Hi Shengjiu Wang,
On Thu, Jan 16, 2025 at 03:30:55PM +0800, Shengjiu Wang wrote:
On Thu, Jan 16, 2025 at 3:24 PM Stefan Eichenberger eichest@gmail.com wrote:
Hi Shengjiu Wang,
On Thu, Jan 16, 2025 at 12:01:13PM +0800, Shengjiu Wang wrote:
On Wed, Jan 15, 2025 at 6:39 PM Stefan Eichenberger eichest@gmail.com wrote:
On Wed, Jan 15, 2025 at 05:14:09PM +0800, Shengjiu Wang wrote:
On Wed, Jan 15, 2025 at 4:33 PM Stefan Eichenberger eichest@gmail.com wrote:
Hi Shengjiu Wang,
On Tue, Jan 14, 2025 at 12:58:24PM +0100, Stefan Eichenberger wrote: > Hi Shengjiu Wang, > > On Tue, Jan 14, 2025 at 03:49:10PM +0800, Shengjiu Wang wrote: > > On Mon, Jan 13, 2025 at 5:54 PM Stefan Eichenberger eichest@gmail.com wrote: > > > > > > From: Stefan Eichenberger stefan.eichenberger@toradex.com > > > > > > Currently, the flags for the ACM clocks are set to 0. This configuration > > > causes the fsl-sai audio driver to fail when attempting to set the > > > sysclk, returning an EINVAL error. The following error messages > > > highlight the issue: > > > fsl-sai 59090000.sai: ASoC: error at snd_soc_dai_set_sysclk on 59090000.sai: -22 > > > imx-hdmi sound-hdmi: failed to set cpu sysclk: -22 > > > > The reason for this error is that the current clock parent can't > > support the rate > > you require (I think you want 11289600). > > > > We can configure the dts to provide such source, for example: > > > > &sai5 { > > + assigned-clocks = <&acm IMX_ADMA_ACM_SAI5_MCLK_SEL>, > > + <&acm IMX_ADMA_ACM_AUD_CLK1_SEL>, > > + <&clk IMX_SC_R_AUDIO_PLL_0 IMX_SC_PM_CLK_PLL>, > > + <&clk IMX_SC_R_AUDIO_PLL_0 IMX_SC_PM_CLK_SLV_BUS>, > > + <&clk IMX_SC_R_AUDIO_PLL_0 IMX_SC_PM_CLK_MST_BUS>, > > + <&clk IMX_SC_R_AUDIO_PLL_1 IMX_SC_PM_CLK_PLL>, > > + <&clk IMX_SC_R_AUDIO_PLL_1 IMX_SC_PM_CLK_SLV_BUS>, > > + <&clk IMX_SC_R_AUDIO_PLL_1 IMX_SC_PM_CLK_MST_BUS>, > > + <&sai5_lpcg 0>; > > + assigned-clock-parents = <&aud_pll_div0_lpcg 0>, <&aud_rec1_lpcg 0>; > > + assigned-clock-rates = <0>, <0>, <786432000>, <49152000>, <12288000>, > > + <722534400>, <45158400>, <11289600>, > > + <49152000>; > > status = "okay"; > > }; > > > > Then your case should work. > > > > > > > > By setting the flag CLK_SET_RATE_NO_REPARENT, we signal that the ACM > > > > I don't think CLK_SET_RATE_NO_REPARENT is a good choice. which will cause > > the driver don't get an error from clk_set_rate(). > > Thanks for the proposal, I will try it out tomorrow. Isn't this a > problem if other SAIs use the same clock source but with different > rates? > > If we have to define fixed rates in the DTS or else the clock driver > will return an error, isn't that a problem? Maybe I should change the > sai driver so that it ignores the failure and just takes the rate > configured? In the end audio works, even if it can't set the requested > rate.
The following clock tree change would allow the driver to work in our scenario: &sai5 { assigned-clocks = <&acm IMX_ADMA_ACM_SAI5_MCLK_SEL>, <&clk IMX_SC_R_AUDIO_PLL_1 IMX_SC_PM_CLK_PLL>; assigned-clock-parents = <&aud_pll_div1_lpcg 0>; assigned-clock-rates = <0>, <11289600>; };
In which we configure PLL_0 for 48KHz series (8kHz/16kHz/32kHz/48kHz), PLL_1 for 44kHz series (11kHz/22kHz/44kHz), which should fit for most audio requirements.
However, this means we have to switch the parent clock to the audio pll
- For the simple setup with two SAIs, one for analog audio and one for
HDMI this wouldn't be a problem. But I'm not sure if this is a good solution if a customer would add a third SAI which requires again a different parent clock rate.
We won't change the PLL's rate in the driver, so as PLL_0 for 48kHz, PLL_1 for 44kHz, even with a third SAI or more, they should work.
One potential solution could be that the SAI driver tries to first derive the clock from the current parent and only if this fails it tries to modify its parent clock. What do you think about this solution?
I did some more testing and I'm still not happy with the current solution. The problem is that if we change the SAI5 mclk clock parent it can either support the 44kHz series or the 48kHz series. However, in the case of HDMI we do not know in advance what the user wants.
This means when testing either this works: speaker-test -D hw:2,0 -r 48000 -c 2 or this works: speaker-test -D hw:2,0 -r 44100 -c 2 With: card 2: imxaudiohdmitx [imx-audio-hdmi-tx], device 0: i.MX HDMI i2s-hifi-0 [i.MX HDMI i2s-hifi-0]
Are you using the setting below? then should not either, should both works &sai5 {
assigned-clocks = <&acm IMX_ADMA_ACM_SAI5_MCLK_SEL>,<&acm IMX_ADMA_ACM_AUD_CLK1_SEL>,<&clk IMX_SC_R_AUDIO_PLL_0 IMX_SC_PM_CLK_PLL>,<&clk IMX_SC_R_AUDIO_PLL_0 IMX_SC_PM_CLK_SLV_BUS>,<&clk IMX_SC_R_AUDIO_PLL_0 IMX_SC_PM_CLK_MST_BUS>,<&clk IMX_SC_R_AUDIO_PLL_1 IMX_SC_PM_CLK_PLL>,<&clk IMX_SC_R_AUDIO_PLL_1 IMX_SC_PM_CLK_SLV_BUS>,<&clk IMX_SC_R_AUDIO_PLL_1 IMX_SC_PM_CLK_MST_BUS>,<&sai5_lpcg 0>;assigned-clock-parents = <&aud_pll_div0_lpcg 0>, <&aud_rec1_lpcg 0>;assigned-clock-rates = <0>, <0>, <786432000>, <49152000>, <12288000>,<722534400>, <45158400>, <11289600>,<49152000>; status = "okay";};
Sorry I didn't communicate that properly. Yes I was trying with these settings but they do not work. The problem does not seem to be that the driver can not adjust the rate for the audio (so e.g. 44kHz or 48kHz) but that snd_soc_dai_set_sysclk in imx-hdmi.c fails with EINVAL. This results in a call to fsl_sai_set_mclk_rate in fsl_sai.c with clk_id 1 (mclk_clk[1]) and a freq of 11289600 which causes the fail. Interestingly, in fsl_sai_set_bclk we then only use clk_get_rate on mclk_clk[0] which is set to audio_ipg_clk (rate 175000000) and we do not use mclk_clk[1] anymore at all. This is why I'm not sure if this call to snd_soc_dai_set_syclk is really necessary?
Could you please check if you have the below commit in your test tree? 35121e9def07 clk: imx: imx8: Use clk_hw pointer for self registered clock in clk_parent_data
if not, please cherry-pick it.
The audio_ipg_clk can be selected if there is no other choice. but the rate 175000000 is not accurate for 44kHz. what we got is 44102Hz. This is the reason I don't like to use this source.
Thanks a lot for the hint, in my test setup I indeed have not applied this commit. I tested it now with cherry-picking the commit and with applying the change to the dts. This made it work. I will do some more testing tomrrow and if I can't find any addtional issue I will use this as a solution.
Should I add the change to our Apalis board dtsi file or directly to imx8qm-ss-audio.dtsi? I think it fixes kind of a general issue or not?
Thanks for your support Stefan