Blob Blame Raw
From 0ee28ebbde0313e54b0c8e0f316aa75f97c87169 Mon Sep 17 00:00:00 2001
From: "Daniel P. Berrange" <>
Date: Thu, 8 Feb 2018 17:50:41 +0100
Subject: [PATCH 27/27] io: skip updates to client if websocket output buffer
 is non-zero

RH-Author: Daniel P. Berrange <>
Message-id: <>
Patchwork-id: 78959
O-Subject: [RHEL-7.5 qemu-kvm PATCH v1 27/27] io: skip updates to client if websocket output buffer is non-zero
Bugzilla: 1518711
RH-Acked-by: Laszlo Ersek <>
RH-Acked-by: Gerd Hoffmann <>
RH-Acked-by: Miroslav Rezanina <>

From: "Daniel P. Berrange" <>

When getting a framebuffer update from the guest, we first check
to see if there's still any queued data in the VNC send buffer.
If there is, we skip the update so that we avoid having the send
buffer grow without bound. Unfortunately, the code is only
monitoring the normal send buffer, and not the websockets send
buffer which is separate.

This flaw causes the websockets send buffer to grow without bound
if the other end of the underlying data channel doesn't
read data being sent. This can be seen with VNC if a client
is on a slow WAN link and the guest OS is sending many screen
updates. A malicious VNC client can act like it is on a slow
link by playing a video in the guest and then reading data
very slowly, causing QEMU host memory to expand arbitrarily.

This issue is assigned CVE-2017-15268, publically reported in

Reviewed-by: Eric Blake <>
Signed-off-by: Daniel P. Berrange <>

The corresponding upstream fix is present in commit
a7b20a8efa28e5f22c26c06cd06c2f12bc863493, however, this
patch is a complete re-implementation since the upstream
code for websockets is completely different to that in
QEMU 1.5.3. As such ignore the Reviewed-by tag above.

Signed-off-by: Miroslav Rezanina <>
 ui/vnc-ws.c | 10 ++++++++--
 ui/vnc.c    |  3 +++
 2 files changed, 11 insertions(+), 2 deletions(-)

diff --git a/ui/vnc-ws.c b/ui/vnc-ws.c
index 7133be9..fbfadde 100644
--- a/ui/vnc-ws.c
+++ b/ui/vnc-ws.c
@@ -173,8 +173,13 @@ long vnc_client_write_ws(VncState *vs)
     long ret;
     VNC_DEBUG("Write WS: Pending output %p size %zd offset %zd\n",
               vs->output.buffer, vs->output.capacity, vs->output.offset);
-    vncws_encode_frame(&vs->ws_output, vs->output.buffer, vs->output.offset);
-    buffer_reset(&vs->output);
+    /* We don't consume more from 'output' unless we've finished
+     * sending the previous websockets frame. This ensures that
+     * we still correctly throttle forced framebuffer updates */
+    if (vs->ws_output.offset == 0) {
+        vncws_encode_frame(&vs->ws_output, vs->output.buffer, vs->output.offset);
+        buffer_reset(&vs->output);
+    }
     ret = vnc_client_write_buf(vs, vs->ws_output.buffer, vs->ws_output.offset);
     if (!ret) {
         return 0;
@@ -183,6 +188,7 @@ long vnc_client_write_ws(VncState *vs)
     buffer_advance(&vs->ws_output, ret);
     if (vs->ws_output.offset == 0) {
+        vs->force_update_offset = 0;
         qemu_set_fd_handler2(vs->csock, NULL, vnc_client_read, NULL, vs);
diff --git a/ui/vnc.c b/ui/vnc.c
index 2be87b8..99b1ab1 100644
--- a/ui/vnc.c
+++ b/ui/vnc.c
@@ -915,6 +915,9 @@ static bool vnc_should_update(VncState *vs)
          * is completely idle.
         if (vs->output.offset < vs->throttle_output_offset &&
+            vs->ws_output.offset  < vs->throttle_output_offset &&
             vs->job_update == VNC_STATE_UPDATE_NONE) {
             return true;