Now that regmap_read_bypassed() has been added to the kernel it is
preferable to wake the device with a read rather than a write as the
utility function can be called at a time before the device has been
identified.
Signed-off-by: Simon Trimmer <simont@xxxxxxxxxxxxxxxxxxxxx>
---
sound/soc/codecs/cs35l56-shared.c | 19 +++++--------------
1 file changed, 5 insertions(+), 14 deletions(-)
diff --git a/sound/soc/codecs/cs35l56-shared.c b/sound/soc/codecs/cs35l56-shared.c
index e7e8d617da94..91b3c1c8575c 100644
--- a/sound/soc/codecs/cs35l56-shared.c
+++ b/sound/soc/codecs/cs35l56-shared.c
@@ -450,32 +450,23 @@ static const struct reg_sequence cs35l56_hibernate_seq[] = {
REG_SEQ0(CS35L56_DSP_VIRTUAL1_MBOX_1, CS35L56_MBOX_CMD_ALLOW_AUTO_HIBERNATE),
};
-static const struct reg_sequence cs35l56_hibernate_wake_seq[] = {
- REG_SEQ0(CS35L56_DSP_VIRTUAL1_MBOX_1, CS35L56_MBOX_CMD_WAKEUP),
-};
-
static void cs35l56_issue_wake_event(struct cs35l56_base *cs35l56_base)
{
+ unsigned int val;
+
/*
* Dummy transactions to trigger I2C/SPI auto-wake. Issue two
* transactions to meet the minimum required time from the rising edge
* to the last falling edge of wake.
*
- * It uses bypassed write because we must wake the chip before
+ * It uses bypassed read because we must wake the chip before
* disabling regmap cache-only.
- *
- * This can NAK on I2C which will terminate the write sequence so the
- * single-write sequence is issued twice.
*/
- regmap_multi_reg_write_bypassed(cs35l56_base->regmap,
- cs35l56_hibernate_wake_seq,
- ARRAY_SIZE(cs35l56_hibernate_wake_seq));
+ regmap_read_bypassed(cs35l56_base->regmap, CS35L56_IRQ1_STATUS, &val);
usleep_range(CS35L56_WAKE_HOLD_TIME_US, 2 * CS35L56_WAKE_HOLD_TIME_US);
- regmap_multi_reg_write_bypassed(cs35l56_base->regmap,
- cs35l56_hibernate_wake_seq,
- ARRAY_SIZE(cs35l56_hibernate_wake_seq));
+ regmap_read_bypassed(cs35l56_base->regmap, CS35L56_IRQ1_STATUS, &val);
cs35l56_wait_control_port_ready();
}
--
2.43.0
