Remaining buffer from socket isn't available anymore after calling socket.recv the first time

[Sworddragon]

BPO	21819

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2014-06-21.10:08:56.480>
created_at = <Date 2014-06-21.05:04:51.229>
labels = ['invalid', 'type-bug', 'library']
title = "Remaining buffer from socket isn't available anymore after calling socket.recv the first time"
updated_at = <Date 2014-06-22.06:40:56.514>
user = 'https://bugs.python.org/Sworddragon'

bugs.python.org fields:

activity = <Date 2014-06-22.06:40:56.514>
actor = 'neologix'
assignee = 'none'
closed = True
closed_date = <Date 2014-06-21.10:08:56.480>
closer = 'neologix'
components = ['Library (Lib)']
creation = <Date 2014-06-21.05:04:51.229>
creator = 'Sworddragon'
dependencies = []
files = []
hgrepos = []
issue_num = 21819
keywords = []
message_count = 4.0
messages = ['221155', '221164', '221182', '221223']
nosy_count = 2.0
nosy_names = ['neologix', 'Sworddragon']
pr_nums = []
priority = 'normal'
resolution = 'not a bug'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue21819'
versions = ['Python 3.4']

[Sworddragon]

If I'm receiving data from a socket (several bytes) and making the first call to socket.recv(1) all is fine but the second call won't get any further data. But doing this again with socket.recv(2) instead will successfully get the 2 bytes. Here is a testcase:

Script:

def icmp_packet(type, code, data):
	length_data = len(data)
	if length_data % 2 == 1:
		data += b'\x00'
	checksum = code | type << 8
	i = 0
	while i < length_data:
		checksum += data[i + 1] | data[i] << 8
		checksum = (checksum & 65535) + (checksum >> 16)
		i += 2
	return bytes([type]) + bytes([code]) + (checksum ^ 65535).to_bytes(2, 'big') + data

import socket

connection = socket.socket(proto = socket.IPPROTO_ICMP, type = socket.SOCK_RAW)
connection.settimeout(1)
connection.sendto(icmp_packet(8, 0, b'\x00\x00\x00\x00'), ('8.8.8.8', 0))
print(connection.recv(2))
connection.close()
connection = socket.socket(proto = socket.IPPROTO_ICMP, type = socket.SOCK_RAW)
connection.settimeout(1)
connection.sendto(icmp_packet(8, 0, b'\x00\x00\x00\x00'), ('8.8.8.8', 0))
print(connection.recv(1))
print(connection.recv(1))
connection.close()

Here is the result:

root@ubuntu:/home/sworddragon/tmp# python3 test.py
b'E\x00'
b'E'
Traceback (most recent call last):
  File "test.py", line 24, in <module>
    print(connection.recv(1))
socket.timeout: timed out

[neologix]

If I'm receiving data from a socket (several bytes) and making the
first call to socket.recv(1) all is fine but the second call won't get
any further data. But doing this again with socket.recv(2) instead will
successfully get the 2 bytes. Here is a testcase:

First, note that Python just calles the underlying recv() syscall, so it's not at fault here.

And actually, noone's at fault here, because what you're trying to do doesn't make sense: ICMP is datagram-oriented, so you should use recvfrom(): and if you try to receive less bytes than the datagram size, the rest will be discarded, like UDP.

[Sworddragon]

and if you try to receive less bytes than the datagram size, the rest will be discarded, like UDP.

I'm wondering how would it be possible then to fetch packets of an unknown size without using an extremely big buffer.

[neologix]

I'm wondering how would it be possible then to fetch packets of an unknown size without using an extremely big buffer.

IP packets are limited to 64K, so just pass a 64K buffer, that's not
"extremely big".
If you really wanted to avoid this, you could try the FIONREAD ioctl,
but I wouldn't advise it.