Using an iMX6 and SGTL5000 for audio, I noticed a 400ms delay when snd_pcm_prepare() is called. Looking at the sgtl5000.c codec driver, there is a sleep of 400ms when the power management powers up and powers down. For some reason the call to power up is always called on snd_pcm_prepare() even when the audio codec power never went down. I added a state to keep track of it in the codec driver as a workaround to not delay unless necessary. While this workaround works, I don't think it's the best solution. Does someone know why the power up section of code is always called in the driver? Thanks
+diff --git a/sound/soc/codecs/sgtl5000.c b/sound/soc/codecs/sgtl5000.c
+index 6668845..3aaf384 100644
+--- a/sound/soc/codecs/sgtl5000.c
++++ b/sound/soc/codecs/sgtl5000.c
+@@ -152,6 +152,8 @@ static int mic_bias_event(struct snd_soc_dapm_widget *w,
+ static int power_vag_event(struct snd_soc_dapm_widget *w,
+ struct snd_kcontrol *kcontrol, int event)
+ {
++ static bool is_power_up = false;
++
+ struct snd_soc_codec *codec = snd_soc_dapm_to_codec(w->dapm);
+ const u32 mask = SGTL5000_DAC_POWERUP | SGTL5000_ADC_POWERUP;
+
+@@ -159,7 +161,12 @@ static int power_vag_event(struct snd_soc_dapm_widget *w,
+ case SND_SOC_DAPM_POST_PMU:
+ snd_soc_update_bits(codec, SGTL5000_CHIP_ANA_POWER,
+ SGTL5000_VAG_POWERUP, SGTL5000_VAG_POWERUP);
+- msleep(400);
++
++ if (is_power_up == false)
++ {
++ msleep(400);
++ is_power_up = true;
++ }
+ break;
+
+ case SND_SOC_DAPM_PRE_PMD:
+@@ -173,6 +180,7 @@ static int power_vag_event(struct snd_soc_dapm_widget *w,
+ snd_soc_update_bits(codec, SGTL5000_CHIP_ANA_POWER,
+ SGTL5000_VAG_POWERUP, 0);
+ msleep(400);
++ is_power_up = false;
+ }
+ break;
+ default:
