From: Eric Anholt Subject: Re: [RFC] drm/i915: always set positive sync in the sdvo register Date: Thu, 08 Dec 2011 15:19:58 -0800 Message-ID: <87hb1aodlt.fsf@eliezer.anholt.net> References: <1323358913-4346-1-git-send-email-przanoni@gmail.com> Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="===============1184615382==" Return-path: In-Reply-To: <1323358913-4346-1-git-send-email-przanoni@gmail.com> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: intel-gfx-bounces+gcfxdi-intel-gfx=m.gmane.org@lists.freedesktop.org Errors-To: intel-gfx-bounces+gcfxdi-intel-gfx=m.gmane.org@lists.freedesktop.org To: przanoni@gmail.com, intel-gfx@lists.freedesktop.org Cc: Paulo Zanoni List-Id: intel-gfx@lists.freedesktop.org --===============1184615382== Content-Type: multipart/signed; boundary="=-=-="; micalg=pgp-sha1; protocol="application/pgp-signature" --=-=-= Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: quoted-printable On Thu, 8 Dec 2011 13:41:53 -0200, przanoni@gmail.com wrote: > From: Paulo Zanoni >=20 > We use struct intel_sdvo_dtd for that too. >=20 > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=3D15766 > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=3D42174 > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=3D43333 >=20 > Signed-off-by: Paulo Zanoni I think the patch this is sort-of-reverting was wrong. Here's what I found: It was non-obvious to me what vblank high/low would mean on a digital link. It turns out that there are 3 magic characters sent down instead of pixel data: blank, hsync, and vsync (and syncs can overlap). The sync polarity field for SDVOB in the PRM (looking at G35 here) says: "Indicates the polarity of Hsync and Vsync. Inverted polarity is transmitted as SYNC-BLANK-SYNC and standard polarity is transmitted as BLANK-SYNC-BLANK. For example, if Vsync is not inverted and Hsync is inverted, an Hsync period transmitted during Vsync would be transmitted as BLANK+VS+HS =E2=80=93 BLANK+VS =E2=80=93 BLANK+VS+HS= ." Looking at my SDVO EDS, there's an example of "Invalid HSync Blank Periods", which includes a sample sequence of "active blank+ hsync blank hsync blank+ active" being invalid, which matches that description for hsync low. The valid hsync examples all match "active blank+ hsync+ blank+ active". (+ here is the regexp meaning of "at least one of the character"). I don't see equivalent examples banning vertical low, but I'd expect so (the valid vertical example is vertical high). However, I think setting these flags pre-gen4 is also wrong -- bit 4 is reserved and bit 3 is RO and has a different meaning. I think the pure revert is the right answer, instead. A couple of interesting notes for people debugging SDVO that aren't reflected in our code: vblank must be at least 3 lines vsync must be at least one line. hblank must be at least hsync characters hsync must be at least 16 characters. --=-=-= Content-Type: application/pgp-signature -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.11 (GNU/Linux) iEYEARECAAYFAk7hRh4ACgkQHUdvYGzw6veMxgCgiXmFc5ECc241zvakvHua9EF2 qwQAnRLShxu/uIP36/Ds1h0akdjAJn1N =GcSa -----END PGP SIGNATURE----- --=-=-=-- --===============1184615382== Content-Type: text/plain; charset="us-ascii" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit Content-Disposition: inline _______________________________________________ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx --===============1184615382==--