Apalis iMX6 RESET_MOCI does not get deasserted

Hi,

Was there any change in 3.0b4 regarding RESET_MOCI pin handling? It doesn’t get deasserted if linux gets shutdown by shutdown -P now command. Tried this with 2.8b7 image version and RESET_MOCI signal gets deasserted when linux shuts down but that’s not the case in 3.0b4 image.

Regards, Sandi Mlinar

Greetings @mlinar,

It seems that there was a change on the behavior of the RESET_MOCI pin between the BSP versions.
Reasoning here is that RESET_MOCI# is controlled by two signals, one from the PMIC and ENET_TX_EN. The idea would be that a poweroff turns off the PMIC and that brings RESET_MOCI# to 0, thus resetting the board. With the newer kernel on newer BSPs, that behavior seems to have been lost. That’s why RESET_MOCI# remains at logic level high.

I’m checking with the BSP team if we have any fix or workaround for this.

Hi @gustavo.tx

Thanks for the info,

So the case here is that we have an ups system on the carrier board which relies on RESET_MOCI signal. So when power is lost, the system goes into shutdown and when we detect RESET_MOCI low level we can shutdown power for CPU also. So this signal and RESET_MOCI functionality are quite important for us.

So if you have any patch or other workaround for this that would be great.

Side question related to this topic and to confirm I am doing it right: disabling device tree pcie node has no impact on RESET_MOCI functionality? I have disabled it in my device tree since we are not using PCIE.

Regards, Mlinar

Hi @mlinar,

I’m escalating this with the BSP team, we should have a proper answer soon.

Regarding the PCIe node, I don’t think it should have an impact there considering you’re not using the peripheral.

Hi @gustavo.tx

Do you have any new info regarding this topic?

I noticed that this issue was reported in backlog.
Also noticed the last four commits on 4.14-2.3 branch, are those four commits fix for this?

Regards, Mlinar

Hi @mlinar!

Exactly, the BSP team created that ticket for this specific issue.

Also, I’m almost sure that these four commits are related to this behavior. I kindly ask you to clone our kernel repository on that branch, build, test and see if that fixes that for you already. Here’s our developer article on building the kernel in case you need: Build U-Boot and Linux Kernel from Source Code | Toradex Developer Center.

Hi @gustavo.tx,

I can confirm that those commits fixed the RESET_MOCI signal not getting low on shutdown.

But the condition is, that PCI support must be enabled in kernel configuration and in device tree file (pcie node in Ixora carrier board case). If either of these is disabled (because of optimization) then the RESET_MOCI does not get high on system start. It gets high when the system shuts down if called shutdown -P.

Regards, Mlinar

Hi @mlinar!

Thanks for your feedback!

Is it viable for you to keep PCIe support enabled, then? I don’t think it will be much of an overhead in most cases.

Thanks for your feedback!

Your welcome.

Is it viable for you to keep PCIe support enabled, then?

Yes, we will keep PCIe enabled for this to work.

I don’t think it will be much of an overhead in most cases.

I know it probably isn’t so much of an overhead, but if you are optimizing boot times, sizes, etc., then you want to disable all the things that your system doesn’t use.

Regards, Mlinar

@mlinar,

Perfect, thank you!