Merge tag 'sound-5.3-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound

Pull sound updates from Takashi Iwai:
 "Many updates in this development cycle are found in ASoC where it got
  a wide range of changes for the continued refactoring.

  Some highlights are below.

  ASoC:

   - Continued refactoring work by Morimoto-san toward the full
     componentization; the changes are seen allover the places

   - Support for force disconnecting muxes in DAPM

   - Continued development of ASoC Intel SOF stuff

   - New drivers for Cirrus Logic CS47L35, CS47L85 and CS47L90, Conexant
     CX2072X, Realtek RT1011 and RT1308

  HD-audio:

   - More fixes and adjustments for ASoC SOF HD-audio

   - Fix for resume problem on some Realtek codecs

  USB-audio:

   - A few fixes for the issues reported by syzbot USB fuzzer

   - Fix for UAC2 extension unit parser

   - Quirks for Line6 Helix, Emgaic Unitor 8

  FireWire:

   - Lots of code refactoring and fixes in most of its components"

* tag 'sound-5.3-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound: (626 commits)
  ALSA: firewire-lib: code refactoring for local variables
  ALSA: firewire-lib: code refactoring for post operation to data block counter
  ALSA: firewire-lib: code refactoring for error path of parser for CIP header
  ALSA: firewire-lib: fix different data block counter between probed event and transferred isochronous packet
  ALSA: firewire-lib: fix initial value of data block count for IR context without CIP_DBC_IS_END_EVENT
  ALSA: firewire-lib/fireface: fix initial value of data block counter for IR context with CIP_NO_HEADER
  ALSA: firewire-lib: fix invalid length of rx packet payload for tracepoint events
  ALSA: usb-audio: fix Line6 Helix audio format rates
  firewire-motu: fix wrong reference count for stream functionality at error path of rawmidi interface
  ALSA: firewire-digi00x: fix wrong reference count for stream functionality at error path of rawmidi interface
  ALSA: dice: fix wrong reference count for stream functionality at error path of rawmidi interface
  ALSA: oxfw: fix wrong reference count for stream functionality at error path of rawmidi interface
  ALSA: fireworks: fix wrong reference count for stream functionality at error path of rawmidi interface
  ALSA: bebob: fix wrong reference count for stream functionality at error path of rawmidi interface
  ASoC: SOF: Intel: implement runtime idle for CNL/APL
  ASoC: SOF: add runtime idle callback
  ASoC: hdac_hdmi: report codec link up/down status to bus
  ASoC: SOF: debug: fix possible memory leak in sof_dfsentry_write()
  ASoC: sunxi: sun50i-codec-analog: Add earpiece
  ASoC: rt5665: remove redundant assignment to variable idx
  ...

1 files changed, 29 insertions, 0 deletions

diff --git a/sound/soc/meson/axg-tdm-formatter.c b/sound/soc/meson/axg-tdm-formatter.c
index 0c6cce5c5773..2e498201139f 100644
--- a/sound/soc/meson/axg-tdm-formatter.c
+++ b/sound/soc/meson/axg-tdm-formatter.c
@@ -7,6 +7,7 @@
 #include <linux/module.h>
 #include <linux/of_platform.h>
 #include <linux/regmap.h>
+#include <linux/reset.h>
 #include <sound/soc.h>
 
 #include "axg-tdm-formatter.h"
@@ -20,6 +21,7 @@ struct axg_tdm_formatter {
 	struct clk *lrclk;
 	struct clk *sclk_sel;
 	struct clk *lrclk_sel;
+	struct reset_control *reset;
 	bool enabled;
 	struct regmap *map;
 };
@@ -76,6 +78,24 @@ static int axg_tdm_formatter_enable(struct axg_tdm_formatter *formatter)
 		return 0;
 
 	/*
+	 * On the g12a (and possibly other SoCs), when a stream using
+	 * multiple lanes is restarted, it will sometimes not start
+	 * from the first lane, but randomly from another used one.
+	 * The result is an unexpected and random channel shift.
+	 *
+	 * The hypothesis is that an HW counter is not properly reset
+	 * and the formatter simply starts on the lane it stopped
+	 * before. Unfortunately, there does not seems to be a way to
+	 * reset this through the registers of the block.
+	 *
+	 * However, the g12a has indenpendent reset lines for each audio
+	 * devices. Using this reset before each start solves the issue.
+	 */
+	ret = reset_control_reset(formatter->reset);
+	if (ret)
+		return ret;
+
+	/*
 	 * If sclk is inverted, invert it back and provide the inversion
 	 * required by the formatter
 	 */
@@ -306,6 +326,15 @@ int axg_tdm_formatter_probe(struct platform_device *pdev)
 		return ret;
 	}
 
+	/* Formatter dedicated reset line */
+	formatter->reset = reset_control_get_optional_exclusive(dev, NULL);
+	if (IS_ERR(formatter->reset)) {
+		ret = PTR_ERR(formatter->reset);
+		if (ret != -EPROBE_DEFER)
+			dev_err(dev, "failed to get reset: %d\n", ret);
+		return ret;
+	}
+
 	return devm_snd_soc_register_component(dev, drv->component_drv,
 					       NULL, 0);
 }