<aka_[m]>
bryanodonoghue: do you remember anything regarding 8939 and power domains?
<aka_[m]>
I can recall seeing some mention of MX domain supporting bit higher state than CX and im almost sure thats stuff also happen on 8976
<bryanodonoghue>
hrmm doesn't ring a bell
<bryanodonoghue>
"Vdd_Mx must be greater than or equal to Vdd_Cx at all times"
<bryanodonoghue>
if that helps
<bryanodonoghue>
the RPM should enforce that
<bryanodonoghue>
I'd see no logical reason to depart from that truism on other SoCs
<aka_[m]>
im still thinking on reboot on 8976 when wcnss core vote for max state which now is set to be turbo_hgh
<aka_[m]>
Konrad noticed ds only mentions turbo_high in notes for CCI,DDR
<bryanodonoghue>
OnePlus One bacon is an 8976 right ?
<bryanodonoghue>
Ahh... CCI
<konradybcio>
bryanodonoghue: does the "MX >= CX" hold for 8939 or all SMD RPM socs?
<aka_[m]>
konradybcio: tho in cpu code there is no turbo_high macro
<konradybcio>
if the latter, we should enforce it in the rpmpd driver
<bryanodonoghue>
You need CPR
<z3ntu>
OnePlus One (bacon) is msm8974
<aka_[m]>
i had some feeling of cpr also using these voltage levels
<aka_[m]>
so it could be used on cpr controlled apcX regulator and not on rpmpd one
<bryanodonoghue>
Yeah but if you are ramping the CCI frequency
<bryanodonoghue>
without upping the voltage
<bryanodonoghue>
likely your root-cause
<bryanodonoghue>
konradybcio IDK
<aka_[m]>
not exactly sure what's all this about but mine issue is if rpmpd have turbo_high defined like now, and wcnss code starts bumping all domains to current max state it reboot
<bryanodonoghue>
But I guess it should hold
<aka_[m]>
if i drop it to turbo its fine
<bryanodonoghue>
Not sure how easy an experiment it would be for you to run
<aka_[m]>
i had some idea that CX cannot be voted for turbo_high
<bryanodonoghue>
but set the CCI CPR voltage to max
<aka_[m]>
but MX could
<bryanodonoghue>
and then mess with the rpm*
<aka_[m]>
CCI CPR reffers to main supply for block?
<aka_[m]>
it appears to be shared with APC0
<bryanodonoghue>
so... turbo_high is defined for 8976 wcnss downstream ?
<bryanodonoghue>
or you are trying to set ?
<aka_[m]>
bryanodonoghue: ds is capping states for wcnss and any remoteproc to turbo
<bryanodonoghue>
Sorry mentioning the CCI just sets me off - because on 8939 the AP needs to do the voltage ramping not the RPM
<aka_[m]>
i cannot find any binding with RPM_SMD_LEVEL_TURBO_HIGH
<bryanodonoghue>
Soo...
<bryanodonoghue>
I'd say its not supported
<aka_[m]>
only CCI devfreq have some lack of notes in freqtables
<aka_[m]>
or notes like < 614400 >; /* STURBO */
<aka_[m]>
bryanodonoghue: i would just send commit decreasing max state but don't want then have others yell at me because it somehow happens MX does support that
<aka_[m]>
so i had idea of changing wcnss code and just trying setting turbo_high on MX only and noticing if it will crash