On 11/14/25 22:12, Michael Tokarev wrote:
On 10/3/25 18:02, Daniel P. Berrangé wrote:
While code is supposed to call qio_channel_close() before releasing the
last reference on an QIOChannel, this is not guaranteed. QIOChannelFile
and QIOChannelSocket both cleanup resources in their finalizer if the
close operation was missed.
This ensures the TLS channel will do the same failsafe cleanup.
Signed-off-by: Daniel P. Berrangé <[email protected]>
---
io/channel-tls.c | 10 ++++++++++
1 file changed, 10 insertions(+)
diff --git a/io/channel-tls.c b/io/channel-tls.c
index 7135896f79..bb460ca7e9 100644
--- a/io/channel-tls.c
+++ b/io/channel-tls.c
@@ -342,6 +342,16 @@ static void qio_channel_tls_finalize(Object *obj)
{
QIOChannelTLS *ioc = QIO_CHANNEL_TLS(obj);
+ if (ioc->hs_ioc_tag) {
+ trace_qio_channel_tls_handshake_cancel(ioc);
+ g_clear_handle_id(&ioc->hs_ioc_tag, g_source_remove);
+ }
+
+ if (ioc->bye_ioc_tag) {
+ trace_qio_channel_tls_bye_cancel(ioc);
+ g_clear_handle_id(&ioc->bye_ioc_tag, g_source_remove);
+ }
+
object_unref(OBJECT(ioc->master));
qcrypto_tls_session_free(ioc->session);
}
This series should be relevant for 7.2.x qemu stable series too, it
looks like. Except that 7.2 does not have ioc->bye_ioc_tag (which
comes from v9.2.0-1773-g30ee88622e "io: tls: Add qio_channel_tls_bye"),
and does not have trace_qio_channel_tls_handshake_cancel defined.
https://gitlab.com/mjt0k/qemu/-/commits/cebdbd038e44af56e74272924dc2bf595a51fd8f
is what I did (4 commits from there). I picked up a trivial change
v8.2.0-1149-g003f15369d "io: add trace event when cancelling TLS
handshake", and removed half of v10.1.0-1703-g2c147611cf56
"io: release active GSource in TLS channel finalizer" (this one)
which releases bye_ioc_tag.
The other (subsequent) two commits from this series are taken as-is.
Does the whole series (besides this change) look sane in the 7.2.x
context?
Thanks,
/mjt
