Hi,
Thanks for getting back to me so quickly.

On 30/07/2019 13:41, Charles Keepax wrote:

On Tue, Jul 30, 2019 at 01:09:37PM +0100, Thomas Preston wrote:

quoted

Add a debugfs device node which initiates the turn-on diagnostic routine
feature of the TDA7802 amplifier. The four status registers (one per
channel) are returned.

Signed-off-by: Thomas Preston <redacted>
---
Changes since v1:
- Rename speaker-test to (turn-on) diagnostics
- Move turn-on diagnostic to debugfs as there is no standard ALSA
  interface for this kind of routine.

 sound/soc/codecs/tda7802.c | 186 ++++++++++++++++++++++++++++++++++++-
 1 file changed, 185 insertions(+), 1 deletion(-)

+static int tda7802_bulk_update(struct regmap *map, struct reg_update *update,
+		size_t update_count)
+{
+	int i, err;
+
+	for (i = 0; i < update_count; i++) {
+		err = regmap_update_bits(map, update[i].reg, update[i].mask,
+				update[i].val);
+		if (err < 0)
+			return err;
+	}
+
+	return i;
+}

This could probably be removed using regmap_multi_reg_write.

The problem is that I want to retain the state of the other bits in those
registers. Maybe I should make a copy of the backed up state, set the bits
I want to off-device, then either:

1. Write the changes with regmap_multi_reg_write
2. Write all six regs again (if my device doesn't support the multi_reg)

quoted

+static int tda7802_probe(struct snd_soc_component *component)
+{
+	struct tda7802_priv *tda7802 = snd_soc_component_get_drvdata(component);
+	struct device *dev = &tda7802->i2c->dev;
+	int err;
+
+	tda7802->debugfs = debugfs_create_dir(dev_name(dev), NULL);
+	if (IS_ERR_OR_NULL(tda7802->debugfs)) {
+		dev_info(dev,
+			"Failed to create debugfs node, err %ld\n",
+			PTR_ERR(tda7802->debugfs));
+		return 0;
+	}
+
+	mutex_init(&tda7802->diagnostic_mutex);
+	err = debugfs_create_file("diagnostic", 0444, tda7802->debugfs, tda7802,
+			&tda7802_diagnostic_fops);
+	if (err < 0) {
+		dev_err(dev,
+			"debugfs: Failed to create diagnostic node, err %d\n",
+			err);
+		goto cleanup_diagnostic;
+	}

You shouldn't be failing the driver probe if debugfs fails, it
should be purely optional.

Got it, thanks.