From 39aa1cf94d43efaa80571b3f9d3dc56297f2f5be Mon Sep 17 00:00:00 2001 From: Lennart Poettering Date: Fri, 31 Jul 2009 02:07:24 +0200 Subject: alsa: revert to first set number of periods, then set buffer size Apparently some ALSA drivers aren't happy with getting the buffer size configured first followed the period size. So swap the order again and document this for future reference so that we don't turn that around again. --- src/modules/alsa/alsa-util.c | 55 +++++++++++++++++++++++--------------------- 1 file changed, 29 insertions(+), 26 deletions(-) diff --git a/src/modules/alsa/alsa-util.c b/src/modules/alsa/alsa-util.c index 1f3e5dcd..a47a8958 100644 --- a/src/modules/alsa/alsa-util.c +++ b/src/modules/alsa/alsa-util.c @@ -233,14 +233,16 @@ int pa_alsa_set_hw_params( goto finish; } - if (_period_size && tsched_size && _periods) { + if (_period_size > 0 && tsched_size > 0 && _periods > 0) { + snd_pcm_uframes_t buffer_size; + unsigned int p; /* Adjust the buffer sizes, if we didn't get the rate we were asking for */ _period_size = (snd_pcm_uframes_t) (((uint64_t) _period_size * r) / ss->rate); tsched_size = (snd_pcm_uframes_t) (((uint64_t) tsched_size * r) / ss->rate); if (_use_tsched) { - snd_pcm_uframes_t buffer_size = 0; + buffer_size = 0; if ((ret = snd_pcm_hw_params_get_buffer_size_max(hwparams, &buffer_size)) < 0) pa_log_warn("snd_pcm_hw_params_get_buffer_size_max() failed: %s", pa_alsa_strerror(ret)); @@ -251,32 +253,33 @@ int pa_alsa_set_hw_params( _periods = 1; } - if (_period_size > 0 && _periods > 0) { - snd_pcm_uframes_t buffer_size; - - buffer_size = _periods * _period_size; - - if ((ret = snd_pcm_hw_params_set_buffer_size_near(pcm_handle, hwparams, &buffer_size)) < 0) - pa_log_info("snd_pcm_hw_params_set_buffer_size_near() failed: %s", pa_alsa_strerror(ret)); - } - - if (_periods > 0) { - - /* First we pass 0 as direction to get exactly what we - * asked for. That this is necessary is presumably a bug - * in ALSA. All in all this is mostly a hint to ALSA, so - * we don't care if this fails. */ - - dir = 0; - if (snd_pcm_hw_params_set_periods_near(pcm_handle, hwparams, &_periods, &dir) < 0) { - dir = 1; - if (snd_pcm_hw_params_set_periods_near(pcm_handle, hwparams, &_periods, &dir) < 0) { - dir = -1; - if ((ret = snd_pcm_hw_params_set_periods_near(pcm_handle, hwparams, &_periods, &dir)) < 0) - pa_log_info("snd_pcm_hw_params_set_periods_near() failed: %s", pa_alsa_strerror(ret)); - } + /* Some ALSA drivers really don't like if we set the buffer + * size first and the number of periods second. (which would + * make a lot more sense to me) So, follow this rule and + * adjust the periods first and the buffer size second */ + + /* First we pass 0 as direction to get exactly what we + * asked for. That this is necessary is presumably a bug + * in ALSA. All in all this is mostly a hint to ALSA, so + * we don't care if this fails. */ + + p = _periods; + dir = 0; + if (snd_pcm_hw_params_set_periods_near(pcm_handle, hwparams, &p, &dir) < 0) { + p = _periods; + dir = 1; + if (snd_pcm_hw_params_set_periods_near(pcm_handle, hwparams, &p, &dir) < 0) { + p = _periods; + dir = -1; + if ((ret = snd_pcm_hw_params_set_periods_near(pcm_handle, hwparams, &p, &dir)) < 0) + pa_log_info("snd_pcm_hw_params_set_periods_near() failed: %s", pa_alsa_strerror(ret)); } } + + /* Now set the buffer size */ + buffer_size = _periods * _period_size; + if ((ret = snd_pcm_hw_params_set_buffer_size_near(pcm_handle, hwparams, &buffer_size)) < 0) + pa_log_info("snd_pcm_hw_params_set_buffer_size_near() failed: %s", pa_alsa_strerror(ret)); } if ((ret = snd_pcm_hw_params(pcm_handle, hwparams)) < 0) -- cgit