<html> <head> <meta content="text/html; charset=utf-8" http-equiv="Content-Type"> </head> <body bgcolor="#FFFFFF" text="#000000"> <div class="moz-cite-prefix">On 2016年05月25日 09:06, Mark yao wrote: </div> <blockquote cite="mid:5744FA7C.3080802@rock-chips.com" type="cite"> <meta content="text/html; charset=utf-8" http-equiv="Content-Type"> <div class="moz-cite-prefix">On 2016年05月24日 18:11, Tomeu Vizoso wrote: </div> <blockquote cite="mid:CAAObsKAj9Ja56buTixvp0EVBk=xE4JCXPPwXdviweHkmUQb2xg@mail.gmail.com" type="cite"> <blockquote type="cite" style="color: #000000;"> <pre wrap="">Hi Tomeu > > Sorry for reply late. > I don't agree the changes: > > - if (!state->enable) > - return VOP_WIN_GET(vop_win->vop, vop_win->data, enable) == 0; > + if (!state->enable && > + VOP_WIN_GET(vop_win->vop, vop_win->data, enable) == 0) > + return true; > > This changes actually would lead a bug. > when we disable a plane, the vop_win_pending_is_complete would always return > true, That is not allowed, would cause fb free too early. </pre> </blockquote> <pre wrap="">Ok, maybe I need to ask you first what the original block of code intended to achieve. The TRM I have is very terse and I don't find any explanation there. The battery of tests I have pass just fine without it. </pre> <blockquote type="cite" style="color: #000000;"> <pre wrap="">> Does this patch is needed for "[PATCH 2/2] drm/rockchip: vop: Wait for > pending events when disabling a CRTC" </pre> </blockquote> <pre wrap="">Yes, this function is currently returning false when the pageflip has been completed but the plan has been already disabled. If you have any different idea of how to fix this bug, please share. Thanks, Tomeu </pre> </blockquote> Hi Tomeu <pre wrap="">@@ -504,6 +506,9 @@ static void vop_crtc_disable(struct drm_crtc *crtc) if (!vop->is_enabled) return; + if (crtc->state->event || vop->event) + vop_crtc_wait_for_update(crtc); + I think above change has some problem, the function stack: ->drm swap state ->vop_crtc_disable ->vop_atomic_begin ->vop_atomic_flush on vop_crtc_disable, crtc->state is new state, the crtc->state->event not yet update to vop, wait for crtc->state->event here is wrong. So I think the patch should be: + if (vop->event) + vop_crtc_wait_for_update(crtc); +</pre> </blockquote> call vop_crtc_wait_for_update(crtc) here also is unsafe, it will reinit the vop->wait_update_complete. I doubt that, since use the serialize outstanding nonblocking commits, only one process can run into the update stack, old vop->event should be free on last time, if we get vop->event here, that should be a bug. <blockquote cite="mid:5744FA7C.3080802@rock-chips.com" type="cite"> <pre wrap=""> Then the patch "drm/rockchip: vop: Do check if an update is pending during disable" should be no needed. Thanks. </pre> -- Ｍark Yao <fieldset class="mimeAttachmentHeader"></fieldset> <pre wrap="">_______________________________________________ dri-devel mailing list <a class="moz-txt-link-abbreviated" href="mailto:dri-devel@lists.freedesktop.org">dri-devel@lists.freedesktop.org</a> <a class="moz-txt-link-freetext" href="https://lists.freedesktop.org/mailman/listinfo/dri-devel">https://lists.freedesktop.org/mailman/listinfo/dri-devel</a> </pre> </blockquote> <pre class="moz-signature" cols="72">-- Ｍark Yao</pre> </body> </html>