pro cpu reset by JTAG

[floitsch_toit]

One of our projects uses an ESP32S3.
Sometimes we see the message `pro cpu reset by JTAG` during boot, and the device then stops shortly after. It manages to execute some instructions but it runs less than a second. This is semi-reproducible. If we wake up from sleep it happens maybe once every 20? times.

There is no JTAG connected to the board.

Where is this message printed? (I grepped through the esp-idf sources, but couldn't find it).
What does it mean?
Any ideas on how to fix/diagnose this?

Code: Select all

ESP-ROM:esp32s3-20210327
Build:Mar 27 2021
rst:0x5 (DSLEEP),boot:0x2a (SPI_FAST_FLASH_BOOT)
pro cpu reset by JTAG
SPIWP:0xee
mode:DIO, clock div:1
load:0x3fce3810,len:0x194

[ESP_Sprite]

That message comes from the internal boot ROM, it's printed when the RTC_CNTL_JTAG_RESET_FLAG_PROCPU bit (which is in a hardware register) is set to 1. From what I understand, that bit should be set to 1 only when JTAG reset the pro CPU (as the name and the message indicates), so I don't know why deep sleep would set that... I'd look at general things that can make the ESP32S3 act wonky: is your reset circuit OK? Is your power supply adequate? Generally, can you also get this behaviour on other boards of the same design, or even on entirely different boards (e.g. devboards)?

[MicroController]

There is no JTAG connected to the board.

Are you using the USB-JTAG-Serial for monitoring?

[floitsch_toit]

Are you using the USB-JTAG-Serial for monitoring?

No. We use the serial port's TX (pin 43 iirc).

[Pjordan]

We have de-soldered the ESP32-S3 from our PCB and hooked it up directly on a USB port with power from a lab power supply and we still see the Pro CPU reset by JTAG. Nothing is connected to the chip at all. We see this error on all our ESP32-S3 chips in production. Any advice on how this is happening or what we can do about it?

[ESP_Sprite]

Can you also get this behaviour on an Espressif devboard? If so, we can replicate it and look into it. If not, it points at a hardware issue.

[Pjordan]

We have just reproduced the same error on a dev board (DollaTek ESP32-S3-DevKitC-1) Any suggestions?

[ESP_Sprite]

Okay, that makes it a software thing. Any chance you can share the code with us? Preferably whittled down to the bare minimum that shows the issue, and preferably the code is shared publicly, but if that's not possible you can send whatever you privately to me as well. (Although do note lunar new year is coming up, so no guarantees when we can look at this, unfortunately.)