From mboxrd@z Thu Jan 1 00:00:00 1970 From: Tomi Valkeinen Subject: Re: [PATCHv2 42/45] drm: omapdrm: add omap_atomic_wait_for_gos() Date: Mon, 8 Jun 2015 11:36:25 +0300 Message-ID: <55755409.5060100@ti.com> References: <1433408582-9828-1-git-send-email-tomi.valkeinen@ti.com> <1433408582-9828-43-git-send-email-tomi.valkeinen@ti.com> <7878755.bG2kh6HLMr@avalon> Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="===============0511013222==" Return-path: Received: from devils.ext.ti.com (devils.ext.ti.com [198.47.26.153]) by gabe.freedesktop.org (Postfix) with ESMTP id 561D66E648 for ; Mon, 8 Jun 2015 01:36:30 -0700 (PDT) In-Reply-To: <7878755.bG2kh6HLMr@avalon> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dri-devel-bounces@lists.freedesktop.org Sender: "dri-devel" To: Laurent Pinchart Cc: dri-devel@lists.freedesktop.org List-Id: dri-devel@lists.freedesktop.org --===============0511013222== Content-Type: multipart/signed; micalg=pgp-sha256; protocol="application/pgp-signature"; boundary="G9K9N00O23A7hsmGmKBMl0t4QTcNVKtO9" --G9K9N00O23A7hsmGmKBMl0t4QTcNVKtO9 Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: quoted-printable On 06/06/15 07:10, Laurent Pinchart wrote: > Hi Tomi, >=20 > Thank you for the patch. >=20 > On Thursday 04 June 2015 12:02:59 Tomi Valkeinen wrote: >> omap_atomic_complete() uses drm_atomic_helper_wait_for_vblanks() to wa= it >> for all operations to finish. That works, but can easily cause waits f= or >> vblanks when no wait is actually necessary. >=20 > It actually doesn't work properly, as a full mode set without any chang= e to=20 > framebuffers will need to reenable planes after enabling the CRTC and t= hus=20 > sets the go bit, but the drm_atomic_helper_wait_for_vblanks() function = called=20 That's not how it goes. In complete(), we first disable the crtcs, then set the planes, then enable the crtcs. We don't first enable the crtcs, and then enable the planes. So everything is already configured at the time we enable the crtc, and no GO bit is needed, and thus no wait is needed. > during atomic commit waits for vblank only on CRTCs that have seen a pa= ge=20 > flip. This leads to missing waits, resulting in a go busy warning if th= e next=20 > .atomic_flush() call occurs too soon. I think those issues came from your patch that moved the plane config to be done after crtc enable. I dropped that patch. Tomi --G9K9N00O23A7hsmGmKBMl0t4QTcNVKtO9 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v2 iQIcBAEBCAAGBQJVdVQJAAoJEPo9qoy8lh71wgYP/jwKDgRdlVXx1PKthyBtCTul iAk2ETTbJI7hsLAz3fjU3kbSY/FE9F6jHPjH/uKQk3lNzAcJZkDs9nLicxV8xUyC He3RykoaoIIIMp1uy3bDpxeDHhnEZSSZNrjMdfOTtFD3Eoso2llRaLZQBX8alfOx gOLegJxEOhb4SJ2/UV5G/IKMClz/wNlvwbfqSm3IRUt74JCPNvvSQGZoesWQf0HD SYh9fUv6DmZY0sFhG6icdo+VSvkdbO2G7lLHIRM2fpWSGWxfWcgyuKa8bHqHicb3 Dewd0yIpCl5I/z5RygHxV3s5Rqdq+FSkdMN8yJFMxLjGdNjTkDSsv7pwcyILia6E XwMdAOrd+zwIlwNbMukHSC7fKp7fAE4avet6+VQr1m7YJEHDkld9HPDldev5jmWF lQNw0tme5vA0PX49zdRW1Xstex/weFpygAnsk8ylSBhykbw/0eUWabUDOxhUKDz7 Q7UAQp2D9cIo36HJFPpyfmZLiQw6Yoy/tMWX7TffXnJMD9qegCifq203Ma9SK9KS U/wL57/O6v+XogVN3DHZo4xO/mlsi64A95tihRbtEYvBqrwopFPCjSxuX1bitklN +KZtrStpIujjK/JhWQaiSIMnCvtMn6/1Sgv31O63MLNC1Axjybq+p41E7YgpwGAV RL+od703Fzsfa50BkCwn =iUyh -----END PGP SIGNATURE----- --G9K9N00O23A7hsmGmKBMl0t4QTcNVKtO9-- --===============0511013222== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: base64 Content-Disposition: inline X19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX18KZHJpLWRldmVs IG1haWxpbmcgbGlzdApkcmktZGV2ZWxAbGlzdHMuZnJlZWRlc2t0b3Aub3JnCmh0dHA6Ly9saXN0 cy5mcmVlZGVza3RvcC5vcmcvbWFpbG1hbi9saXN0aW5mby9kcmktZGV2ZWwK --===============0511013222==--