[Nouveau] Only stereo sound with gtx570 over hdmi (regression)

[Ben Skeggs]

On Tue, Feb 17, 2015 at 11:41 PM, Farid BENAMROUCHE <fariouche at yahoo.fr> wrote:
> Hello Ben,
>
> The working kernel code I based my investigation was a 1.3.17 + gentoo patch (so yes, pretty old)
>
> The new one below, not working is the vanilla 3.19 (from the gentoo repo, should be identical to the latest 3.19 stable)
>
> I've narrowed down the issue to the size of the eld.
> The new patch (that indeed uses drm_eld_size) (with this patch, I have a working 3.19 kernel):
> I will need to restart my computer with the non working kernel to check the eld output (by the way, what is the ELD?, is it the EDID?)
>
> Thanks
>
> --- drivers/gpu/drm/nouveau/nv50_display.c.orig 2015-02-15 20:04:10.583860698 +0100
> +++ drivers/gpu/drm/nouveau/nv50_display.c      2015-02-15 20:46:43.518148777 +0100
> @@ -1700,7 +1700,7 @@
>         memcpy(args.data, nv_connector->base.eld, sizeof(args.data));
>
>         nvif_mthd(disp->disp, 0, &args,
> -                 sizeof(args.base) + drm_eld_size(args.data));
> +                 sizeof(args.base));
That's... Odd.  Can you boot with "nouveau.debug=drm=trace" both with
and without the patch and attach the results here?

Thanks,
Ben.

>  }
>
>  static void
>
> --------------------------------------------
> En date de : Mar 17.2.15, Ben Skeggs <skeggsb at gmail.com> a écrit :
>
>  Objet: Re: [Nouveau] Only stereo sound with gtx570 over hdmi (regression)
>  À: "Farid BENAMROUCHE" <fariouche at yahoo.fr>
>  Cc: "nouveau at lists.freedesktop.org" <nouveau at lists.freedesktop.org>
>  Date: Mardi 17 février 2015, 0h47
>
>  On Mon, Feb 16, 2015 at
>  5:03 AM, Farid BENAMROUCHE <fariouche at yahoo.fr>
>  wrote:
>  > Hello all,
>  >
>  > I'm using gentoo,
>  with kernel 3.17.0-p1-pf and at some point, a patch was
>  included in this branch of the gentoo kernel that broke hdmi
>  audio.
>  >
>  > I've
>  checked with the latest 3.19 vanilla kernel, and I still
>  have the same problem. I cannot output multichannel sound
>  over hdmi.
>  >
>  > After
>  some investigations, I've narrowed down the issue to the
>  following lines in the file nv50_display.c:
>  >
>  > (the structure
>  slightly changed and declared "packed". Reverting
>  this change made the multichannel audio back)
>  > I still do not known exactly when this
>  modification was introoduced. If you need more details,
>  please let me know.
>  There were bugs in the
>  implementation for a while that could've caused
>  this problem.  However, those should be fixed
>  already by 3.19.  What
>  kernel version is
>  your revert against?  It's not the latest code,
>  because that uses drm_eld_size().
>
>  Testing on my TV at home
>  (3.19.0-rc5), I see this in /proc/asound/card1/eld#0.0:
>
>  monitor_present 1
>  eld_valid 1
>  monitor_name
>  SAMSUNG
>  connection_type HDMI
>  eld_version [0x2] CEA-861D or below
>  edid_version [0x3] CEA-861-B, C or D
>  manufacture_id 0x2d4c
>  product_id 0x69b
>  port_id 0x0
>  support_hdcp 0
>  support_ai 1
>  audio_sync_delay 0
>  speakers
>  [0xf] FL/FR LFE FC RL/RR
>  sad_count 4
>  sad0_coding_type [0x1] LPCM
>  sad0_channels 2
>  sad0_rates
>  [0x6e0] 32000 44100 48000 88200 96000
>  sad0_bits [0xe0000] 16 20 24
>  sad1_coding_type [0x1] LPCM
>  sad1_channels 6
>  sad1_rates
>  [0x6e0] 32000 44100 48000 88200 96000
>  sad1_bits [0xe0000] 16 20 24
>  sad2_coding_type [0x2] AC-3
>  sad2_channels 6
>  sad2_rates
>  [0xe0] 32000 44100 48000
>  sad2_max_bitrate
>  640000
>  sad3_coding_type [0x7] DTS
>  sad3_channels 6
>  sad3_rates
>  [0xe0] 32000 44100 48000
>  sad3_max_bitrate
>  1536000
>
>  Which looks
>  correct, and like the code you mention is doing the right
>  thing.
>
>  Thanks,
>  Ben.
>
>
>  >
>  > ---
>  working_kernel/drivers/gpu/drm/nouveau/nv50_display.c
>  > +++
>  not_working_kernel/drivers/gpu/drm/nouveau/nv50_display.c
>  > @@ -1653,15 +1653,17 @@
>  >         struct
>  nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
>  >         struct
>  nouveau_connector *nv_connector;
>  >
>       struct nv50_disp *disp =
>  nv50_disp(encoder->dev);
>  > -
>     struct {
>  > -
>       struct nv50_disp_mthd_v1 base;
>  > -               struct
>  nv50_disp_sor_hda_eld_v0 eld;
>  > +
>     struct __packed {
>  > +
>             struct {
>  > +
>                       struct
>  nv50_disp_mthd_v1 mthd;
>  > +
>               struct nv50_disp_sor_hda_eld_v0
>  eld;
>  > +               }
>  base;
>  >
>     u8 data[sizeof(nv_connector->base.eld)];
>  >         } args = {
>  > -
>     .base.version = 1,
>  > -
>             .base.method  =
>  NV50_DISP_MTHD_V1_SOR_HDA_ELD,
>  > -
>             .base.hasht   =
>  nv_encoder->dcb->hasht,
>  > -
>           .base.hashm   =
>  nv_encoder->dcb->hashm,
>  > +
>           .base.mthd.version = 1,
>  > +
>     .base.mthd.method  =
>  NV50_DISP_MTHD_V1_SOR_HDA_ELD,
>  > +
>             .base.mthd.hasht   =
>  nv_encoder->dcb->hasht,
>  > +
>           .base.mthd.hashm   =
>  nv_encoder->dcb->hashm,
>  >
>     };
>  >
>  >
>         nv_connector =
>  nouveau_encoder_connector_get(nv_encoder);
>  > @@ -1671,7 +1673,7 @@
>  >
>     drm_edid_to_eld(&nv_connector->base,
>  nv_connector->edid);
>  >
>     memcpy(args.data, nv_connector->base.eld,
>  sizeof(args.data));
>  >
>  > -
>     nvif_mthd(disp->disp, 0, &args,
>  sizeof(args));
>  > +
>     nvif_mthd(disp->disp, 0, &args,
>  sizeof(args.base) + args.data[2] * 4);
>  >  }
>  >
>  >  static void
>  >
>  >
>  >
>  _______________________________________________
>  > Nouveau mailing list
>  >
>  Nouveau at lists.freedesktop.org
>  > http://lists.freedesktop.org/mailman/listinfo/nouveau
>