On Sat, 30.08.08 09:38, Kevin Gilbert (kevin952 at tpg.com.au) wrote: > > Umpf. there seems to be something very broken here in the audio driver > > or alsa-libs. We only support interleaved channels. This assert above > > is always true for interleaved channels. We ask the ALSA driver for > > interleaved channels. But apparently it gave us some more complex > > channel arrangement. > > > > Please report to the ALSA people. But first try to update your kernel > > driver and alsa-libs to newest from upstream. > > Before reporting to the ALSA folk I thought I should check with the PS3 kernel > people as I am running one of their kernels (no use reporting a problem to > ALSA if there is PS3 kernel specific code that could be the problem). > > See http://ozlabs.org/pipermail/cbe-oss-dev/2008-August/005423.html for that > thread. > > In short: > > "PS3 sound driver supports only SNDRV_PCM_INFO_NONINTERLEAVED format as well > as many ALSA driver like AC97 support only SNDRV_PCM_INFO_INTERLEAVED. Non- > interleaved format is just not usual, but it is not broken. It's defined in > ALSA. With a brief looking, other mainlined drivers -- gus, emu10k, es1968, > rme9652 and aica -- seem to support only non-interleaved samples. [...] PS3 > sound chip's hardware is not suitable for the interleaved samples due to its > FIFO and DMAC design. I have no plan to support interleave permutation in the > kernel driver level. I think the best way is that the user space renderer > takes care of the various interleave formats as well as sample endianness and > bit width." The thing is that we try to enable SND_PCM_ACCESS_MMAP_INTERLEAVED and when that fails SND_PCM_ACCESS_RW_INTERLEAVED. Which apparently suceeded on your setup, since you got past that initialization. You subsequently hit an assert when we checked whether the sample layout actually matches what SND_PCM_ACCESS_xxx_INTERLEAVED promises. It's totally fine to support only NONINTERLEAVED channel setups, but then the alsa driver needs to refuse enabling INTERLEAVED mode, i.e. snd_pcm_hw_params_set_access() must fail -- which it doesn't. The relevant code in PA you find here: http://pulseaudio.org/browser/src/modules/alsa-util.c#L324 Lennart -- Lennart Poettering Red Hat, Inc. lennart [at] poettering [dot] net ICQ# 11060553 http://0pointer.net/lennart/ GnuPG 0x1A015CC4
