Mbed-os: crash_log_parser crashes

[bmcdonnell-ionx]

I'm troubleshooting an intermittent "imprecise" HardFault. crash_log_parser crashed when trying to parse a crash log from firmware where I had disabled the write buffer use during default memory map access with the instruction SCnSCB->ACTLR &= ~(SCnSCB_ACTLR_DISDEFWBUF_Msk);.

References:
https://community.nxp.com/docs/DOC-103810
http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.dui0553a/CHDCBHEE.html

EDIT: Update here: https://github.com/ARMmbed/mbed-os/issues/6444#issuecomment-376232035.

Description

• Type: Bug
• Priority: Major | Minor

---

Bug

Target
LPC4088

Toolchain:
GCC_ARM

Toolchain version:
MCUXpresso IDE v10.1.1 [Build 606] [2018-01-02]
(and whatever version of GCC_ARM that comes with)

mbed-cli version:
1.5.0 (with https://github.com/ARMmbed/mbed-cli/pull/642)

mbed-os sha:
ad284b280 (tag: mbed-os-5.8.0-rc2, tag: mbed-os-5.8.0, arm/mbed-os-5.8, mbed-os-5.8) Merge pull request #6352 from ARMmbed/release-candidate

DAPLink version:

Expected behavior

crash_log_parser successfully parses a crash log to produce crash info.

Actual behavior

me@pc ~/Documents/proj/troubleshooting
$ ./crash_log_parser.py crash-log.txt fw-build-dbg.axf fw-build-dbg.map

Crash Info:
Traceback (most recent call last):
  File "./crash_log_parser.py", line 194, in <module>
    main(args.crashlog, elfhelper)
  File "./crash_log_parser.py", line 164, in main
    print("\tCrash location = %s [0x%s] (based on PC value)" % (pc_name.strip(), str(pc_val)))
UnboundLocalError: local variable 'pc_name' referenced before assignment

Steps to reproduce (maybe)

• Write a program that crashes with a HardFault on an ARM Cortex-M4. (Might need to be "imprecise"; IDK. Cortex-M3 would probably work.)
• Include the instruction SCnSCB->ACTLR &= ~(SCnSCB_ACTLR_DISDEFWBUF_Msk); to force precise faults.
• Export to MCUXpresso IDE. Build. Load and run firmware.
• Make/watch it crash. Capture the MbedOS Fault Handler output, and run the crash_log_parser on it.
• Observe failure to parse as above. Maybe.

[0xc0170]

@SenRamakri Please review

cc @ARMmbed/team-nxp

[mmahadevan108]

@bmcdonnell-ionx Thank you for finding this. Is this a fix to the DAPLink HICHAL driver?

Could you submit a patch.

[bmcdonnell-ionx]

I did have some crashes later which it could parse, so I'm not sure if you'll be able to reproduce this. Maybe you can give it a shot, and if not, then we wait and see.

(Sorry I can't share the files with you to troubleshoot. It's application firmware.)

@mmahadevan108

Could you submit a patch.

I would have if I knew how to fix the issue.

Is this a fix to the DAPLink HICHAL driver?

I don't know where the problem lies.

[bmcdonnell-ionx]

I'm using an Embedded Artists' LPC4088 QuickStart Board.

[SenRamakri]

As far as I see this is an issue in the tool itself, nothing to do with DAPLink driver. I'll look into this.

[SenRamakri]

@bmcdonnell-ionx - Do you have the crash dump which caused the tool to fail?

[bmcdonnell-ionx]

@SenRamakri

Do you have the crash dump which caused the tool to fail?

I'm glad you asked...

On further inspection, I found that the crash dump file I was using used CR (\r) for linebreaks, whereas another that was fine used CRLF (\r\n). I'm not sure how that happened; maybe I captured one from a PuTTY terminal window, and the other from a log file, or something.

Can you make the parser more tolerant?

crash-log-01--CR.txt
crash-log-01--CRLF.txt

[SenRamakri]

@bmcdonnell-ionx - Thanks for sharing the info and feedback. I will look into making the tools more tolerant. Eventually, we may integrate this script/feature into other mbed debug tools.

[ciarmcom]

ARM Internal Ref: MBOTRIAGE-208

[SenRamakri]

Fix in process - https://github.com/ARMmbed/mbed-os/pull/7356

[SenRamakri]

Fix has been merged.