On Fri, Sep 13, 2024 at 06:28:34PM -0300, Geraldo Nascimento wrote:
On Fri, Sep 13, 2024 at 10:12:39PM +0100, Hugh Cole-Baker wrote:
I added printk to show the value of AUD_CONF2, and found that on 6.1.23, the value is 0 before the code in this patch sets the insert_pcuv bit. On 6.10.3 the value is 4, i.e. insert_pcuv is already set.
According to the RK3399 TRM, the value-after-reset of the insert_pcuv bit is 1, so apparently on the 6.1.23 kernel something is clearing the bit after HW reset but before this driver sets the hw_params, and this patch sets it back to the correct value. On 6.10.3 the bit is not cleared, i.e. this patch is seemingly no longer necessary (but is a harmless no-op).
Hi Hugh,
Thank you for your extensive testing. It seems then there's no action we need to take for mainline, as it's already fixed there.
Unless Neil wants to pick-up it up for Stable? Neil, although not a regression, this is definitely a show-stopper for sound on RK3399 for older, still supported kernels. And thanks to Hugh detailed report we now have confirmation that this happens on vanilla RK3399 and is not a quirk of my Rock Pi N10 board or sink.
Thanks, Geraldo Nascimento