Dne 18.8.2018 v 11:42 Robert Jarzmik napsal(a):
petrcvekcz@gmail.com writes:
From: Petr Cvek petrcvekcz@gmail.com
The switch of PXA to dmaengine caused a regression for devices which are using SSP/I2S and probably AC97. The problem is in multiple allocation of DMA by function pxa2xx_soc_pcm_new() called by pxa-pcm-audio and (SSP/I2S/AC97). The regression was found on HTC Magician and after playback kernel generates an oops:
With AC97, this is still continuously tested and works fine, and the pcm part is what makes aplay happy for playing music. This is used both in a devicetree and non devicetree build, which makes me thing the AC97 part is fine, and the SSP/I2S needs some love.
I was looking into the code and if there is no .platform_name present soc_bind_dai_link() will just use "snd-soc-dummy" as the name.
With your patch, a compilation of mioa701 fails with [1].
My bad I've forgot to add "select SND_PXA2XX_LIB" from removed SND_PXA2XX_SOC, now it should compile, can you test it if it works now? fixed patch in [1].
BTW can you try insmod-aplay-stop-rmmod with the current -next on your board?
Therefore, I'll NAK this patch, as removing SND_PAX2XX_SOC config is a bad idea.
As a way forward, please submit a patch for magician only, which makes it work again (by work I mean you can play and hear a song with aplay or equivalent).
Then hopefully Daniel can interpolate to all the SSP/I2S platforms. And maybe Daniel can test on his platform too so that we're safe on both pxa2xx and pxa3xx side.
OK I can do that, (I can speak only for I2S and SSP) but if fixed patch work it should be fine.
best regards Petr
Attached: [1] Fixed patch v2, forgotten "select SND_PXA2XX_LIB"