[alsa-devel] [PATCH 1/1] ALSA: hda - bug fix for invalid connection list of Haswell HDMI codec pin
From: Mengdong Lin mengdong.lin@intel.com
Haswell HDMI codec pins may report invalid connection list entries, which will cause failure to play audio via HDMI.
So this patch - allow generic hdmi patch to pick up and apply fixups for different codecs. - add fixup for Haswell to workaround this hardware issue: enable DP1.2 mode and override invalid connection list entries reported by hardware.
Signed-off-by: Mengdong Lin mengdong.lin@intel.com Signed-off-by: Xingchao Wang xingchao.wang@intel.com
diff --git a/sound/pci/hda/patch_hdmi.c b/sound/pci/hda/patch_hdmi.c index 71555cc..2bf1758 100644 --- a/sound/pci/hda/patch_hdmi.c +++ b/sound/pci/hda/patch_hdmi.c @@ -38,6 +38,7 @@ #include <sound/tlv.h> #include "hda_codec.h" #include "hda_local.h" +#include "hda_auto_parser.h" #include "hda_jack.h"
static bool static_hdmi_pcm; @@ -79,6 +80,7 @@ struct hdmi_spec_per_pin { };
struct hdmi_spec { + struct hda_gen_spec gen; int num_cvts; struct hdmi_spec_per_cvt cvts[MAX_HDMI_CVTS];
@@ -1687,6 +1689,69 @@ static const struct hda_codec_ops generic_hdmi_patch_ops = { .unsol_event = hdmi_unsol_event, };
+static void intel_haswell_fixup_enable_dp12_mode(struct hda_codec *codec, + const struct hda_fixup *fix, int action) +{ + unsigned int vendor_param; + + if (action != HDA_FIXUP_ACT_PRE_PROBE) + return; + + vendor_param = snd_hda_codec_read(codec, 0x08, 0, 0xf81, 0); + if (vendor_param == -1 || vendor_param & 0x02) + return; + + vendor_param |= 0x02; + snd_hda_codec_read(codec, 0x08, 0, 0x781, vendor_param); +} + +static void intel_haswell_fixup_connect_list(struct hda_codec *codec, + const struct hda_fixup *fix, int action) +{ + hda_nid_t list[3] = {0x2, 0x3, 0x4}; + unsigned int vendor_param; + + if (action != HDA_FIXUP_ACT_PROBE) + return; + + vendor_param = snd_hda_codec_read(codec, 0x08, 0, 0xf81, 0); + if (vendor_param == -1 || !(vendor_param & 0x02)) + return; + + snd_hda_override_conn_list(codec, 0x06, 3, list); + hdmi_read_pin_conn(codec, 1); +} + + +/* available models for fixup */ +enum { + INTEL_HASWELL, + INTEL_HASWELL_CONNECT_LIST, +}; + +static const struct hda_model_fixup hdmi_models[] = { + {.id = INTEL_HASWELL, .name = "Haswell"}, + {} +}; + +static const struct snd_pci_quirk hdmi_fixup_tbl[] = { + SND_PCI_QUIRK(0x8086, 0x2010, "Haswell", INTEL_HASWELL), + {} /* terminator */ +}; + +static const struct hda_fixup hdmi_fixups[] = { + [INTEL_HASWELL] = { + .type = HDA_FIXUP_FUNC, + .v.func = intel_haswell_fixup_enable_dp12_mode, + .chained = true, + .chain_id = INTEL_HASWELL_CONNECT_LIST, + }, + [INTEL_HASWELL_CONNECT_LIST] = { + .type = HDA_FIXUP_FUNC, + .v.func = intel_haswell_fixup_connect_list, + }, +}; + static int patch_generic_hdmi(struct hda_codec *codec) { struct hdmi_spec *spec; @@ -1696,11 +1761,19 @@ static int patch_generic_hdmi(struct hda_codec *codec) return -ENOMEM;
codec->spec = spec; + snd_hda_gen_init(&spec->gen); + + snd_hda_pick_fixup(codec, hdmi_models, hdmi_fixup_tbl, hdmi_fixups); + snd_hda_apply_fixup(codec, HDA_FIXUP_ACT_PRE_PROBE); + if (hdmi_parse_codec(codec) < 0) { codec->spec = NULL; kfree(spec); return -EINVAL; } + + snd_hda_apply_fixup(codec, HDA_FIXUP_ACT_PROBE); + codec->patch_ops = generic_hdmi_patch_ops; generic_hdmi_init_per_pins(codec);
At Fri, 14 Dec 2012 13:57:30 -0500, mengdong.lin@intel.com wrote:
How is DP1.2 enablement involved with this? I don't see a link between DP1.2 and the lack of connection list...
Isn't this needed after S4?
Is this ACT_PROBE, that is, applied only after parsing the codec? Usually a fixup for overriding the connection list is applied before parsing the tree.
thanks,
Takashi
Yes, this fixup is applied after parsing the codec. Because during parsing, the pin nid#6 may still report invalid connection list entries 0x00000000 in snd_hda_get_raw_connections(), which should be 0x00040302 to stand for 3 output convertors. So we need to overwrite its connection list with the proper value.
But, do you mean we can change this to the pre-probe stage, so that snd_hda_get_raw_connections() need not be called during the parsing the codec? I'll do a test.
Thanks Mengdong
At Fri, 14 Dec 2012 07:26:28 +0000, Lin, Mengdong wrote:
Yes. It may work currently but it might be broken if the parser code is changed, for example. I see no reason why the fixup must be applied after parsing the tree.
Takashi
Hi Takashi,
You're right. Overriding the connection list can work after being merged into the pre-probe stage and so only one fix-up is enough.
I'll update the patch. Many thanks for your advice!
- Mengdong
participants (3)
-
Lin, Mengdong -
mengdong.lin＠intel.com
-
Takashi Iwai