The tcp_chr_free_connection & tcp_chr_disconnect methods both
skip all of their cleanup work unless the 's->connected' flag
is set. This flag is set when the incoming client connection
is ready to use. Crucially this is *after* the TLS handshake
has been completed. So if the TLS handshake fails and we try
to cleanup the failed client, all the cleanup is skipped as
's->connected' is still false.
The only important thing that should be skipped in this case
is sending of the CHR_EVENT_CLOSED, because we never got as
far as sending the corresponding CHR_EVENT_OPENED. Every other
bit of cleanup can be robust against being called even when
s->connected is false.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
---
Changed in v2:
- Remove conditional checks for NULL (Marc-André)
- Don't use camelCase variable name (Marc-André)
chardev/char-socket.c | 19 +++++++++----------
1 file changed, 9 insertions(+), 10 deletions(-)
diff --git a/chardev/char-socket.c b/chardev/char-socket.c
index e65148fe97..53eda8ef00 100644
--- a/chardev/char-socket.c
+++ b/chardev/char-socket.c
@@ -332,10 +332,6 @@ static void tcp_chr_free_connection(Chardev *chr)
SocketChardev *s = SOCKET_CHARDEV(chr);
int i;
- if (!s->connected) {
- return;
- }
-
if (s->read_msgfds_num) {
for (i = 0; i < s->read_msgfds_num; i++) {
close(s->read_msgfds[i]);
@@ -394,22 +390,25 @@ static void update_disconnected_filename(SocketChardev *s)
s->is_listen, s->is_telnet);
}
+/* NB may be called even if tcp_chr_connect has not been
+ * reached, due to TLS or telnet initialization failure,
+ * so can *not* assume s->connected == true
+ */
static void tcp_chr_disconnect(Chardev *chr)
{
SocketChardev *s = SOCKET_CHARDEV(chr);
-
- if (!s->connected) {
- return;
- }
+ bool emit_close = s->connected;
tcp_chr_free_connection(chr);
- if (s->listen_ioc) {
+ if (s->listen_ioc && s->listen_tag == 0) {
s->listen_tag = qio_channel_add_watch(
QIO_CHANNEL(s->listen_ioc), G_IO_IN, tcp_chr_accept, chr, NULL);
}
update_disconnected_filename(s);
- qemu_chr_be_event(chr, CHR_EVENT_CLOSED);
+ if (emit_close) {
+ qemu_chr_be_event(chr, CHR_EVENT_CLOSED);
+ }
if (s->reconnect_time) {
qemu_chr_socket_restart_timer(chr);
}
--
2.13.5
On 10/05/2017 10:50 AM, Daniel P. Berrange wrote: > The tcp_chr_free_connection & tcp_chr_disconnect methods both > skip all of their cleanup work unless the 's->connected' flag > is set. This flag is set when the incoming client connection > is ready to use. Crucially this is *after* the TLS handshake > has been completed. So if the TLS handshake fails and we try > to cleanup the failed client, all the cleanup is skipped as > 's->connected' is still false. > > The only important thing that should be skipped in this case > is sending of the CHR_EVENT_CLOSED, because we never got as > far as sending the corresponding CHR_EVENT_OPENED. Every other > bit of cleanup can be robust against being called even when > s->connected is false. > > Signed-off-by: Daniel P. Berrange <berrange@redhat.com> > --- > > Changed in v2: > > - Remove conditional checks for NULL (Marc-André) > - Don't use camelCase variable name (Marc-André) Reviewed-by: Eric Blake <eblake@redhat.com> -- Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
On Thu, Oct 5, 2017 at 5:50 PM, Daniel P. Berrange <berrange@redhat.com> wrote:
> The tcp_chr_free_connection & tcp_chr_disconnect methods both
> skip all of their cleanup work unless the 's->connected' flag
> is set. This flag is set when the incoming client connection
> is ready to use. Crucially this is *after* the TLS handshake
> has been completed. So if the TLS handshake fails and we try
> to cleanup the failed client, all the cleanup is skipped as
> 's->connected' is still false.
>
> The only important thing that should be skipped in this case
> is sending of the CHR_EVENT_CLOSED, because we never got as
> far as sending the corresponding CHR_EVENT_OPENED. Every other
> bit of cleanup can be robust against being called even when
> s->connected is false.
>
> Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
> ---
>
> Changed in v2:
>
> - Remove conditional checks for NULL (Marc-André)
> - Don't use camelCase variable name (Marc-André)
>
> chardev/char-socket.c | 19 +++++++++----------
> 1 file changed, 9 insertions(+), 10 deletions(-)
>
> diff --git a/chardev/char-socket.c b/chardev/char-socket.c
> index e65148fe97..53eda8ef00 100644
> --- a/chardev/char-socket.c
> +++ b/chardev/char-socket.c
> @@ -332,10 +332,6 @@ static void tcp_chr_free_connection(Chardev *chr)
> SocketChardev *s = SOCKET_CHARDEV(chr);
> int i;
>
> - if (!s->connected) {
> - return;
> - }
> -
> if (s->read_msgfds_num) {
> for (i = 0; i < s->read_msgfds_num; i++) {
> close(s->read_msgfds[i]);
> @@ -394,22 +390,25 @@ static void update_disconnected_filename(SocketChardev *s)
> s->is_listen, s->is_telnet);
> }
>
> +/* NB may be called even if tcp_chr_connect has not been
> + * reached, due to TLS or telnet initialization failure,
> + * so can *not* assume s->connected == true
> + */
> static void tcp_chr_disconnect(Chardev *chr)
> {
> SocketChardev *s = SOCKET_CHARDEV(chr);
> -
> - if (!s->connected) {
> - return;
> - }
> + bool emit_close = s->connected;
>
> tcp_chr_free_connection(chr);
>
> - if (s->listen_ioc) {
> + if (s->listen_ioc && s->listen_tag == 0) {
> s->listen_tag = qio_channel_add_watch(
> QIO_CHANNEL(s->listen_ioc), G_IO_IN, tcp_chr_accept, chr, NULL);
> }
> update_disconnected_filename(s);
> - qemu_chr_be_event(chr, CHR_EVENT_CLOSED);
> + if (emit_close) {
> + qemu_chr_be_event(chr, CHR_EVENT_CLOSED);
> + }
> if (s->reconnect_time) {
> qemu_chr_socket_restart_timer(chr);
> }
> --
> 2.13.5
>
>
--
Marc-André Lureau
On 05/10/2017 17:50, Daniel P. Berrange wrote:
> The tcp_chr_free_connection & tcp_chr_disconnect methods both
> skip all of their cleanup work unless the 's->connected' flag
> is set. This flag is set when the incoming client connection
> is ready to use. Crucially this is *after* the TLS handshake
> has been completed. So if the TLS handshake fails and we try
> to cleanup the failed client, all the cleanup is skipped as
> 's->connected' is still false.
>
> The only important thing that should be skipped in this case
> is sending of the CHR_EVENT_CLOSED, because we never got as
> far as sending the corresponding CHR_EVENT_OPENED. Every other
> bit of cleanup can be robust against being called even when
> s->connected is false.
>
> Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
> ---
>
> Changed in v2:
>
> - Remove conditional checks for NULL (Marc-André)
> - Don't use camelCase variable name (Marc-André)
>
> chardev/char-socket.c | 19 +++++++++----------
> 1 file changed, 9 insertions(+), 10 deletions(-)
>
> diff --git a/chardev/char-socket.c b/chardev/char-socket.c
> index e65148fe97..53eda8ef00 100644
> --- a/chardev/char-socket.c
> +++ b/chardev/char-socket.c
> @@ -332,10 +332,6 @@ static void tcp_chr_free_connection(Chardev *chr)
> SocketChardev *s = SOCKET_CHARDEV(chr);
> int i;
>
> - if (!s->connected) {
> - return;
> - }
> -
> if (s->read_msgfds_num) {
> for (i = 0; i < s->read_msgfds_num; i++) {
> close(s->read_msgfds[i]);
> @@ -394,22 +390,25 @@ static void update_disconnected_filename(SocketChardev *s)
> s->is_listen, s->is_telnet);
> }
>
> +/* NB may be called even if tcp_chr_connect has not been
> + * reached, due to TLS or telnet initialization failure,
> + * so can *not* assume s->connected == true
> + */
> static void tcp_chr_disconnect(Chardev *chr)
> {
> SocketChardev *s = SOCKET_CHARDEV(chr);
> -
> - if (!s->connected) {
> - return;
> - }
> + bool emit_close = s->connected;
>
> tcp_chr_free_connection(chr);
>
> - if (s->listen_ioc) {
> + if (s->listen_ioc && s->listen_tag == 0) {
> s->listen_tag = qio_channel_add_watch(
> QIO_CHANNEL(s->listen_ioc), G_IO_IN, tcp_chr_accept, chr, NULL);
> }
> update_disconnected_filename(s);
> - qemu_chr_be_event(chr, CHR_EVENT_CLOSED);
> + if (emit_close) {
> + qemu_chr_be_event(chr, CHR_EVENT_CLOSED);
> + }
> if (s->reconnect_time) {
> qemu_chr_socket_restart_timer(chr);
> }
>
Queued, thanks.
Paolo
© 2016 - 2026 Red Hat, Inc.