chardev/char-udp.c | 2 -- tests/unit/test-char.c | 2 ++ 2 files changed, 2 insertions(+), 2 deletions(-)
This patch removes the `*be_opened = false` override for the UDP chardev
backend. Since UDP is connectionless it never sends a `CHR_EVENT_OPENED`
so it is never marked open. This causes some frontends (e.g. virtio-serial)
to never perform any operations on the socket.
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2993
Signed-off-by: Eric K <erickra@cs.utexas.edu>
---
chardev/char-udp.c | 2 --
tests/unit/test-char.c | 2 ++
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/chardev/char-udp.c b/chardev/char-udp.c
index 572fab0ad1..1025f577a0 100644
--- a/chardev/char-udp.c
+++ b/chardev/char-udp.c
@@ -215,8 +215,6 @@ static void qmp_chardev_open_udp(Chardev *chr,
g_free(name);
s->ioc = QIO_CHANNEL(sioc);
- /* be isn't opened until we get a connection */
- *be_opened = false;
}
static void char_udp_class_init(ObjectClass *oc, const void *data)
diff --git a/tests/unit/test-char.c b/tests/unit/test-char.c
index 8a98e42cad..2869c4e09d 100644
--- a/tests/unit/test-char.c
+++ b/tests/unit/test-char.c
@@ -1012,6 +1012,8 @@ static void char_udp_test_internal(Chardev *reuse_chr, int sock)
qemu_chr_fe_init(fe, chr, &error_abort);
}
+ g_assert(chr->be_open);
+
d.chr = chr;
qemu_chr_fe_set_handlers(fe, socket_can_read_hello, socket_read_hello,
NULL, NULL, &d, NULL, true);
--
2.52.0
Hi Eric On Tue, Dec 9, 2025 at 5:18 AM Eric K <erickra@cs.utexas.edu> wrote: > > This patch removes the `*be_opened = false` override for the UDP chardev > backend. Since UDP is connectionless it never sends a `CHR_EVENT_OPENED` > so it is never marked open. This causes some frontends (e.g. virtio-serial) > to never perform any operations on the socket. > > Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2993 > Signed-off-by: Eric K <erickra@cs.utexas.edu> UDP is connection-less, so it will not be notified when a client is ready to receive. If we make the chardev always open, the device may send data too early though. At the same time, a chardev that only reads isn't very useful. I don't think we need to introduce a property for the change of behaviour. Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com> > --- > chardev/char-udp.c | 2 -- > tests/unit/test-char.c | 2 ++ > 2 files changed, 2 insertions(+), 2 deletions(-) > > diff --git a/chardev/char-udp.c b/chardev/char-udp.c > index 572fab0ad1..1025f577a0 100644 > --- a/chardev/char-udp.c > +++ b/chardev/char-udp.c > @@ -215,8 +215,6 @@ static void qmp_chardev_open_udp(Chardev *chr, > g_free(name); > > s->ioc = QIO_CHANNEL(sioc); > - /* be isn't opened until we get a connection */ > - *be_opened = false; > } > > static void char_udp_class_init(ObjectClass *oc, const void *data) > diff --git a/tests/unit/test-char.c b/tests/unit/test-char.c > index 8a98e42cad..2869c4e09d 100644 > --- a/tests/unit/test-char.c > +++ b/tests/unit/test-char.c > @@ -1012,6 +1012,8 @@ static void char_udp_test_internal(Chardev *reuse_chr, int sock) > qemu_chr_fe_init(fe, chr, &error_abort); > } > > + g_assert(chr->be_open); > + > d.chr = chr; > qemu_chr_fe_set_handlers(fe, socket_can_read_hello, socket_read_hello, > NULL, NULL, &d, NULL, true); > -- > 2.52.0 > > -- Marc-André Lureau
On Tue, Dec 09, 2025 at 12:02:10PM +0400, Marc-André Lureau wrote: > Hi Eric > > On Tue, Dec 9, 2025 at 5:18 AM Eric K <erickra@cs.utexas.edu> wrote: > > > > This patch removes the `*be_opened = false` override for the UDP chardev > > backend. Since UDP is connectionless it never sends a `CHR_EVENT_OPENED` > > so it is never marked open. This causes some frontends (e.g. virtio-serial) > > to never perform any operations on the socket. > > > > Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2993 > > Signed-off-by: Eric K <erickra@cs.utexas.edu> > > UDP is connection-less, so it will not be notified when a client is > ready to receive. If we make the chardev always open, the device may > send data too early though. > > At the same time, a chardev that only reads isn't very useful. Loosing data due to no client in the receive end is the price users willing accept when they choose to use the UDP backend. If they need reliable data transfer we have better chardev backends than UDP available to use. > I don't think we need to introduce a property for the change of behaviour. > > Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com> > > > --- > > chardev/char-udp.c | 2 -- > > tests/unit/test-char.c | 2 ++ > > 2 files changed, 2 insertions(+), 2 deletions(-) > > > > diff --git a/chardev/char-udp.c b/chardev/char-udp.c > > index 572fab0ad1..1025f577a0 100644 > > --- a/chardev/char-udp.c > > +++ b/chardev/char-udp.c > > @@ -215,8 +215,6 @@ static void qmp_chardev_open_udp(Chardev *chr, > > g_free(name); > > > > s->ioc = QIO_CHANNEL(sioc); > > - /* be isn't opened until we get a connection */ > > - *be_opened = false; > > } > > > > static void char_udp_class_init(ObjectClass *oc, const void *data) > > diff --git a/tests/unit/test-char.c b/tests/unit/test-char.c > > index 8a98e42cad..2869c4e09d 100644 > > --- a/tests/unit/test-char.c > > +++ b/tests/unit/test-char.c > > @@ -1012,6 +1012,8 @@ static void char_udp_test_internal(Chardev *reuse_chr, int sock) > > qemu_chr_fe_init(fe, chr, &error_abort); > > } > > > > + g_assert(chr->be_open); > > + > > d.chr = chr; > > qemu_chr_fe_set_handlers(fe, socket_can_read_hello, socket_read_hello, > > NULL, NULL, &d, NULL, true); > > -- > > 2.52.0 > > > > > > > -- > Marc-André Lureau > With regards, Daniel -- |: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
© 2016 - 2025 Red Hat, Inc.