MicroPython RP2040 Machine.WDT and socket.getaddrinfo (WDT not fed in time!)

[OscamSatUser]

Situation : Pico continually reboot.

Replication to create reboot:

o NO INTERNET SCENARIO
o WDT(timeout=8000)
o socket.getaddrinfo()

GetAddrinfo hangs for what appears to be about 15 seconds, well beyond the max WDT timeout allotment. System reboots before the getaddrinfo returns.

I attempted to use asyncio to make the call, but the same scenario occurs, the getaddrinfo gets hung within the function for 14 seconds. Pico reboots before getaddrinfo returns.

I did google and AI searching, and there seems to be no timeout setting for getaddrinfo.

Im using getaddrinfo before setting the time on the PICO using a NTP server. I suppose i could use the direct IP address for NTP servers, instead of using "pool.ntp.org" , but pool.ntp.org of course frowns upon that.

Any thoughts?

[karfas]

You might use one of the DNS client libs, e.g. https://github.com/vshymanskyy/aiodns and look up the hostname/address using these.

However, I think I don't grasp your problem in full.
Looking up pool.ntp.org will not work in a "no internet scenario", as well as NTP requests to pool.ntp.org, so I think your app should not call any of these functions in absence of "internet access".

[OscamSatUser]

I could do a ping before calling the getaddrinfo , but that just seems like unnecessary layers, and it doesn't guarantee the getaddrinfo will still return within the WDT expiration.

[jomasnash]

Which begs the question of how you test for the presence of the internet :). A nontrivial requirement.

[EDIT] One solution, if using aiodns is to subject the DNS query to a timeout (via asyncio.wait_for). If it times out on a query to known good server, you can deduce that the internet is absent. The question is how long you're prepared to wait for an answer, or how long a delay counts as down?

I'd be interested to hear of better ways to do this.

I think asyncio.wait_for will also not work when using a blocking (getaddrinfo) function. The only solution is a nonblocking async getaddrinfo. Since this is not implemented yet, I wrote my own async getaddrinfo. It might not be perfect but is works for ME.

import asyncio
import struct
from socket import socket, AF_INET, AF_INET6, SOCK_STREAM, SOCK_DGRAM, IPPROTO_TCP
from network import WLAN
import select
import time

async def getaddrinfo(host, port, af=AF_INET, type=SOCK_STREAM, proto=0, flags=0, dns=None, timeout=1000):
    if not dns:
        dns = WLAN(WLAN.IF_STA).ipconfig('gw4')
    urls = host.split('.')
    size=len(urls)
    for u in urls:
        size += len(u)
    url_pkt = bytearray(size)
    idx = 0
    for u in urls:
        url_pkt[idx] = len(u)
        idx += 1
        for c in u:
            url_pkt[idx] = ord(c)
            idx += 1
    idp = struct.pack( "<H", time.ticks_us() & 0xffff )
    hdr = b'\x01\x00\x00\x01\x00\x00\x00\x00\x00\x00'
    tail = b'\x00\x00\x01\x00\x01'
    packet = idp + hdr + url_pkt + tail
    s = socket(AF_INET, SOCK_DGRAM)
    s.setblocking(False)
    poller = select.poll()
    poller.register(s, select.POLLIN)
    msg = b''
    ts = time.ticks_ms()
    s.sendto(packet, (dns, 53))
    while time.ticks_diff(time.ticks_ms(), ts) < timeout:
        try:
            if poller.poll(0):
                msg += s.recv(512)
                RCODE = msg[3] & 0b00001111
                if RCODE != 0:
                    raise Exception(f"RCODE={RCODE}")
                ANCOUNT = 256*msg[6] + msg[7]
                if ANCOUNT == 0:
                    raise Exception(f"ANCOUNT={ANCOUNT}")
                idx = 12
                while msg[idx] != 0:
                    idx += 1
                idx += 5
                for i in range(ANCOUNT):
                    TYPE = 256*msg[idx+2] + msg[idx+3]
                    CLASS = 256*msg[idx+4] + msg[idx+5]
                    RDLEN = 256*msg[idx+10] + msg[idx+11]
                    if TYPE == 1 and CLASS == 1 and RDLEN == 4:
                        ip=f'{msg[idx+12]}.{msg[idx+13]}.{msg[idx+14]}.{msg[idx+15]}'
                        s.close()
                        return [(AF_INET, SOCK_STREAM, proto, host, (ip,port))]
                    idx += 12 + RDLEN
                s.close()
                raise Exception("NO-IP")
            await asyncio.sleep(0)
        except IndexError:
            continue
        except Exception as e:
            s.close()
            raise(e)
    s.close()
    raise Exception("NO-IP")

[peterhinch]

I think asyncio.wait_for will also not work when using a blocking (getaddrinfo) function.

You may have missed my point here. I wrote "One solution, if using aiodns is to subject the DNS query to a timeout (via asyncio.wait_for)." - aiodns is an asynchronous, nonblocking, DNS client.

[jomasnash]

aiodns is not asynchronous because it uses getaddrinfo which IS blocking.

[karfas]

@jomasnash: I think you are wrong here. Yes, aiodns uses the original getaddrinfo(), but only with IP addresses as first parameter. With a name as parameter, aiodns.getaddrinfo() runs a DNS query similar to your solution...

[peterhinch]

Unfortunately DNS lookup via getaddrinfo blocks, so a slow internet connection or DNS server can cause problems. There is an unofficial asynchronous DNS client which would enable you to feed the WDT while waiting.

[OscamSatUser]

Thank you. Ill look into the asynchronous dns client. aiodns. I think that might be the way to go. In the meantime, i stopped using a hostname and switched to ipv4 so that i am not using the 'getaddrinfo' function. I'm really surprised you cannot set a timeout as a passed parameter when calling the function.

[OscamSatUser]

thank you @jomasnash , I modified your modification because I do not want to utilize asyncio, and i only want a IP address returned. The timeout feature is exactly what i need. I have it jump out at 5 seconds. If it doesn't catch it by then, it aborts out returning null and allowing me to feed the WDT and prevent un wanted reboot. The code also doesn't add any additional overhead to what i already have within existing code.

Thank you.

def getaddrinfo(host, port=80, af=AF_INET, type=SOCK_STREAM, proto=0, flags=0, dns=None, timeout=5000):
    if not dns:
        sta_if = network.WLAN(network.STA_IF)
        dns = sta_if.ifconfig()[2] # Index 2 is often the Gateway/DNS
    urls = host.split('.')
    size=len(urls)
    for u in urls:
        size += len(u)
    url_pkt = bytearray(size)
    idx = 0
    for u in urls:
        url_pkt[idx] = len(u)
        idx += 1
        for c in u:
            url_pkt[idx] = ord(c)
            idx += 1
    idp = struct.pack( "<H", time.ticks_us() & 0xffff )
    hdr = b'\x01\x00\x00\x01\x00\x00\x00\x00\x00\x00'
    tail = b'\x00\x00\x01\x00\x01'
    packet = idp + hdr + url_pkt + tail
    s = socket(AF_INET, SOCK_DGRAM)
    s.setblocking(False)
    poller = select.poll()
    poller.register(s, select.POLLIN)
    msg = b''
    ts = time.ticks_ms()
    s.sendto(packet, (dns, 53))
    while time.ticks_diff(time.ticks_ms(), ts) < timeout:
        try:
            if poller.poll(0):
                msg += s.recv(512)
                RCODE = msg[3] & 0b00001111
                if RCODE != 0:
                    raise Exception(f"RCODE={RCODE}")
                ANCOUNT = 256*msg[6] + msg[7]
                if ANCOUNT == 0:
                    raise Exception(f"ANCOUNT={ANCOUNT}")
                idx = 12
                while msg[idx] != 0:
                    idx += 1
                idx += 5
                for i in range(ANCOUNT):
                    TYPE = 256*msg[idx+2] + msg[idx+3]
                    CLASS = 256*msg[idx+4] + msg[idx+5]
                    RDLEN = 256*msg[idx+10] + msg[idx+11]
                    if TYPE == 1 and CLASS == 1 and RDLEN == 4:
                        ip=f'{msg[idx+12]}.{msg[idx+13]}.{msg[idx+14]}.{msg[idx+15]}'
                        s.close()
                        return ip
                    idx += 12 + RDLEN
                s.close()
                #raise Exception("NO-IP")
            time.sleep_ms(100) 
        except IndexError:
            continue
        except Exception as e:
            s.close()
            #raise(e)
    s.close()
    #raise Exception("NO-IP")