SD Card reading --> ESP32 freezes (connection to server/WLAN gets lost)

[basementmedia2]

Board

ESP32 (Wemos D1 mini)

Device Description

Custom ESP32, got no photo here sorry (am in the office right now)

Hardware Configuration

SD-Card Shield connected

Version

v2.0.3

IDE Name

Arduino IDE

Operating System

Linux Manjaro

Flash frequency

40

PSRAM enabled

no

Upload speed

115200

Description

I have a sketch which works as follows:

On client-side (webserver) the user selects a date which is then sent (per Ajax) to the ESP.
if the date is e.g. "2022-06-08" a file named "2022-06-08.bin" is opened for reading and then read line by line.

With esp core 1.0.6 it works perfectly.
Same code with esp core 2.0.2 was very slow (cause there was still a bug i think)
Now with esp core 2.0.3 it is really fast again, but after a few date selects (means a few SD-card-Read-cycles) the esp totally freezes (webserver is not available any more).

Sketch

// Just Parts of the Sketch
// I cannot publish complete Sketch here because of copyrights (customer project)

struct data {
  uint16_t value1;
  uint16_t value2;
  uint16_t value3;
  // 10 more values, just an example 
};

void tagesauswertung (char* datum_select, int startpos, int endpos, int tag, int monat, int jahr, int rand_key_sent) {
  struct data Messung;
  DynamicJsonDocument auswertung_arr(4800);

  sprintf(filename_auswertung, "/e/%s.bin", datum_select);
  auswertungenFile = SD.open(filename_auswertung, FILE_READ);
  if (auswertungenFile) {
    
    int i = 0;
    int alive = 0;
    while (auswertungenFile.available()) {

      auswertungenFile.read((uint8_t *)&Messung, sizeof(Messung));
      
      if (i >= startpos && i <= endpos) {
        // As the file is very big, the JSON-Array would get to o big
        // So the server sends "startpos" and "endpos" and thus the content is transferred partly to the client
        alive = 1;
        auswertung_arr[i - startpos]["val1"] = Messung.value1;
        auswertung_arr[i - startpos]["val2"] = Messung.value2;
        auswertung_arr[i - startpos]["val3"] = Messung.value3;
        // ten more values, just an example
      }
      i++;
    }

    if (alive == 1) {
      // Still Data left --> Send it to the client
      String jsonString;
      serializeJson(auswertung_arr, jsonString);
      Serial.println(jsonString);
      server.send(200, "application/json", jsonString);
    } else {
        server.send(200, "text/plain", "Done");
    }

  } else {
    server.send(200, "text/plain", "Fehler");
  } 
}

Debug Message

ESP Freezes (Webserver is not available anymore)

Other Steps to Reproduce

No response

I have checked existing issues, online documentation and the Troubleshooting Guide

• I confirm I have checked existing issues, online documentation and Troubleshooting guide.

[VojtechBartoska]

Can you please take a look @P-R-O-C-H-Y?

[basementmedia2]

Hi @VojtechBartoska

what do you mean? Does @P-R-O-C-H-Y have a solution for this problem?
Or should i post my issue in his repo?

Best wishes
Daniel

[P-R-O-C-H-Y]

Hi @basementmedia2

I take care of issues about SD lib :) Thats why @VojtechBartoska assigned me and pinged me in comment.
I will take a look on you issue soon, no need to repost issue anywhere ;)

[P-R-O-C-H-Y]

Hi @basementmedia2,
Can you try it with latest 2.0.4 release? Thanks

[basementmedia2]

Hi @P-R-O-C-H-Y ,

sure, I will check it these days and come back with my experience.
Thank you for your help!

Best wishes
Daniel

[basementmedia2]

Hi @P-R-O-C-H-Y ,

unfortunately updating to 2.0.4 does not fix the problem.

I have created a test sketch, which I think you can use to reproduce the problem yourself.

#include <WiFi.h>
#include <WiFiAP.h>
#include <HTTPClient.h>
#include <WebServer.h>
WebServer server(80);
#include <WiFiClient.h>
#include <WiFiClientSecure.h>

#include <FS.h>
#include <SD.h>
#include <ArduinoJson.h>

char firmware_update_url[150]="some url to json file";
char IPlocal[20];

unsigned long connTimer=0;
unsigned long dauer = 10000;
const char ssid[14] = "MySSID";
const char password[21] = "MyPW";
const char apssid[10] = "MyESP32";
const char appassword[10] = "123456789";

unsigned long previousMillis_reconn = 0;
unsigned long interval_reconn = 30000;

unsigned long previousMillis_fwc = 0;
unsigned long interval_fwc = 20000;

unsigned long previousMillis_sd_loop = 0;
unsigned long interval_sd_loop = 20000;

int counter=0;

static const char indexPage[] PROGMEM = R"=====(
<!doctype html>
<html lang="de">
<body>
Hello World
</body>
</html>
)=====";

struct data {
    uint16_t tag;
    uint16_t stunde;
    float liter;
};

char filename_auswertung[35];
char* datum;
File auswertungenFile;


void display_root() {
    const char * httpType PROGMEM = "text/html";
    server.send_P(200, httpType, indexPage);
}


void setup() {
    delay(2000);
    Serial.begin(115200);
    if(!SD.begin()){
        Serial.println("Card Mount Failed");
    }


    xTaskCreatePinnedToCore(
        keepWiFiAlive,
        "Keep Wifi Alive",
        5000,
        NULL,
        2,
        NULL,
        CONFIG_ARDUINO_RUNNING_CORE
    );


    WiFi.mode(WIFI_AP_STA);
    WiFi.begin(ssid, password);

    connTimer = millis();

    while (WiFi.status() != WL_CONNECTED) {
        Serial.print(".");
        delay(500);
    }
    if (WiFi.status() == WL_CONNECTED) {
        Serial.println("We are connected...");
        strcpy(IPlocal, WiFi.localIP().toString().c_str());
    } else {
        Serial.println("We are not connected");
    }

    WiFi.softAP(apssid, appassword);

    server.on("/", display_root);
    server.on("/auswertung", auswertung);

    server.begin();
}

void loop() {
    server.handleClient();
    unsigned long currentMillis_reconn = millis();
    if ((WiFi.status() != WL_CONNECTED) && (currentMillis_reconn - previousMillis_reconn >=interval_reconn)) {
        Serial.println("Reconnecting to WiFi...");
        WiFi.disconnect();
        WiFi.begin(ssid, password);
        connTimer = millis();
        while (WiFi.status() != WL_CONNECTED && (millis() - connTimer < dauer)) {
            Serial.print(".");
            delay(500);
        }
        previousMillis_reconn = currentMillis_reconn;
    } else {
        if (millis() - previousMillis_fwc > interval_fwc) {
            Serial.println("Firmware Check!");
            check_firmware_version();
            previousMillis_fwc = millis();
        }
    }

    if (millis() - previousMillis_sd_loop > interval_sd_loop) {
        if(SD.begin()){
            auswertung();
            previousMillis_sd_loop = millis();
        }
    }
}


void keepWiFiAlive(void * parameters) {
    for (;;) {
        if (WiFi.status() == WL_CONNECTED) {
            Serial.println("We are connected");
            vTaskDelay(10000 / portTICK_PERIOD_MS);
            check_firmware_version();
            continue;
        }
        WiFi.disconnect();
        WiFi.begin(ssid, password);
        unsigned long startAttemptTime=millis();
        while (WiFi.status() != WL_CONNECTED && millis() - startAttemptTime < 5000) {

        }
        if (WiFi.status() != WL_CONNECTED) {
            Serial.println("Not connected! ");
            continue;
        } else {
            // Again connected
            continue;
        }
    }
}


void auswertung() {
    Serial.println(WiFi.localIP());
    Serial.print("Starting SD-Reading ->");
    datum="2022-05-26";
    for (int i = 0; i < 10; i = i + 1) {
        tagesauswertung(datum, i*16, (i*16)+16, 26, 5, 2022);
    }
    datum="2022-05-27";
    for (int i = 0; i < 10; i = i + 1) {
        tagesauswertung(datum, i*16, (i*16)+16, 27, 5, 2022);
    }
    counter++;
    Serial.println(counter);
}

void tagesauswertung (char* datum_select, int startpos, int endpos, int tag, int monat, int jahr) {
    struct data Messung;
    DynamicJsonDocument auswertung_arr(4800);
    sprintf(filename_auswertung, "/e/%s.bin", datum_select);
    auswertungenFile = SD.open(filename_auswertung, FILE_READ);
    if (auswertungenFile) {
        Serial.print("Opening file ");
        Serial.println(filename_auswertung);
        int i = 0;
        int alive = 0;
        while (auswertungenFile.available()) {
            auswertungenFile.read((uint8_t *)&Messung, sizeof(Messung));
            if (i >= startpos && i <= endpos) {
                alive = 1;
                auswertung_arr[i - startpos]["t"] = Messung.tag;
                auswertung_arr[i - startpos]["s"] = Messung.stunde;
                auswertung_arr[i - startpos]["l"] = Messung.liter;
            }
            i++;
        }

        if (alive == 1) {
            String jsonString;
            serializeJson(auswertung_arr, jsonString);
            Serial.println(jsonString);
            server.send(200, "application/json", jsonString);
        } else {
            // We are finished with reading
            server.send(200, "text/plain", "Fertig");
        }
    } else {
        Serial.println("File could not be opened");
    }
}

void check_firmware_version() {
    if (WiFi.status() == WL_CONNECTED) {
        WiFiClientSecure *client = new WiFiClientSecure;
        if(client) {
            client->setInsecure(); {
                HTTPClient https;
                https.useHTTP10(true);
                if (https.begin(*client, firmware_update_url)) { //Abfrage-URL
                    int httpsCode = https.GET(); //Antwort des Servers abrufen
                    if (httpsCode == 200) {
                        Serial.println("Firmware-File found!");
                    }
                    https.end();
                }
            }
            delete client;
        }
    } else {
        Serial.println("Not connected --> No FW-Check!");
    }
}

Here are the bin-files, used in the function "auswertung"

bin-files.zip

The JSON-File you must pölace somewhere on your server (but i don't know if this part is really relevant).

If you run the sketch with 1.0.6, it runs endlessly correctly through - the function "day evaluation" is started again and again. With 2.0.3 and with 2.0.4 after a few runs (sometimes more, sometimes less) the loop-task is shot down. Then only the separate task "keepWiFiAlive" is running.

Could it be that the heap memory is running full and therefore at some point the void loop can no longer be executed?

I also noticed that any execution of SD.begin() when no SD card is inserted causes the heap to grow until then eventually the memory overflows. No idea if this is intentional or if it's a bug.

You can test this with the following sketch and NOT putting an SD-Card in the SD-Card-Reader:

#include <SD.h> // SD-Kartenleser

unsigned long millisMerker= 0;
unsigned long dauerBisRepeat = 20000;

int sd_card_status =0;

void setup() {
  Serial.begin(115200);
  if (!SD.begin(5)) {
    sd_card_status=0;
  } else {
    sd_card_status=1;
  }
}

void loop() {
  if (!SD.begin(5)) {
    sd_card_status=0;
  } else {
    sd_card_status=1;
  }
  if (millis() - millisMerker > dauerBisRepeat) {
    uint32_t heapSize=ESP.getHeapSize(); //total heap size
    uint32_t freeHeap=ESP.getFreeHeap(); //available heap
    uint32_t minFreeHeap=ESP.getMinFreeHeap(); //lowest level of free heap since boot
    uint32_t maxAllocHeap=ESP.getMaxAllocHeap(); //largest block of heap that can be allocated at once
    Serial.println(" -> Total: " + (String)heapSize + ", Frei: " + (String)freeHeap + " Min: "+ (String)minFreeHeap + ", Max: " + (String)maxAllocHeap) + " SD-Card:" + sd_card_status;
    millisMerker = millis(); 
  }
}

Would be great to get further support.

Best wishes
Daniel

[P-R-O-C-H-Y]

Hi @basementmedia2 Thanks for really nice report 👍

Gonna take a look on the issue, and for sure the heap cannot grow. There has to be some bug.
I will inform you about the progress :)