hostboot icon indicating copy to clipboard operation
hostboot copied to clipboard

What is the purpose of this commit

Open Grubby0624 opened this issue 1 year ago • 2 comments

https://github.com/open-power/hostboot/blob/1e3984c6540ba6d327d0c6f7aaccd783647ef27f/src/import/chips/ocmb/common/procedures/hwp/pmic/lib/utils/pmic_enable_utils.C#L476

I have a few questions about this commit: 43fd55aa243fe570b989da8f8ff9fcb893c7d218 "Added recovery procedure to DDR4 pmic_n_made_detect": 1. What is the purpose of this commit? If I don't add this commit and restart the machine after setting the PMIC output voltage bias, it can start up normally. However, if I add this commit, the I2C link to this PMIC will report an I2C No ACK error after booting up 2. I am using TPS53830 PMIC, and I understand that RA3 represents the I2C address for this model of PMIC. Why is its value changed to 0x00 here? Thanks!!

Grubby0624 avatar Jul 29 '24 03:07 Grubby0624

I think there was a typo in your commit hash above. Did you mean 43fd55aa243fe570b989da8f8bf9fcb893c7d218? If so, the commit is adding some logic to recover a dead PMIC at runtime. It has been working fine for us. Where in the IPL do you see the failure?

I'm asking around about the second question. If that is a different PMIC model than we use then our code may not support it. In our case RA3 seems to be just a scratch register with no functional use.

dcrowell77 avatar Aug 06 '24 13:08 dcrowell77

Are you testing with DDIMMs or some other configuration? Do you have DDR4 or DDR5 memory? It appears that you are going down a DDR4 code path using a PMIC that we only support on the DDR5 DDIMMs. There are some implicit assumptions about which PMIC models are supported for which DDR versions that you seem to be violating.

dcrowell77 avatar Aug 06 '24 14:08 dcrowell77