base64 encode resets esp32 on certain strings

[euquiq]

Hi there!

I think I found a bug: Sometimes my DOIT ESP32 DEVKIT V1 starts resetting itself when encoding a string by using the base64 library. .

When trying to encode certain quantity of characters, my ESP32 will give the following error:

CORRUPT HEAP: Bad tail at 0x3ffc9aac. Expected 0xbaad5678 got 0xbaad5600
 assertion "head != NULL" failed: file "/Users/ficeto/Desktop/ESP32/ESP32/esp-idf-public/components/heap/./multi_heap_poisoning.c", line 199, function: multi_heap_free
 abort() was called at PC 0x400d2af3 on core 1
 
 Backtrace: 0x400870ac:0x3ffcee50 0x400871ab:0x3ffcee70 0x400d2af3:0x3ffcee90 0x40086dd9:0x3ffceec0 0x40082796:0x3ffceee0 0x40082d19:0x3ffcef00 0x4000bec7:0x3ffcef20 0x400d0905:0x3ffcef40 0x400d0931:0x3ffcef80 0x400d0880:0x3ffcefa0 0x400dc84f:0x3ffceff0
 
 Rebooting...

I can reproduce the error with the following code:

#include <base64.h>
base64 b;

void setup()
{
	Serial.begin(115200);
	Serial.println("12345678 = " + b.encode("12345678"));
	delay(1000);
	Serial.println("123456 = " + b.encode("123456"));
	delay(1000);
	Serial.println("1234567 = " + b.encode("1234567")); //THIS LINE WILL CRASH ESP32
}

void loop()
{

}

I also tried this simple code in two other ESP32 boards, the geekworm EASY KIT ESP32-01 (revision 0 silicon) and another one with revision 1 silicon, which says ESP32_Core_board-V2.

Same behavior in the three of them.

Regards,

Enrique

[me-no-dev]

how about Serial.println(String("123456 = ") + b.encode("123456"));

[euquiq]

Hi me-no-dev!

It is not that line of code which you suggest to change that breaks the Arduino ESP32: it is the following one. Anyways I added the String() to the one you say, and ALSO the one that really crashes the ESP32, with the same behavior as before.

The line:

Serial.println("1234567 = " + b.encode("1234567")); //THIS LINE WILL CRASH ESP32

is a no-go. And the use of the String function helps not:

Opening port
Port open

12345678 = MTIzNDU2Nzg=
123456 = MTIzNDU2
CORRUPT HEAP: Bad tail at 0x3ffcc894. Expected 0xbaad5678 got 0xbaad5600
assertion "head != NULL" failed: file "/Users/ficeto/Desktop/ESP32/ESP32/esp-idf-public/components/heap/./multi_heap_poisoning.c", line 199, function: multi_heap_free
abort() was called at PC 0x400d2bc7 on core 1

Backtrace: 0x40087120:0x3ffcecb0 0x4008721f:0x3ffcecd0 0x400d2bc7:0x3ffcecf0 0x40086e4d:0x3ffced20 0x40082842:0x3ffced40 0x40082dc5:0x3ffced60 0x4000bec7:0x3ffced80 0x400d09a1:0x3ffceda0 0x400d09cd:0x3ffcede0 0x400d08d7:0x3ffcee00 0x400dc923:0x3ffcee50

Rebooting...

If I delete the third b.encode with that specific 7 letter string to encode, the code works as expected.

The encode function crashes the ESP32 chip when asked to encode a string with 7 chars.

Weird bug!

Regards,

Enrique.

[me-no-dev]

ahhh OK will have a look :)

[anteph]

Hi!

Same problem here with the code below.

#include <base64.h>

void setup() {
  Serial.begin(115200);
  delay(1000);
  Serial.println(base64::encode("test"));
}

void loop() {
}

Then the ESP reboots a couple of times and then stops rebooting. The message I get is the following:

rst:0xc (SW_CPU_RESET),boot:0x13 (SPI_FAST_FLASH_BOOT)
configsip: 0, SPIWP:0xee
clk_drv:0x00,q_drv:0x00,d_drv:0x00,cs0_drv:0x00,hd_drv:0x00,wp_drv:0x00
mode:DIO, clock div:1
load:0x3fff0018,len:4
load:0x3fff001c,len:812
load:0x40078000,len:0
load:0x40078000,len:10212
entry 0x40078a00
CORRUPT HEAP: Bad tail at 0x3ffcc8d7. Expected 0xbaad5678 got 0xbaad003d
assertion "head != NULL" failed: file "/Users/ficeto/Desktop/ESP32/ESP32/esp-idf-public/components/heap/./multi_heap_poisoning.c", line 199, function: multi_heap_free
abort() was called at PC 0x400d2a97 on core 1

Backtrace: 0x40087120:0x3ffcec20 0x4008721f:0x3ffcec40 0x400d2a97:0x3ffcec60 0x40086e4d:0x3ffcec90 0x40082842:0x3ffcecb0 0x40082dc5:0x3ffcecd0 0x4000bec7:0x3ffcecf0 0x400d0c69:0x3ffced10 0x400d0c95:0x3ffced50 0x400d0838:0x3ffced70 0x400dc7f3:0x3ffcedb0

Rebooting...
ets Jun  8 2016 00:22:57

Best regards,
Nuno Santos

[me-no-dev]

can you please decode the backtraces and post them here?

[anteph]

Sorry I've never done that, is there any guide for how to decode the backtraces?

Thank you

[me-no-dev]

yup: https://github.com/me-no-dev/EspExceptionDecoder

[euquiq]

@me-no-dev Here I paste my backtrace:

Decoding 12 results
0x400d2b7f: __assert_func at /Users/ivan/e/newlib_xtensa-2.2.0-bin/newlib_xtensa-2.2.0/xtensa-esp32-elf/newlib/libc/stdlib/../../../.././newlib/libc/stdlib/assert.c line 63 (discriminator 8)
0x40087120: invoke_abort at /Users/ficeto/Desktop/ESP32/ESP32/esp-idf-public/components/esp32/./panic.c line 553
0x4008721f: abort at /Users/ficeto/Desktop/ESP32/ESP32/esp-idf-public/components/esp32/./panic.c line 553
0x400d2b7f: __assert_func at /Users/ivan/e/newlib_xtensa-2.2.0-bin/newlib_xtensa-2.2.0/xtensa-esp32-elf/newlib/libc/stdlib/../../../.././newlib/libc/stdlib/assert.c line 63 (discriminator 8)
0x40086e4d: multi_heap_free at /Users/ficeto/Desktop/ESP32/ESP32/esp-idf-public/components/heap/./multi_heap_poisoning.c line 199 (discriminator 1)
0x40082842: heap_caps_free at /Users/ficeto/Desktop/ESP32/ESP32/esp-idf-public/components/heap/./heap_caps.c line 275
0x40082dc5: _free_r at /Users/ficeto/Desktop/ESP32/ESP32/esp-idf-public/components/newlib/./syscalls.c line 42
0x400d0d51: base64::encode(unsigned char*, unsigned int) at C:\Users\Enrique\Documents\Arduino\hardware\espressif\esp32\cores\esp32/base64.cpp line 50
0x400d0d7d: base64::encode(String) at C:\Users\Enrique\Documents\Arduino\hardware\espressif\esp32\cores\esp32/base64.cpp line 63
0x400d08cc: setup() at C:\Users\Enrique\Documents\Arduino\probando_base64/probando_base64.ino line 11
0x400dc8db: loopTask(void*) at C:\Users\Enrique\Documents\Arduino\hardware\espressif\esp32\cores\esp32/main.cpp line 15

It crashes, as before, when reaching the line with the "1234567" string:

#include <base64.h>
base64 b;

void setup()
{
  Serial.begin(115200);
  Serial.println("12345678 = " + b.encode("12345678"));
  delay(1000);
  Serial.println("123456 = " + b.encode("123456"));
  delay(1000);
  Serial.println("1234567 = " + b.encode("1234567")); //THIS LINE WILL CRASH ESP32
}

void loop()
{

}

Hopefully it helps.

Regards,

Enrique

[anteph]

Sorry for the delay in posting the backtrace but I'm facing some issues running the decoder tool. I will post it as soon as I can solve them.

Best regards,
Nuno Santos

[euquiq]

@me-no-dev It would seem that the base64 code in espressif IDF, which in turn arduino calls for encoding in base64 is flawed ? Maybe it has a "runaway heap" due to an uncontrolled memory / array / String space usage or something like that (I infer this from reading the:

0x40082842: heap_caps_free at /Users/ficeto/Desktop/ESP32/ESP32/esp-idf-public/components/heap/./heap_caps.c line 275

... on my backtrace).

But then, I might be completely wrong. I posted this because I would love this bug not to be stalled into oblivion. If there is something I can do, please, tell me. Maybe I can post this problem as an espressif / IDF issue ?

[me-no-dev]

no worries :) I have been completely overwhelmed with other work, so I had not time to really sit down and dig into this (i imagine it will require some time). Source for base64 is in Arduino itself: https://github.com/espressif/arduino-esp32/tree/master/cores/esp32/libb64

[anteph]

Finally had some time to update my old Arduino IDE version and decode the backtrace :)

0x400d1e47: __assert_func at /Users/ivan/e/newlib_xtensa-2.2.0-bin/newlib_xtensa-2.2.0/xtensa-esp32-elf/newlib/libc/stdlib/../../../.././newlib/libc/stdlib/assert.c line 63 (discriminator 8)
0x4008720c: invoke_abort at /Users/ficeto/Desktop/ESP32/ESP32/esp-idf-public/components/esp32/./panic.c line 572
0x4008730b: abort at /Users/ficeto/Desktop/ESP32/ESP32/esp-idf-public/components/esp32/./panic.c line 572
0x400d1e47: __assert_func at /Users/ivan/e/newlib_xtensa-2.2.0-bin/newlib_xtensa-2.2.0/xtensa-esp32-elf/newlib/libc/stdlib/../../../.././newlib/libc/stdlib/assert.c line 63 (discriminator 8)
0x40086f39: multi_heap_free at /Users/ficeto/Desktop/ESP32/ESP32/esp-idf-public/components/heap/./multi_heap_poisoning.c line 199 (discriminator 1)
0x400829a2: heap_caps_free at /Users/ficeto/Desktop/ESP32/ESP32/esp-idf-public/components/heap/./heap_caps.c line 275
0x40082ef9: _free_r at /Users/ficeto/Desktop/ESP32/ESP32/esp-idf-public/components/newlib/./syscalls.c line 42
0x400d0ca8: base64::encode(unsigned char*, unsigned int) at C:\Users\N\Documents\Arduino\hardware\espressif\esp32\cores\esp32/base64.cpp line 50
0x400d0cd5: base64::encode(String) at C:\Users\N\Documents\Arduino\hardware\espressif\esp32\cores\esp32/base64.cpp line 63
0x400d0859: setup() at C:\Users\N\Desktop\Blog\Posts\212 - ESP32 Arduino base64 encode\Base64_encode/Base64_encode.ino line 8
0x400de22f: loopTask(void*) at C:\Users\N\Documents\Arduino\hardware\espressif\esp32\cores\esp32/main.cpp line 15

Hope it helps with finding the bug.

Meanwhile I've written a small tutorial about using base64 encoding, will leave it here in case someone needs help getting started:
https://techtutorialsx.com/2017/12/09/esp32-arduino-base64-enconding/

Best regards,
Nuno Santos

[sagreen83]

Any updates on this topic?

I'm getting the same error when encoding the string "system1"

Thanks,
Scott...

[euquiq]

No update yet. Which compels me to write the following: Since I found no way of private messaging any of the moderators on espressif/arduino-esp32, maybe I can take the space to ask a couple of questions (and I hope I do not offend anyone):

1. Is espressif in charge of the arduino-esp32 branch development (meaning: are the moderators in here espressif employees or at least paid by them somehow)?

2. Since espressif got now their own IDE (IDE-IDF) ... is this ESP32-Arduino branch being deprecated or at least left to the resources of "free will - time" of enthusiasts like the moderators ? (in case answer to earlier question 1) is that they are guys "like us" and not being compensated in any way ?)

I am trying to understand the Arduino-ESP32 "scene" and decide if it is a pragmatic sound decision to move away from good old Arduino framework, shifting my efforts into espressif IDF framework (which would be more than cumbersome, given the apparent (for me) lack of libraries and third party sensors / peripherals support).

Most probably I got it all wrong, again I am sorry for any impoliteness I may have incurred, as it is not my intention other than grasp the way this espressif/Arduino-ESP32 branch works.

Regards,
Enrique

[me-no-dev]

If you consider the fact that there are a whole bunch of things that i am working on (non-arduino) you will understand. On a more positive note i can say that i am almost done with all other priority tasks and will take care of this soon. Then also this is the same lib not written by us and in use by esp8266 arduino also. You can write whatever you want but that would only bring more negativity if anything.