Uart smashing Wifi Stack

[jurstu]

Hi!

My application goes like this:
-WifiAP with accepting sockets (tcp-socket-server)
-UART1 for other comunication than programming/debugging

when I leave Wifi on in AP mode, with or without accepting sockets -> too many messages from UART1 and stack gets smashed or the core panics

If I leave only UART1, it works perfectly fine!

I can provide anything from code to config.

example core panic dump

Code: Select all

D (7275) btCmdHandler: LC cmd, entering handler
I (7285) whichTarget LC CMD: 16
Guru Meditation Error: Core  0 panic'ed (IllegalInstruction)
. Exception was unhandled.
Core 0 register dump:
PC      : 0x400f0b58  PS      : 0x00060330  A0      : 0x80104440  A1      : 0x3ffc3570  
0x400f0b58: ieee80211_hostap_send_beacon_process at ??:?

A2      : 0x3ffb8818  A3      : 0x00000000  A4      : 0x00000001  A5      : 0x00000000  
A6      : 0x3ffb4bf8  A7      : 0x3ffb4bfc  A8      : 0x800f0b18  A9      : 0x3ffc3550  
A10     : 0x00000000  A11     : 0x00000001  A12     : 0x00000000  A13     : 0x3ffc3648  
A14     : 0x3ffb5238  A15     : 0x3ffc8610  SAR     : 0x0000001f  EXCCAUSE: 0x00000000  
EXCVADDR: 0x00000000  LBEG    : 0x4000c2e0  LEND    : 0x4000c2f6  LCOUNT  : 0xffffffff  

Backtrace: 0x400f0b58:0x3ffc3570 0x4010443d:0x3ffc35f0 0x40104726:0x3ffc3610 0x40087284:0x3ffc3630
0x400f0b58: ieee80211_hostap_send_beacon_process at ??:?

0x4010443d: wifi_station_stop at ??:?

0x40104726: ieee80211_timer_do_process at ??:?

0x40087284: ppTask at ??:?


Rebooting...
I (10) boot: ESP-IDF v3.1-dev-1381-g9a55b42-dirty 2nd stage bootloader
I (10) boot: compile time 17:39:10
I (10) boot: Enabling RNG early entropy source...
I (16) boot: SPI Speed      : 40MHz
I (20) boot: SPI Mode       : DIO
I (24) boot: SPI Flash Size : 4MB
I (28) boot: Partition Table:
I (32) boot: ## Label            Usage          Type ST Offset   Length
I (39) boot:  0 nvs              WiFi data        01 02 00009000 00006000
I (46) boot:  1 phy_init         RF data          01 01 0000f000 00001000
I (54) boot:  2 factory          factory app      00 00 00010000 00200000
I (61) boot: End of partition table
I (65) esp_image: segment 0: paddr=0x00010020 vaddr=0x3f400020 size=0x165f0 ( 91632) map
I (106) esp_image: segment 1: paddr=0x00026618 vaddr=0x3ffb0000 size=0x03a20 ( 14880) load
I (113) esp_image: segment 2: paddr=0x0002a040 vaddr=0x3ffb3a20 size=0x00000 (     0) load
I (113) esp_image: segment 3: paddr=0x0002a048 vaddr=0x40080000 size=0x00400 (  1024) load
0x40080000: _iram_start at /home/jur/Desktop/esp/esp-idf/components/freertos/xtensa_vectors.S:1685

I (123) esp_image: segment 4: paddr=0x0002a450 vaddr=0x40080400 size=0x05bc0 ( 23488) load
I (141) esp_image: segment 5: paddr=0x00030018 vaddr=0x400d0018 size=0x69828 (432168) map
0x400d0018: _stext at ??:?

I (292) esp_image: segment 6: paddr=0x00099848 vaddr=0x40085fc0 size=0x0b198 ( 45464) load
0x40085fc0: ppSearchTxframe at ??:?

I (312) esp_image: segment 7: paddr=0x000a49e8 vaddr=0x400c0000 size=0x00000 (     0) load
I (312) esp_image: segment 8: paddr=0x000a49f0 vaddr=0x50000000 size=0x00000 (     0) load
E (318) esp_image: Checksum failed. Calculated 0xa0 read 0x26
E (325) boot: Factory app partition is not bootable
E (330) boot: No bootable app partitions in the partition table
user code done

[ESP_Angus]

Hi jurstu,

When you say the stack is being smashed, what's the basis for this? The example backtrace looks to have a valid call stack (in the WiFi ppTask).

Do you get any warnings or errors from the Task or Interrupt WDT in your code?

The oddest thing about the log output is that the device resets after crashing and then fails to boot due to invalid flash contents. Does this only happen immediately after crashing (ie the ESP32 boots successfully if reset again?)

What hardware are you using, is it possible there are power stability issues? (If you're using a dev board, try swapping the USB port or cable.)

Also, what ESP-IDF version are you using?

EDIT: I see you're using v3.1-dev from June. Can you try updating to latest master branch, please?

Thanks,

Angus