Core Dump

Overview

ESP-IDF provides support to generate core dumps on unrecoverable software errors. This useful technique allows post-mortem analysis of software state at the moment of failure. Upon the crash system enters panic state, prints some information and halts or reboots depending configuration. User can choose to generate core dump in order to analyse the reason of failure on PC later on. Core dump contains snapshots of all tasks in the system at the moment of failure. Snapshots include tasks control blocks (TCB) and stacks. So it is possible to find out what task, at what instruction (line of code) and what callstack of that task lead to the crash. It is also possible dumping variables content on demand if previously attributed accordingly. ESP-IDF provides special script espcoredump.py to help users to retrieve and analyse core dumps. This tool provides two commands for core dumps analysis:

  • info_corefile - prints crashed task’s registers, callstack, list of available tasks in the system, memory regions and contents of memory stored in core dump (TCBs and stacks)

  • dbg_corefile - creates core dump ELF file and runs GDB debug session with this file. User can examine memory, variables and tasks states manually. Note that since not all memory is saved in core dump only values of variables allocated on stack will be meaningful

For more information about core dump internals see the - Core dump internals

Configurations

There are a number of core dump related configuration options which user can choose in project configuration menu (idf.py menuconfig).

Core dump data destination (Components -> Core dump -> Data destination)

  • Save core dump to Flash (Flash)

  • Print core dump to UART (UART)

  • Disable core dump generation (None)

Core dump data format (Components -> Core dump -> Core dump data format)

  • ELF format (Executable and Linkable Format file for core dump)

  • Binary format (Basic binary format for core dump)

The ELF format contains extended features and allow to save more information about broken tasks and crashed software but it requires more space in the flash memory. This format of core dump is recommended for new software designs and is flexible enough to extend saved information for future revisions.

The Binary format is kept for compatibility standpoint, it uses less space in the memory to keep data and provides better performance.

Core dump data integrity check (Components -> Core dump -> Core dump data integrity check)

  • Use CRC32 for core dump integrity verification

Maximum number of tasks snapshots in core dump (Components -> Core dump -> Maximum number of tasks)

Delay before core dump is printed to UART (Components -> Core dump -> Delay before print to UART)

The value is in ms.

Handling of UART core dumps in IDF Monitor (Components -> Core dump -> Delay before print to UART)

The value is base64 encoded.

  • Decode and show summary (info_corefile)

  • Don’t decode

Reserved stack size (Components -> Core dump -> Reserved stack size)

Size of the memory to be reserved for core dump stack. If 0 core dump process will run on the stack of crashed task/ISR, otherwise special stack will be allocated. To ensure that core dump itself will not overflow task/ISR stack set this to the value above 800.

Save core dump to flash

When this option is selected core dumps are saved to special partition on flash. When using default partition table files which are provided with ESP-IDF it automatically allocates necessary space on flash, But if user wants to use its own layout file together with core dump feature it should define separate partition for core dump as it is shown below:

# Name,   Type, SubType, Offset,  Size
# Note: if you have increased the bootloader size, make sure to update the offsets to avoid overlap
nvs,      data, nvs,     0x9000,  0x6000
phy_init, data, phy,     0xf000,  0x1000
factory,  app,  factory, 0x10000, 1M
coredump, data, coredump,,        64K

There are no special requirements for partition name. It can be chosen according to the user application needs, but partition type should be ‘data’ and sub-type should be ‘coredump’. Also when choosing partition size note that core dump data structure introduces constant overhead of 20 bytes and per-task overhead of 12 bytes. This overhead does not include size of TCB and stack for every task. So partition size should be at least 20 + max tasks number x (12 + TCB size + max task stack size) bytes.

The example of generic command to analyze core dump from flash is: espcoredump.py -p </path/to/serial/port> info_corefile </path/to/program/elf/file> or espcoredump.py -p </path/to/serial/port> dbg_corefile </path/to/program/elf/file>

ROM Functions in Backtraces

It is possible situation that at the moment of crash some tasks or/and crashed task itself have one or more ROM functions in their callstacks. Since ROM is not part of the program ELF it will be impossible for GDB to parse such callstacks, because it tries to analyse functions’ prologues to accomplish that. In that case callstack printing will be broken with error message at the first ROM function. To overcome this issue, you can use the ROM ELF provided by Espressif. You can find the esp32c3’s corresponding ROM ELF file from the list of released archives. The ROM ELF file can then be passed to espcoredump.py. More details about ROM ELFs can be found here.

Dumping variables on demand

Sometimes you want to read the last value of a variable to understand the root cause of a crash. Core dump supports retrieving variable data over GDB by attributing special notations declared variables.

Supported notations and RAM regions

  • COREDUMP_DRAM_ATTR places variable into DRAM area which will be included into dump.

  • COREDUMP_RTC_ATTR places variable into RTC area which will be included into dump.

  • COREDUMP_RTC_FAST_ATTR places variable into RTC_FAST area which will be included into dump.

Example

  1. In Project Configuration Menu, enable COREDUMP TO FLASH, then save and exit.

  2. In your project, create a global variable in DRAM area as such as:

// uint8_t global_var;
COREDUMP_DRAM_ATTR uint8_t global_var;

  1. In main application, set the variable to any value and assert(0) to cause a crash.

global_var = 25;
assert(0);

  1. Build, flash and run the application on a target device and wait for the dumping information.

  2. Run the command below to start core dumping in GDB, where PORT is the device USB port:

espcoredump.py -p PORT dbg_corefile <path/to/elf>

  1. In GDB shell, type p global_var to get the variable content:

(gdb) p global_var
$1 = 25 '\031'

Running espcoredump.py

Generic command syntax: espcoredump.py [options] command [args]

Script Options
–chip {auto,esp32,esp32s2,esp32s3,esp32c3}

Target chip type. Default value is “auto”

--port PORT, -p PORT

Serial port device. Either “chip” or “port” need to be specified to determine the port when you have multi-target connected at the same time.

--baud BAUD, -b BAUD

Serial port baud rate used when flashing/reading

--gdb-timeout-sec GDB_TIMEOUT_SEC

Overwrite the default internal delay for gdb responses

Commands

dbg_corefile Starts GDB debugging session with specified corefile

info_corefile Print core dump info from file

Command Arguments
--debug DEBUG, -d DEBUG

Log level (0..3)

--gdb GDB, -g GDB

Path to gdb

--core CORE, -c CORE

Path to core dump file (if skipped core dump will be read from flash)

–core-format {b64,elf,raw}, -t {b64,elf,raw}

File specified with “-c” is an ELF (“elf”), raw (raw) or base64-encoded (b64) binary

--off OFF, -o OFF

Offset of coredump partition in flash (type “make partition_table” to see).

--save-core SAVE_CORE, -s SAVE_CORE

Save core to file. Otherwise temporary core file will be deleted. Does not work with “-c”

--rom-elf ROM_ELF, -r ROM_ELF

Path to ROM ELF file. Will use “<target>_rom.elf” if not specified

--print-mem, -m

Print memory dump. Only valid when info_corefile.

<prog> Path to program ELF file.

Provide feedback about this document