ESP32-C3: failing to connect GDB (OCD-580)

[tcbennun]

Development Kit

N/A

Module or chip used

ESP32-C3-MINI-1

Debug Adapter

Built-in USB

OpenOCD version

v0.11.0-esp32-20211220

Operating System

Debian Linux (kernel 5.18)

Using an IDE ?

N/A

OpenOCD command line

openocd -f board/esp32c3-builtin.cfg

JTAG Clock Speed

Any (tried default, tried 1 MHz)...

ESP-IDF version

v4.4.1

Problem Description

1. Connect module to PC via USB cable.
2. Run openocd -l openocd.log -d3 -f board/esp32c3-builtin.cfg in a new terminal with ESP-IDF already exported.
3. Create file named gdbinit with the following contents:

target remote :3333
set remote hardware-watchpoint-limit 2
mon reset halt
flushregs
thb app_main
c

4. Run riscv32-esp-elf-gdb -ex "set remotelogfile gdb.log" -x gdbinit build/blecent.elf in another new terminal with ESP-IDF already exported.

The result is a failure for GDB to connect to the device. OpenOCD initially connects correctly, but problems occur when GDB is invoked. See logs.

Key line in the OpenOCD log:
Error: 1043 69168 riscv.c:1898 riscv_wait_algorithm(): Algorithm timed out after 40001 ms.

Debug Logs

OpenOCD log: https://gist.github.com/tcbennun/e1625a2c6fd85873721a51334ed65098

GDB log:

w +$qSupported:multiprocess+;swbreak+;hwbreak+;qRelocInsn+;fork-events+;vfork-events+;exec-events+;vContSupported+;QThreadEvents+;no-resumed+#df
r +<Timeout: 2 seconds>
w -
r <Timeout: 2 seconds>
w -
r <Timeout: 2 seconds>
w -+$vMustReplyEmpty#3a
r <Timeout: 2 seconds>
w $vMustReplyEmpty#3a
r <Timeout: 2 seconds>
w $vMustReplyEmpty#3a
r <Timeout: 2 seconds>
w $vMustReplyEmpty#3a
r <Timeout: 2 seconds><Timeout: 2 seconds>
w -
r <Timeout: 2 seconds>
w -
r <Timeout: 2 seconds>
w -+
End of log

Expected behavior

Expected GDB to connect to OpenOCD and for the program to stop at a temporary breakpoint at the start of app_main.

Screenshots

No response

[erhankur]

Do you see the same issue with hello_world example?

[tcbennun]

I get the same in the hello_world example, @erhankur.

However, as it turns out, this is apparently fixed by forcing the chip into Download Boot by pulling GPIO9 low on powerup. (Not yet sure whether this is a permanent fix, but I've managed to hit the entry point successfully.)

Is this usually a requirement?

[tcbennun]

I'm now finding that GDB freezes after hitting that initial breakpoint - I can't enter any commands at all. If I interrupt with Ctrl+C, I can enter commands but nothing appears to actually happen...? OpenOCD reports nothing out of sorts?

[tcbennun]

... Yeah, so going into Download Boot means that the program doesn't execute at all... the builtin JTAG doesn't seem to notice and just freezes as if execution is continuing.

[gerekon]

@tcbennun

openocd -l openocd.log -d3 -f board/esp32c3-builtin.cfg

Could you post collected openocd.log?

[tcbennun]

@gerekon I did, I posted it as a gist link just above the GDB log, since it's so long.

[gerekon]

Sorry. Can not find attached file with openocd log. I see only GDB log posted as a part of the initial message.

Does your program uses WIFI/BLE?
In IDF 4.4.1 USB JTAG operation can be corrupted due USB PHY is disabled (espressif/esp-idf#8046, espressif/arduino-esp32#6264).
To overcome this you can

• use add a fix described here
• or config builtin USB/CDC to be used as a console (see ESP_PHY_ENABLE_USB)

[tcbennun]

@gerekon

The gist link is there... it's in the verbose-styled part of the issue just under the "Debug Logs" header. Here's the link again: https://gist.github.com/tcbennun/e1625a2c6fd85873721a51334ed65098

Does your program uses WIFI/BLE?

It does use BLE. I just retried hello_world with success this time - so I assume it's PHY related after all.

I already had ESP_PHY_ENABLE_USB enabled, and it looks like esp_phy_enable() calls phy_bbpll_en_usb(true) already in ESP-IDF 4.4.1.

(edit: which I believe covers the fixes you mentioned?)

[erhankur]

@tcbennun If you still have an issue with the latest OpenOCD feel free to re-open.