CI: systest: safer random_rfc1918_subnet

Our aarch64 CI system uses 172.31.0.0/20. Because I was (and am)
lazy, my random_rfc1918_subnet() helper was only checking /24.
This causes flakes.

Solution is to actually do it right: binary arithmetic, prefix
matching. This is effectively impossible in bash, so, use a
hairy perl helper and add copious tests.

Fixes: #18693

Signed-off-by: Ed Santiago <[email protected]>

test/system/helpers.network.bash

@@ -126,11 +126,13 @@ function random_rfc1918_subnet() {
     local retries=1024
 
     while [ "$retries" -gt 0 ];do
-        local cidr=172.$(( 16 + $RANDOM % 16 )).$(( $RANDOM & 255 ))
+        # 172.16.0.0 -> 172.31.255.255
+        local n1=172
+        local n2=$(( 16 + $RANDOM & 15 ))
+        local n3=$(( $RANDOM & 255 ))
 
-        in_use=$(ip route list | grep -F $cidr)
-        if [ -z "$in_use" ]; then
-            echo "$cidr"
+        if ! subnet_in_use $n1 $n2 $n3; then
+            echo "$n1.$n2.$n3"
             return
         fi
 
@@ -140,6 +142,69 @@ function random_rfc1918_subnet() {
     die "Could not find a random not-in-use rfc1918 subnet"
 }
 
+# subnet_in_use() - true if subnet already routed on host
+function subnet_in_use() {
+    local subnet_script=${PODMAN_TMPDIR-/var/tmp}/subnet-in-use
+    rm -f $subnet_script
+
+    # This would be a nightmare to do in bash. ipcalc, ipcalc-ng, sipcalc
+    # would be nice but are unavailable some environments (cough RHEL).
+    # Likewise python/perl netmask modules. So, use bare-bones perl.
+    cat >$subnet_script <<"EOF"
+#!/usr/bin/env perl
+
+use strict;
+use warnings;
+
+# 3 octets, in binary: 172.16.x -> 1010 1100 0000 1000 xxxx xxxx ...
+my $subnet_to_check = sprintf("%08b%08b%08b", @ARGV);
+
+my $found = 0;
+
+# Input is "ip route list", one or more lines like '10.0.0.0/8 via ...'
+while (<STDIN>) {
+    # Only interested in x.x.x.x/n lines
+    if (m!^([\d.]+)/(\d+)!) {
+        my ($ip, $bits) = ($1, $2);
+
+        # Our caller has /24 granularity, so treat /30 on host as /24.
+        $bits = 24 if $bits > 24;
+
+        # Temporary: entire subnet as binary string. 4 octets, split,
+        # then represented as a 32-bit binary string.
+        my $net = sprintf("%08b%08b%08b%08b", split(/\./, $ip));
+
+        # Now truncate those 32 bits down to the route's netmask size.
+        # This is the actual subnet range in use on the host.
+        my $net_truncated = sprintf("%.*s", $bits, $net);
+
+        # Desired subnet is in use if it matches a host route prefix
+#        print STDERR "--- $subnet_to_check in $net_truncated (@ARGV in $ip/$bits)\n";
+        $found = 1 if $subnet_to_check =~ /^$net_truncated/;
+    }
+}
+
+# Convert to shell exit status (0 = success)
+exit !$found;
+EOF
+
+    chmod 755 $subnet_script
+
+    # This runs 'ip route list', converts x.x.x.x/n to its binary prefix,
+    # then checks if our desired subnet matches that prefix (i.e. is in
+    # that range). Existing routes with size greater than 24 are
+    # normalized to /24 because that's the granularity of our
+    # random_rfc1918_subnet code.
+    #
+    # Contrived examples:
+    #    127.0.0.0/1   -> 0
+    #    128.0.0.0/1   -> 1
+    #    10.0.0.0/8    -> 00001010
+    #
+    # I'm so sorry for the ugliness.
+    ip route list | $subnet_script $*
+}
+
 # ipv4_get_route_default() - Print first default IPv4 route reported by netlink
 # $1:	Optional output of 'ip -j -4 route show' from a different context
 function ipv4_get_route_default() {

test/system/helpers.t

@@ -18,6 +18,10 @@ die() {
 testnum=0
 rc=0
 
+# Possibly used by the code we're testing
+PODMAN_TMPDIR=$(mktemp -d --tmpdir=${TMPDIR:-/tmp} podman_helper_tests.XXXXXX)
+trap 'rm -rf $PODMAN_TMPDIR' 0
+
 ###############################################################################
 # BEGIN test the parse_table helper
 
@@ -242,6 +246,42 @@ done < <(parse_table "$table")
 
 # END   ipv6_to_procfs
 ###############################################################################
+# BEGIN subnet_in_use  ...  because that's complicated
+
+# Override ip command
+function ip() {
+    echo "default foo"
+    echo "192.168.0.0/16"
+    echo "172.17.2.3/30"
+    echo "172.128.0.0/9"
+}
+
+# x.y.z | result (1 = in use, 0 = not in use - opposite of exit code)
+table="
+172 |   0 |   0  | 0
+172 |   0 | 255  | 0
+172 |   1 |   1  | 0
+172 |   1 |   2  | 0
+172 |   1 |   3  | 0
+172 |  17 |   1  | 0
+172 |  17 |   2  | 1
+172 |  17 |   3  | 0
+172 | 127 |   0  | 0
+172 | 128 |   0  | 1
+172 | 255 |   2  | 1
+192 | 168 |   1  | 1
+"
+
+while read n1 n2 n3 expect; do
+    subnet_in_use $n1 $n2 $n3
+    actual=$?
+    check_result "$((1 - $actual))" "$expect" "subnet_in_use $n1.$n2.$n3"
+done < <(parse_table "$table")
+
+unset -f ip
+
+# END   subnet_in_use
+###############################################################################
 # BEGIN check_assert
 #
 # This is way, way more complicated than it should be. The purpose is