Searching for tester of Linux Kernel drivers

[swahren]

And i must confess that i didn't test with IRQCHIP_SKIP_SET_WAKE flag. Since you describe a test scenario i should be able to test it and add a patch.

[joerg.krause]

I noticed another issue with the SAIF of the i.MX28. When the CPU goes into supsend mode, the MCLK and BCLK pins are going sometimes into a HIGH state and sometimes into a LOW state. Not sure what the expected behavior should be, so I started a question on the NXP community platform: https://community.nxp.com/message/818211

[swahren]

Unfortunately it seems that i toasted my Olinuxino Maxi, so i wasn't able to test my last change IRQCHIP_SKIP_SET_WAKE with i.MX23 and GPIO wake up. Did it work for you (already commited the change to my repo)?

It don't have access to a board with SAIF. Sorry for the naive question but what's the problem behind the random pin level in suspend?
Higher power consumption?
Are you able to verify that mxs_saif_trigger is called with SNDRV_PCM_TRIGGER_SUSPEND in case of suspend? In this case i would expect that both clocks are going to be disabled.

I think in case we need a proper solution the alsa mailing list and Fabio Estevam from NXP should be contacted.

Btw: You should mention in your question to NXP community that your using Mainline Linux or does the Vendor Kernel show the same behavior?

[joerg.krause]

Unfortunately it seems that i toasted my Olinuxino Maxi, so i wasn't able to test my last change IRQCHIP_SKIP_SET_WAKE with i.MX23 and GPIO wake up.

Oops! That's bad!

Did it work for you (already commited the change to my repo)?

Sorry, I didn't noticed. I'll check tomorrow.

It don't have access to a board with SAIF. Sorry for the naive question but what's the problem behind the random pin level in suspend?
Higher power consumption?

Yes, the DAC has an auto power-down feature which is activated if the MCLK stays LOW.

Are you able to verify that mxs_saif_trigger is called with SNDRV_PCM_TRIGGER_SUSPEND in case of suspend? In this case i would expect that both clocks are going to be disabled.

No, I haven't done yet. However, I guess what's happening. The I2S clocks are continuous clocks and each clock keeps its level it has when the CPU goes into standby mode. But, that's only guessing...

I think in case we need a proper solution the alsa mailing list and Fabio Estevam from NXP should be contacted.

Good idea! I'll ask for help on the alsa mailing list tomorrow.

Btw: You should mention in your question to NXP community that your using Mainline Linux or does the Vendor Kernel show the same behavior?

I have only tested with mainline. I'll update my post, thanks!

[joerg.krause]

Hi Stefan,

I've tested your patch which sets IRQCHIP_SKIP_SET_WAKE and it works! There is no warning printed anymore. Thanks!

Jörg

[swahren]

Okay, what's the state of the SAIF issue?

[joerg.krause]

Hi Stefan,

NXP replied that Mainline Linux is not supported and I should use the BSP :-(

I did not had the time to prepare a board with the ancient BSP and I'll be in vacation the next two weeks, so nothing new here.

[swahren]

I only want to mention that i rebased all off-tree changes on top a recent linux-next tree:

https://github.com/lategoodbye/linux-mxs-power/tree/rebase-4.12

Unfortunately i didn't had the time to test it with a Olimex iM233 board.

[lambda]

Hi Stefan,

you have there some very nice pieces of code. Any plans to
get them merged into mainline?

Thanks for your work!

[swahren]

This branch is "only" a rebase of my last attempt to get it mainline. It doesn't address all the comments for the version. Furthermore the syscon part must be replaced with a MXS specific mfd driver layer which handles the concurrent register access and the power interrupts.

Since i'm the bcm2835 co-maintainer now i've little time for this series. I would be happy in case someone could continue and finally merge it.