[Qemu-devel] [PATCH 0/2] Small fixes for qemu-sockets ipv4/ipv6 logic.

[Qemu-devel] [PATCH 0/2] Small fixes for qemu-sockets ipv4/ipv6 logic.

[Paolo Bonzini]

While reviewing the L2TPv3 submission, I wanted to tell Anton that
IPv4/IPv6 is already done in qemu-sockets.c.  Turns out it is not
exactly the case; these patches fixes the two issues I spotted.

Paolo Bonzini (2):
  socket: treat ipv4=on,ipv6=on uniformly
  socket: handle ipv4/ipv6 in socket_dgram

 util/qemu-sockets.c | 39 ++++++++++++++++++++++++++-------------
 1 file changed, 26 insertions(+), 13 deletions(-)

-- 
1.8.5.3

[Qemu-devel] [PATCH 1/2] socket: treat ipv4=on,ipv6=on uniformly

[Paolo Bonzini]

In some cases, "ipv4=on,ipv6=on" means "try both kinds of address";
in others, it means "try IPv6 only" just by virtue of how the code
is structured.

Fix this to make things more consistent, and adjust coding style too.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 util/qemu-sockets.c | 33 ++++++++++++++++++++-------------
 1 file changed, 20 insertions(+), 13 deletions(-)

diff --git a/util/qemu-sockets.c b/util/qemu-sockets.c
index 8818d7c..7a9065b 100644
--- a/util/qemu-sockets.c
+++ b/util/qemu-sockets.c
@@ -127,10 +127,13 @@ int inet_listen_opts(QemuOpts *opts, int port_offset, Error **errp)
     addr = qemu_opt_get(opts, "host");
 
     to = qemu_opt_get_number(opts, "to", 0);
-    if (qemu_opt_get_bool(opts, "ipv4", 0))
-        ai.ai_family = PF_INET;
-    if (qemu_opt_get_bool(opts, "ipv6", 0))
-        ai.ai_family = PF_INET6;
+    if (qemu_opt_get_bool(opts, "ipv4", 0) != qemu_opt_get_bool(opts, "ipv6", 0)) {
+        if (qemu_opt_get_bool(opts, "ipv4", 0)) {
+            ai.ai_family = PF_INET;
+        } else {
+            ai.ai_family = PF_INET6;
+        }
+    }
 
     /* lookup */
     if (port_offset)
@@ -319,11 +322,12 @@ static struct addrinfo *inet_parse_connect_opts(QemuOpts *opts, Error **errp)
         return NULL;
     }
 
-    if (qemu_opt_get_bool(opts, "ipv4", 0)) {
-        ai.ai_family = PF_INET;
-    }
-    if (qemu_opt_get_bool(opts, "ipv6", 0)) {
-        ai.ai_family = PF_INET6;
+    if (qemu_opt_get_bool(opts, "ipv4", 0) != qemu_opt_get_bool(opts, "ipv6", 0)) {
+        if (qemu_opt_get_bool(opts, "ipv4", 0)) {
+            ai.ai_family = PF_INET;
+        } else {
+            ai.ai_family = PF_INET6;
+        }
     }
 
     /* lookup */
@@ -418,10 +422,13 @@ int inet_dgram_opts(QemuOpts *opts, Error **errp)
         return -1;
     }
 
-    if (qemu_opt_get_bool(opts, "ipv4", 0))
-        ai.ai_family = PF_INET;
-    if (qemu_opt_get_bool(opts, "ipv6", 0))
-        ai.ai_family = PF_INET6;
+    if (qemu_opt_get_bool(opts, "ipv4", 0) != qemu_opt_get_bool(opts, "ipv6", 0)) {
+        if (qemu_opt_get_bool(opts, "ipv4", 0)) {
+            ai.ai_family = PF_INET;
+        } else {
+            ai.ai_family = PF_INET6;
+        }
+    }
 
     if (0 != (rc = getaddrinfo(addr, port, &ai, &peer))) {
         error_setg(errp, "address resolution failed for %s:%s: %s", addr, port,
-- 
1.8.5.3

Re: [Qemu-devel] [PATCH 1/2] socket: treat ipv4=on, ipv6=on uniformly

[Markus Armbruster]

Paolo Bonzini <pbonzini@redhat.com> writes:

> In some cases, "ipv4=on,ipv6=on" means "try both kinds of address";
> in others, it means "try IPv6 only" just by virtue of how the code
> is structured.
>
> Fix this to make things more consistent, and adjust coding style too.

Begs the question which one you're picking as the consistent
interpretation, "both" or "v6 only".

Re: [Qemu-devel] [PATCH 1/2] socket: treat ipv4=on, ipv6=on uniformly

[Gerd Hoffmann]

On Fr, 2014-02-28 at 11:16 +0100, Paolo Bonzini wrote:
> In some cases, "ipv4=on,ipv6=on" means "try both kinds of address";
> in others, it means "try IPv6 only" just by virtue of how the code
> is structured.

> @@ -127,10 +127,13 @@ int inet_listen_opts(QemuOpts *opts, int port_offset, Error **errp)

> -    if (qemu_opt_get_bool(opts, "ipv4", 0))
> -        ai.ai_family = PF_INET;
> -    if (qemu_opt_get_bool(opts, "ipv6", 0))
> -        ai.ai_family = PF_INET6;
> +    if (qemu_opt_get_bool(opts, "ipv4", 0) != qemu_opt_get_bool(opts, "ipv6", 0)) {
> +        if (qemu_opt_get_bool(opts, "ipv4", 0)) {
> +            ai.ai_family = PF_INET;
> +        } else {
> +            ai.ai_family = PF_INET6;
> +        }
> +    }

This is wrong for the listening side.

ipv6 sockets can listen on both ipv4 and ipv6.  qemu configures ipv6
sockets to do that, unconditionally.

So ipv4=yes,ipv6=no works correctly.
ipv4=yes,ipv6=yes works too.
ipv4=no,ipv6=yes doesn't work, but to fix that you have to set the
IPV6_V6ONLY socket option according to the ipv4 option.

Canevat: Listening on both ipv4+ipv6 with a single file handle works for
the wildcard address only.  Specifying -- say --
host=localhost,ipv4=yes,ipv6=yes, then expect qemu to listen on both
127.0.0.1 and ::1 doesn't work.

This can only be fixed by allowing multiple listening filehandles.
Which is a non-trivial change as this also affects the public api (which
will have to report a list of listening addresses instead of a single
address).
 
> @@ -319,11 +322,12 @@ static struct addrinfo *inet_parse_connect_opts(QemuOpts *opts, Error **errp)

> -    if (qemu_opt_get_bool(opts, "ipv4", 0)) {
> -        ai.ai_family = PF_INET;
> -    }
> -    if (qemu_opt_get_bool(opts, "ipv6", 0)) {
> -        ai.ai_family = PF_INET6;
> +    if (qemu_opt_get_bool(opts, "ipv4", 0) != qemu_opt_get_bool(opts, "ipv6", 0)) {
> +        if (qemu_opt_get_bool(opts, "ipv4", 0)) {
> +            ai.ai_family = PF_INET;
> +        } else {
> +            ai.ai_family = PF_INET6;
> +        }

For connect this change is fine I think.

> @@ -418,10 +422,13 @@ int inet_dgram_opts(QemuOpts *opts, Error **errp)

> -    if (qemu_opt_get_bool(opts, "ipv4", 0))
> -        ai.ai_family = PF_INET;
> -    if (qemu_opt_get_bool(opts, "ipv6", 0))
> -        ai.ai_family = PF_INET6;
> +    if (qemu_opt_get_bool(opts, "ipv4", 0) != qemu_opt_get_bool(opts, "ipv6", 0)) {
> +        if (qemu_opt_get_bool(opts, "ipv4", 0)) {
> +            ai.ai_family = PF_INET;
> +        } else {
> +            ai.ai_family = PF_INET6;
> +        }
> +    }

Not fully sure, but I think this is fine too.

cheers,
  Gerd

Re: [Qemu-devel] [PATCH 1/2] socket: treat ipv4=on, ipv6=on uniformly

[Paolo Bonzini]

Il 28/02/2014 13:23, Gerd Hoffmann ha scritto:
> On Fr, 2014-02-28 at 11:16 +0100, Paolo Bonzini wrote:
>> In some cases, "ipv4=on,ipv6=on" means "try both kinds of address";
>> in others, it means "try IPv6 only" just by virtue of how the code
>> is structured.
>
>> @@ -127,10 +127,13 @@ int inet_listen_opts(QemuOpts *opts, int port_offset, Error **errp)
>
>> -    if (qemu_opt_get_bool(opts, "ipv4", 0))
>> -        ai.ai_family = PF_INET;
>> -    if (qemu_opt_get_bool(opts, "ipv6", 0))
>> -        ai.ai_family = PF_INET6;
>> +    if (qemu_opt_get_bool(opts, "ipv4", 0) != qemu_opt_get_bool(opts, "ipv6", 0)) {
>> +        if (qemu_opt_get_bool(opts, "ipv4", 0)) {
>> +            ai.ai_family = PF_INET;
>> +        } else {
>> +            ai.ai_family = PF_INET6;
>> +        }
>> +    }
>
> This is wrong for the listening side.
>
> ipv6 sockets can listen on both ipv4 and ipv6.  qemu configures ipv6
> sockets to do that, unconditionally.
>
> So ipv4=yes,ipv6=no works correctly.
> ipv4=yes,ipv6=yes works too.
> ipv4=no,ipv6=yes doesn't work, but to fix that you have to set the
> IPV6_V6ONLY socket option according to the ipv4 option.
>
> Canevat: Listening on both ipv4+ipv6 with a single file handle works for
> the wildcard address only.  Specifying -- say --
> host=localhost,ipv4=yes,ipv6=yes, then expect qemu to listen on both
> 127.0.0.1 and ::1 doesn't work.
>
> This can only be fixed by allowing multiple listening filehandles.
> Which is a non-trivial change as this also affects the public api (which
> will have to report a list of listening addresses instead of a single
> address).

Thanks for teaching me!

Do you think patch 2 is okay alone?

Paolo

[Qemu-devel] [PATCH 2/2] socket: handle ipv4/ipv6 in socket_dgram

[Paolo Bonzini]

We were forgetting to set the ipv4 and ipv6 QemuOpts properties, so
socket_dgram was always trying _both_ ipv4 and ipv6.  Forward the
properties correctly.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 util/qemu-sockets.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/util/qemu-sockets.c b/util/qemu-sockets.c
index 7a9065b..7391aee 100644
--- a/util/qemu-sockets.c
+++ b/util/qemu-sockets.c
@@ -958,15 +958,22 @@ int socket_dgram(SocketAddress *remote, SocketAddress *local, Error **errp)
 
     opts = qemu_opts_create(&socket_optslist, NULL, 0, &error_abort);
     switch (remote->kind) {
-    case SOCKET_ADDRESS_KIND_INET:
+    case SOCKET_ADDRESS_KIND_INET: {
+        bool ipv4 = remote->inet->ipv4 || !remote->inet->has_ipv4;
+        bool ipv6 = remote->inet->ipv6 || !remote->inet->has_ipv6;
         qemu_opt_set(opts, "host", remote->inet->host);
         qemu_opt_set(opts, "port", remote->inet->port);
         if (local) {
             qemu_opt_set(opts, "localaddr", local->inet->host);
             qemu_opt_set(opts, "localport", local->inet->port);
         }
+        if (!ipv4 || !ipv6) {
+            qemu_opt_set_bool(opts, "ipv4", ipv4);
+            qemu_opt_set_bool(opts, "ipv6", ipv6);
+        }
         fd = inet_dgram_opts(opts, errp);
         break;
+    }
 
     default:
         error_setg(errp, "socket type unsupported for datagram");
-- 
1.8.5.3

Re: [Qemu-devel] [PATCH 2/2] socket: handle ipv4/ipv6 in socket_dgram

[Gerd Hoffmann]

On Fr, 2014-02-28 at 11:16 +0100, Paolo Bonzini wrote:
> -    case SOCKET_ADDRESS_KIND_INET:
> +    case SOCKET_ADDRESS_KIND_INET: {
> +        bool ipv4 = remote->inet->ipv4 || !remote->inet->has_ipv4;
> +        bool ipv6 = remote->inet->ipv6 || !remote->inet->has_ipv6;
>          qemu_opt_set(opts, "host", remote->inet->host);
>          qemu_opt_set(opts, "port", remote->inet->port);
>          if (local) {
>              qemu_opt_set(opts, "localaddr", local->inet->host);
>              qemu_opt_set(opts, "localport", local->inet->port);
>          }
> +        if (!ipv4 || !ipv6) {
> +            qemu_opt_set_bool(opts, "ipv4", ipv4);
> +            qemu_opt_set_bool(opts, "ipv6", ipv6);
> +        }

I'd go for a simple

	if (remote->inet->has_ipv4) {
		qemu_opt_set_bool(opts, "ipv4", remote->inet->ipv4)
        }

cheers,
  Gerd