On Wed, Nov 19, 2014 at 10:52:45AM -0800, Kenneth Westfield wrote:
+ .periods_min = LPASS_MI2S_NO_OF_PERIODS,
+ .periods_max = LPASS_MI2S_NO_OF_PERIODS,
As Pierre said this is really odd - it appears to be just a single
possible value.
+static irqreturn_t lpass_pcm_mi2s_irq(int intrsrc, void *data)
+{
+ int dma_ch;
+ uint32_t ret = IRQ_NONE;
+ uint32_t has_xrun, pending;
+
+ struct snd_pcm_substream *substream = data;
+ struct snd_pcm_runtime *runtime = substream->runtime;
+ struct lpass_runtime_data_t *prtd = runtime->private_data;
Here we rely on runtime.
+
+ if (prtd) {
+ dma_ch = prtd->lpaif_info.dma_ch;
+ } else {
+ pr_debug("%s: received interrupt w/o runtime\n", __func__);
+ return IRQ_NONE;
+ }
Here we print (as a debug message, not a dev_err()) an error saying we
lack a runtime (actually it's private data).
+ if (unlikely(has_xrun) && substream->runtime &&
+ snd_pcm_running(substream)) {
Here we check if runtime (which we already dereferenced) is non-NULL.
+ if (pending & LPAIF_UNDER_CH(dma_ch)) {
+ snd_pcm_period_elapsed(substream);
+ pr_debug("%s: xrun warning\n", __func__);
+ ret = IRQ_HANDLED;
+ }
+
+ if (pending & LPAIF_ERR_CH(dma_ch)) {
+ pr_debug("%s: Bus access warning\n", __func__);
+ ret = IRQ_HANDLED;
+ }
These errors should be logged as such.
+static int lpass_pcm_mi2s_hw_params(struct snd_pcm_substream *substream,
+ struct snd_pcm_hw_params *params)
+{
+ struct snd_pcm_runtime *runtime = substream->runtime;
+ struct lpass_runtime_data_t *prtd = runtime->private_data;
+
+ snd_pcm_set_runtime_buffer(substream, &substream->dma_buffer);
+ prtd->pcm_stream_info.pcm_prepare_start = 0;
+ prtd->pcm_stream_info.period_index = 0;
+ return 0;
+}
This appears to ignore the params passed so hw_params() seems like the
wrong place - open()?