On Thu, Dec 06, 2012 at 04:57:27PM +0100, Daniel Vetter wrote: > On Thu, Dec 6, 2012 at 3:23 PM, Thierry Reding > <thierry.reding@xxxxxxxxxxxxxxxxx> wrote: > >> > - /* sdvo spec says that the ecc is handled by the hw, and it looks like > >> > - * we must not send the ecc field, either. */ > >> > - memcpy(sdvo_data, &avi_if, 3); > >> > - sdvo_data[3] = avi_if.checksum; > >> > >> This difference between how we need to send the infoframe to the hw > >> between sdvo (without ecc field present) and native hdmi (ecc field > >> present, but still computed by the hw) seems to have been lost. On a quick > >> read from your infoframe packer it looks like the ecc field isn't part of > >> the packed infoframe, so this patch would break native hdmi ports (in the > >> intel_hdmi.c file). > > > > Actually this ECC field seems to be something specific to Intel > > hardware. It doesn't appear in the HDMI specification nor in CEA-861-D. > > So it the binary infoframe representation doesn't contain it, therefore > > it should be handled by the respective drivers. So in fact the code > > should be working fine for HDMI and SDVO, since for the HDMI ports I > > explicitly left the fourth byte of the header empty, while for SDVO, the > > checksum is the one as computed by hdmi_avi_infoframe_pack(), which is > > at byte 3 and therefore matches the previous code. > > Oh dear, I've failed to read the patch correctly and totally missed > the new intel_dip_pack function and assorted logic changes. Sorry for > the confusion. At first I thought to split that out in a separate patch but then decided it wasn't worth the trouble. But there's always the possibility that I've broken things horribly here since the patch is rather messy, so it's good that you look closely and some testing will certainly be good to validate that things still work as before. Thierry
Attachment:
pgpaVLeQfKmmP.pgp
Description: PGP signature
_______________________________________________ dri-devel mailing list dri-devel@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/dri-devel
