STM32MP15 ECO 5.0.0 OPTEE Panic set_clock_then_voltage

[debugging]

On STM32MP1 discrete power.  512MB DDR , TF-A & OPTEE configure in SYSRAM

E/TC:0 0 Panic at core/arch/arm/plat-stm32mp1/cpu_opp.c:106 <set_clock_then_voltage>

Log attached.

This message seems to unique for stm32mp1 in OP-TEE  code. no other platform source file seems to use this message.

Looking further this cpu_opp.c file is not part of OP-TEE on github in ST's stm32mp1 , thus ST unique and added afterwards to the ECO system. Seems to be patch. So that anyone familiar, outside ST, with what this code is doing is probably zero. The source does not reveal much about what is is doing. (no comments)  w/o support this may take much more time to debug.

Move the regulator part in the DT (in root USER area) after the CPU clocks. Same result. what does ST's code expect ?

Added some debug traces:

D/TC:0 0 stm32mp1_cpu_opp_get_dt_subnode:296 Found OPP 0 (650000kHz/1200mV) from DT
D/TC:0 0 stm32mp1_cpu_opp_get_dt_subnode:302 DEBUG: clk_cpu: 650000000
D/TC:0 0 stm32mp1_cpu_opp_get_dt_subnode:303 DEBUG: freq_khz: 650000
D/TC:0 0 stm32mp1_cpu_opp_get_dt_subnode:313 DEBUG: freq_khz*1000 <= clk_cpu. calling set_clock_then_voltage
D/TC:0 0 set_clock_then_voltage:101 DEBUG: opp_set_voltage 1200
D/TC:0 0 set_clock_then_voltage:108 DEBUG: current_opp = 1
D/TC:0 0 set_clock_then_voltage:109 DEBUG: Going to panic due to current_opp = cpu_opp.opp_count....
E/TC:0 0 Panic at core/arch/arm/plat-stm32mp1/cpu_opp.c:110 <set_clock_then_voltage>
E/TC:0 0 TEE load address @ 0xde000000

E/TC:0 0 Call stack:
E/TC:0 0 0xde00e703
E/TC:0 0 0xde03da89
E/TC:0 0 0xde013d3b
E/TC:0 0 0xde0142bf
E/TC:0 0 0xde014397
E/TC:0 0 0xde03c3ab
E/TC:0 0 0xde03c799
E/TC:0 0 0xde03ccc7
E/TC:0 0 0xde03f5e9
E/TC:0 0 0xde00e425
E/TC:0 0 0xde00e52b
E/TC:0 0 0xde00019c
M/TC: CPU : 0
M/TC: usr_sp : 0x50044110
M/TC: usr_lr : 0x2080a000
M/TC: irq_spsr : 0x00071500
M/TC: irq_sp : 0x9200a48a
M/TC: irq_lr : 0x08500012
M/TC: fiq_spsr : 0xa28baa02
M/TC: fiq_sp : 0x50414465
M/TC: fiq_lr : 0x80282200
M/TC: svc_spsr : 0x00514014
M/TC: svc_sp : 0x0ac00264
M/TC: svc_lr : 0x41015450
M/TC: abt_spsr : 0xa0080800
M/TC: abt_sp : 0x55145c50
M/TC: abt_lr : 0x0280a280
M/TC: und_spsr : 0x3080c400
M/TC: und_sp : 0x00a8802a
M/TC: und_lr : 0x50454511
M/TC: pmcr : 0x2020a080

A voltage check if failing: it seems OPT-EE is expecting 1.2V instead of 1.35V for core ?

 

 

 

 

[debugging]

Removed the CPU nodes in USER section, mentioned in workshop material, and that that causes an error even earlier in the boot. it seems the CPU nodes are needed in the DT. Perhaps they are missing some information ?

&cpu0 {
cpu-supply = <&vddcore>;
};
&cpu1 {
cpu-supply = <&vddcore>;
};

[debugging]

For some reason it requites this for vdd_core..

regulator-min-microvolt = <1200000>; 

[Olivier GALLIEN]

Hi @debugging ,
Yes you are right. 

You are using STM32MP157AAA3 which doesn't support overdrive mode @1.35V. 
And even on part supporting it, this mode is not set in BSP but only at kernel level through opp_mode mechanism.
How to change the CPU frequency - stm32mpu

Olivier 

 

[michiel.tjampens]

If not supporting overdrive at 1,35V is the reason, shouldn't setting regulator-max-microvolt at 1350000 also be an issue?

[debugging]

Thank you for pointing out the mechanism  for the issue.

It make sens Kernel manages speed and voltage setting, it would be similar to IPMI/ACPI.  Though since this is a discrete board. Some questions:

1/ it seems OPTEE knows it needs  1.2V for this STM32MP157AAAC1 @ 650Mhz , Why would it have to panic. This DT entry could as well be removed or ignore instead of causing a panic. The code could also give a more explicit.For example" "Vdd_voltage core in DT not matching expected core voltage of X.XX". since this codes is ST specific, it is full under control of ST to add some improvements.

2/ How does one know which entries (for MPU power rails) to set in the DT for a specific MPU  in TF-A, OPTEE and Uboot as well as the kernel in the "USER" areas..? Not only for power but also for clocks among others. A kind of "decision tree" based on the characteristic of the MPU and PMIC/discrete power would really make things much easier than reading through thousands of lines in WIKI's and PDF's to bring up a minimum viable  configuration with and USB/SD boot  and UART (serial debug.input I/O) to Linux.

3/ The informal trouble shooting guide and bring up WIKI for MP1 is more than 4 years old. Since then there should be many issues ST partners and customers walked into  On the STM CN site there is a short one: https://wiki.stmicroelectronics.cn/stm32mpu/index.php?title=Visual_troubleshooting_grid&oldid=62487  but the link in that page to the "latest" , results in Page not found" Could ST put some effort in creating such guide ? Of-course not all situations can be covered. 20% of the  of most common issues and common solutions could save 80% of developer's time. Many common issues can be found in the community If those could be linked to of most common causes, it may help a lot. And when new one's arise the page could be updated. "Common" should be al least 2 users having the same issue.