Re: [Qemu-devel] [Bug 1837909] [NEW] test-char fails if host has no network interfaces

[Philippe Mathieu-Daudé]

On 7/25/19 4:20 PM, Cole Robinson wrote:
> Public bug reported:
> 
> # ./tests/test-char 
> # random seed: R02S8602535bf831a74bca571d8c416d8161
> 1..34
> # Start of char tests
> ...
> ok 12 /char/websocket
> # Start of stdio tests
> # End of stdio tests
> # Start of socket tests
> # Start of server tests
> # Start of mainloop tests
> Unexpected error in inet_parse_connect_saddr() at util/qemu-sockets.c:421:
> # 
> # address resolution failed for 127.0.0.1:42275: Name or service not known
> # 
> 
> Aborted (core dumped)
> 
> 
> # ip a
> 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group 
> default qlen 1000
>     link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
>     inet 127.0.0.1/8 scope host lo
>        valid_lft forever preferred_lft forever
>     inet6 ::1/128 scope host 
>        valid_lft forever preferred_lft forever
> 
> 
> This seems to be related to use of AI_ADDRCONFIG in qemu-sockets.c 
> inet_parse_connect_saddr, dropping it fixes the test. 'man getaddrinfo' makes 
> it sound like AI_ADDRCONFIG requires the host to have a non-loopback ipv4 or 
> ipv6 address available

GETADDRINFO(3)

  If hints.ai_flags includes the AI_ADDRCONFIG flag, then IPv4
  addresses are returned in the list pointed to by res only if
  the local system has at least one IPv4 address configured, and
  IPv6 addresses are returned only if the local system has at
  least one IPv6 address configured.  The loopback address is not
  considered for this case as valid as a configured address.
  This flag is useful on, for example, IPv4-only systems, to
  ensure  that  getaddrinfo() does not return IPv6 socket addresses
  that would always fail in connect(2) or bind(2).

I'm a little confused, and I don't feel fluent enough with English to be
sure that "only if A and only if B" is equivalent to "requires (A or
B)". Maybe the man page should use 'or' instead of 'and' here...

> This host setup may seem niche, but it is what the 'mock' RPM build tool
> has by default. In Fedora we run the test suite during the RPM build, so
> the failing test causes a bit of pain for certain workflows

Would this diff snippet help?

-- >8 --
diff --git a/util/qemu-sockets.c b/util/qemu-sockets.c
index a5092dbd12..9ad775270d 100644
--- a/util/qemu-sockets.c
+++ b/util/qemu-sockets.c
@@ -417,7 +417,7 @@ static struct addrinfo
*inet_parse_connect_saddr(InetSocketAddress *saddr,
         ai.ai_flags &= ~AI_V4MAPPED;
         rc = getaddrinfo(saddr->host, saddr->port, &ai, &res);
     }
-    if (rc != 0) {
+    if (rc and rc != EAI_NONAME) {
         error_setg(errp, "address resolution failed for %s:%s: %s",
                    saddr->host, saddr->port, gai_strerror(rc));
         return NULL;
---

> 
> ** Affects: qemu
>      Importance: Undecided
>          Status: New
>