Skip to contentSkip to conversation

ALSA: hda: hdmi - fix regression in connect list handling

PATCH Superseded Kai Vehmanen 2 participants 4 messages Nov 27, 2019 Sound

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ALSA: hda: hdmi - fix regression in connect list handling

Fix regression in how intel_haswell_fixup_connect_list() results are used in hda_read_pin_conn(). Use of snd_hda_get_raw_connections() in hda_read_pin_conn() bypasses the cache and thus also bypasses the overridden pin connection list. On platforms that require the connection list fixup, mux list will be empty and HDMI playback will fail to -EBUSY at open.

Move intel_haswell_fixup_connect_list() call to hda_read_pin_conn() as it's not used elsewhere.

Fixes 9c32fea83692 ("ALSA: hda - Add DP-MST support for non-acomp codecs")
Cc Nikhil Mahale <[email protected]>
---
 sound/pci/hda/patch_hdmi.c | 24 ++++++++++++++++--------
 1 file changed, 16 insertions(+), 8 deletions(-)
sound/pci/hda/patch_hdmi.c +16 -8
index 55d20e40a195..c111bdd566c9 100644
@@ -1293,6 +1293,9 @@ static int hdmi_pcm_open(struct hda_pcm_stream *hinfo,
 	return err;
 }
 
+static void intel_haswell_fixup_connect_list(struct hda_codec *codec,
+					     hda_nid_t nid);
+
 /*
  * HDA/HDMI auto parsing
  */
@@ -1302,6 +1305,7 @@ static int hdmi_read_pin_conn(struct hda_codec *codec, int pin_idx)
 	struct hdmi_spec_per_pin *per_pin = get_pin(spec, pin_idx);
 	hda_nid_t pin_nid = per_pin->pin_nid;
 	int dev_id = per_pin->dev_id;
+	int conns;
 
 	if (!(get_wcaps(codec, pin_nid) & AC_WCAP_CONN_LIST)) {
 		codec_warn(codec,
@@ -1312,10 +1316,19 @@ static int hdmi_read_pin_conn(struct hda_codec *codec, int pin_idx)
 
 	snd_hda_set_dev_select(codec, pin_nid, dev_id);
 
+	if (spec->intel_hsw_fixup) {
+		intel_haswell_fixup_connect_list(codec, pin_nid);
+		conns = snd_hda_get_connections(codec, pin_nid,
+						per_pin->mux_nids,
+						HDA_MAX_CONNECTIONS);
+	} else {
+		conns = snd_hda_get_raw_connections(codec, pin_nid,
+						    per_pin->mux_nids,
+						    HDA_MAX_CONNECTIONS);
+	}
+
 	/* all the device entries on the same pin have the same conn list */
-	per_pin->num_mux_nids =
-		snd_hda_get_raw_connections(codec, pin_nid, per_pin->mux_nids,
-					    HDA_MAX_CONNECTIONS);
+	per_pin->num_mux_nids = conns;
 
 	return 0;
 }
@@ -1713,9 +1726,6 @@ static void hdmi_repoll_eld(struct work_struct *work)
 	mutex_unlock(&spec->pcm_lock);
 }
 
-static void intel_haswell_fixup_connect_list(struct hda_codec *codec,
-					     hda_nid_t nid);
-
 static int hdmi_add_pin(struct hda_codec *codec, hda_nid_t pin_nid)
 {
 	struct hdmi_spec *spec = codec->spec;
@@ -1790,8 +1800,6 @@ static int hdmi_add_pin(struct hda_codec *codec, hda_nid_t pin_nid)
 		per_pin->dev_id = i;
 		per_pin->non_pcm = false;
 		snd_hda_set_dev_select(codec, pin_nid, i);
-		if (spec->intel_hsw_fixup)
-			intel_haswell_fixup_connect_list(codec, pin_nid);
 		err = hdmi_read_pin_conn(codec, pin_idx);
 		if (err < 0)
 			return err;
Signature · 6 lines
2.17.1

_______________________________________________
Alsa-devel mailing list
[email protected]
https://mailman.alsa-project.org/mailman/listinfo/alsa-devel
#1 lore raw

Hi Takashi and Nikhil,

Quoted 4 lines · Kai Vehmanen

Fix regression in how intel_haswell_fixup_connect_list() results are used in hda_read_pin_conn(). Use of snd_hda_get_raw_connections() in hda_read_pin_conn() bypasses the cache and thus also bypasses the overridden pin connection

the original series did pass our SOF CI, but as we have now results from wider testing, it unfortunately seems a regression slipped through. Re-reviewing the code, this bit seems clearly wrong, we can't use get_raw_connections() if we depend on the connection overrides.

This patch should not affect non-Intel platforms, but please review in any case.

Br, Kai
_______________________________________________
Alsa-devel mailing list
[email protected]
https://mailman.alsa-project.org/mailman/listinfo/alsa-devel

#2 lore raw
In reply to Kai Vehmanen
Quoted 8 lines · Kai Vehmanen

Fix regression in how intel_haswell_fixup_connect_list() results are used in hda_read_pin_conn(). Use of snd_hda_get_raw_connections() in hda_read_pin_conn() bypasses the cache and thus also bypasses the overridden pin connection list. On platforms that require the connection list fixup, mux list will be empty and HDMI playback will fail to -EBUSY at open.

Ah, thanks for catching this. This is an obvious overlook.

Quoted 4 lines
sound/pci/hda/patch_hdmi.c +0 -0
@@ -1312,10 +1316,19 @@ static int hdmi_read_pin_conn(struct hda_codec *codec, int pin_idx)
 
 	snd_hda_set_dev_select(codec, pin_nid, dev_id);
 

sound/pci/hda/patch_hdmi.c static int hdmi_read_pin_conn(struct hda_codec *codec, int pin_idx) · line 1328

Quoted 10 lines
sound/pci/hda/patch_hdmi.c +10 -0
+	if (spec->intel_hsw_fixup) {
+		intel_haswell_fixup_connect_list(codec, pin_nid);
+		conns = snd_hda_get_connections(codec, pin_nid,
+						per_pin->mux_nids,
+						HDA_MAX_CONNECTIONS);
+	} else {
+		conns = snd_hda_get_raw_connections(codec, pin_nid,
+						    per_pin->mux_nids,
+						    HDA_MAX_CONNECTIONS);
+	}

Actually intel_haswell_fixup_connect_list() doesn't influence on the hardware setup but just updates the software cache. So, basically we can copy the values directly from spec->cvt_nids here without the override hack as we have now.

That is, something like

	if (spec->intel_hsw_fixup) {
		conns = spec->cvt_nids;
		memcpy(per_pin->mux_nids, spec->num_cvts,
		       sizeof(hda_nid_t) * conns);
	} else {
	  	snd_hda_set_dev_select(codec, pin_nid, dev_id);
		conns = snd_hda_get_raw_connections(codec, pin_nid,
						    per_pin->mux_nids,
						    HDA_MAX_CONNECTIONS);
	}

Could you check whether this works?

thanks,

Takashi
_______________________________________________
Alsa-devel mailing list
[email protected]
https://mailman.alsa-project.org/mailman/listinfo/alsa-devel

#3 lore raw

Hi,

Quoted 2 lines · Takashi Iwai
Quoted 1 line · Kai Vehmanen
(quoted diff) +0 -0
@@ -1312,10 +1316,19 @@ static int hdmi_read_pin_conn(struct hda_codec *codec, int pin_idx)

[...]

Quoted 29 lines

sound/pci/hda/patch_hdmi.c static int hdmi_read_pin_conn(struct hda_codec *codec, int pin_idx) · line 1328

Quoted 10 lines
sound/pci/hda/patch_hdmi.c +10 -0
+	if (spec->intel_hsw_fixup) {
+		intel_haswell_fixup_connect_list(codec, pin_nid);
+		conns = snd_hda_get_connections(codec, pin_nid,
+						per_pin->mux_nids,
+						HDA_MAX_CONNECTIONS);
+	} else {
+		conns = snd_hda_get_raw_connections(codec, pin_nid,
+						    per_pin->mux_nids,
+						    HDA_MAX_CONNECTIONS);
+	}

Actually intel_haswell_fixup_connect_list() doesn't influence on the hardware setup but just updates the software cache. So, basically we can copy the values directly from spec->cvt_nids here without the override hack as we have now.

That is, something like

	if (spec->intel_hsw_fixup) {
		conns = spec->cvt_nids;
		memcpy(per_pin->mux_nids, spec->num_cvts,
		       sizeof(hda_nid_t) * conns);
	} else {
	  	snd_hda_set_dev_select(codec, pin_nid, dev_id);
		conns = snd_hda_get_raw_connections(codec, pin_nid,
						    per_pin->mux_nids,
						    HDA_MAX_CONNECTIONS);
	}
Quoted 1 line

Could you check whether this works?

hmm, you are right, this should work. spec->cvt_nids and spec->num_cvts were reversed in the proto code :), but otherwise looks ok based on a few tests. I'll send a v2 patch shortly. This simplifies the code as well.

Br, Kai
_______________________________________________
Alsa-devel mailing list
[email protected]
https://mailman.alsa-project.org/mailman/listinfo/alsa-devel