common: do not use DNS to determine if address is local

[tobtoht]

Alternative to #8849.

I'm not convinced we need to support DNS lookup at all here. This PR considers a node "local" if the host is "localhost" or a IPv{4,6} address in the loopback range. The is_privacy_preserving_network check is no longer needed.

[plowsof]

May i confirm my method of seeing the bug (after looking at 8849) - simply using wireshark to show that dns A records at the onion daemon address are being looked up at wallet open time?

[tobtoht]

@plowsof, yes, A or AAAA records. It should no longer happen with 8849 or this PR.

[pokkst]

@plowsof You can test it in just about every wallet too. The only wallet I've tested so far that didn't seem to experience the issue is Feather. For other wallets, it happens for onion addresses, i2p addresses, and clearnet addresses (clearnet is not really an issue if proxy is disabled, of course).

If you need an app for Android to observe network traffic then PCAPdroid on F-Droid is good.

[plowsof]

with wireshark open and a display filter of dns.qry.name contains ".onion"
with release and this PR i still see the same number of dns calls when i e.g. unlock the wallet. Am i testing with the wrong parameters? (no --proxy or torsocks, so the inputs wont work)
./monero-wallet-cli --daemon-address plowsof3t5hogddwabaeiyrno25efmzfxyro2vligremt7sxpsclfaid.onion:18089

[tobtoht]

@plowsof --proxy is needed for testing. A DNS leak happens elsewhere if --proxy is omitted, because it assumes it's connecting a clearnet host. It might be a good idea to have simplewallet warn* if a user tries to connect to a .onion/i2p daemon address without proxy, but this is a separate issue.

* Or refuse to connect, though this would frustrate users with transparent Tor proxying (e.g. Whonix).

[plowsof]

confirming that this PR fixes the issue (no dns leak with an .onion daemon and --proxy where release would otherwise do so). *using a clearnet daemon with proxy no leak

[j-berman]

Small comment on the warning log, otherwise looks good to me.

I wrote some unit tests:

#include "common/util.h"

TEST(LocalAddress, localhost) { ASSERT_TRUE(tools::is_local_address("localhost")); }
TEST(LocalAddress, localhost_port) { ASSERT_TRUE(tools::is_local_address("localhost:18081")); }
TEST(LocalAddress, localhost_suffix) { ASSERT_TRUE(tools::is_local_address("test.localhost")); }
TEST(LocalAddress, loopback) { ASSERT_TRUE(tools::is_local_address("127.0.0.1")); }
TEST(LocalAddress, loopback_port) { ASSERT_TRUE(tools::is_local_address("127.0.0.1:18081")); }
TEST(LocalAddress, loopback_protocol) { ASSERT_TRUE(tools::is_local_address("http://127.0.0.1")); }
TEST(LocalAddress, loopback_hi) { ASSERT_TRUE(tools::is_local_address("127.255.255.255")); }
TEST(LocalAddress, loopback_lo) { ASSERT_TRUE(tools::is_local_address("127.0.0.0")); }
TEST(LocalAddress, loopback_ipv6) { ASSERT_TRUE(tools::is_local_address("[0:0:0:0:0:0:0:1]")); }

TEST(LocalAddress, onion) { ASSERT_FALSE(tools::is_local_address("vww6ybal4bd7szmgncyruucpgfkqahzddi37ktceo3ah7ngmcopnpyyd.onion")); }
TEST(LocalAddress, i2p) { ASSERT_FALSE(tools::is_local_address("xmrto2bturnore26xmrto2bturnore26xmrto2bturnore26xmr2.b32.i2p")); }
TEST(LocalAddress, valid_ip) { ASSERT_FALSE(tools::is_local_address("1.2.3.4")); }
TEST(LocalAddress, valid_ipv6) { ASSERT_FALSE(tools::is_local_address("[0:0:0:0:0:0:0:2]")); }
TEST(LocalAddress, valid_domain) { ASSERT_FALSE(tools::is_local_address("getmonero.org")); }
TEST(LocalAddress, local_prefix) { ASSERT_FALSE(tools::is_local_address("localhost.com")); }
TEST(LocalAddress, invalid) { ASSERT_FALSE(tools::is_local_address("test")); }
TEST(LocalAddress, empty) { ASSERT_FALSE(tools::is_local_address("")); }

[tobtoht]

Added your unit tests and suggested change, rebased so Windows build passes. Thanks for the review. :)