function,
>>> - updated frame counter on flush,
>>> - misc style fixes (thank Inki)
>> [...]
>>
>>> @@ -579,6 +636,23 @@ static const struct component_ops decon_component_ops = {
>>> .unbind = decon_unbind,
>>> };
>>>
>>> +static void decon_handle_vblank(struct decon_context *ctx)
>>> +{
>>> + u32 frm;
>>> +
>>> + spin_lock(&ctx->vblank_lock);
>>> +
>>> + frm = decon_get_frame_count(ctx, true);
>>> +
>>> + if (frm != ctx->frame_id) {
>>> + if (frm > ctx->frame_id)
>>> + drm_crtc_handle_vblank(&ctx->crtc->base);
>> This comparison isn't safe WRT the counter value returned by
>> decon_get_frame_count wrapping around.
>
> And knowing that max framerate is 60fps it will happen after:
>
> 0xffffffff / 60fps / 60sec / 60min / 24h / 365days = 2.3 years
>
> So after 2.3 years of uninterrupted work of panel/TV we will lose one or
> two vblanks :) Highly improbable and even then low damage.
>
>> If it goes all the way up to
>> 0xffffffff before wrapping around to zero, it can be handled e.g. by
>>
>> if ((s32)(frm - ctx->frame_id) > 0)
>>
>> otherwise it gets a bit more complicated.
>>
>>
> But the fix looks simple so I think it is worth to fix it. Thanks.
Andrzej, v3?
>
> Regards
>
> Andrzej
>
>
>
>
>
>
_______________________________________________
dri-devel mailing list
dri-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/dri-devel
