Hi Takashi,
On Tue, 2018-08-21 at 14:17 +0900, Takashi Sakamoto wrote:
+static int aaf_mclk_timeout_capture(snd_pcm_aaf_t *aaf) +{
ssize_t n;uint64_t expirations;n = read(aaf->timer_fd, &expirations, sizeof(uint64_t));if (n < 0) {SNDERR("Failed to read() timer");return -errno;}if (expirations != 1)pr_debug("Missed %llu presentation time(s) ",expirations - 1);
while (expirations--) {snd_pcm_sframes_t len;aaf_inc_ptr(&aaf->hw_ptr, aaf->frames_per_pkt, aaf-boundary);
len = aaf->hw_virt_ptr - aaf->hw_ptr;if (len < 0)len += aaf->boundary;if (len > aaf->buffer_size) {/* If the distance between hw virtual pointerand hw
* pointer is greater than the buffer size,it means we
* had an overrun error so -EPIPE isreturned.
*/return -EPIPE;}aaf->mclk_ticks++;}return 0;+}
In your code, -EPIPE can be returned in a call of 'snd_pcm_poll_revents()'. In this case, typical applications follow recovery process below:
->snd_pcm_poll_revents() ->snd_pcm_prepare() ->snd_pcm_start()
In this case, status of corresponding PCM substream is reset (see 'snd_pcm_ioplug_prepare()'). In my opinion, in this case, backend is expected to stop data streaming, then initialize hardware and restart data streaming. In your driver, timer_fd/sk_fd is leaked. It's better to release them by checking status of PCM substream in .prepare callback.
I think you're right. I'll fix it.
Thank you,
Andre