[qemu-xen master] ui: place a hard cap on VNC server output buffer size

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

[qemu-xen master] ui: place a hard cap on VNC server output buffer size

patchbot
commit 0c85a40e710d4b71656b28f1f5e1ae5e3780d369
Author:     Daniel P. Berrange <[hidden email]>
AuthorDate: Mon Dec 18 19:12:26 2017 +0000
Commit:     Michael Roth <[hidden email]>
CommitDate: Mon Feb 12 18:34:26 2018 -0600

    ui: place a hard cap on VNC server output buffer size
   
    The previous patches fix problems with throttling of forced framebuffer updates
    and audio data capture that would cause the QEMU output buffer size to grow
    without bound. Those fixes are graceful in that once the client catches up with
    reading data from the server, everything continues operating normally.
   
    There is some data which the server sends to the client that is impractical to
    throttle. Specifically there are various pseudo framebuffer update encodings to
    inform the client of things like desktop resizes, pointer changes, audio
    playback start/stop, LED state and so on. These generally only involve sending
    a very small amount of data to the client, but a malicious guest might be able
    to do things that trigger these changes at a very high rate. Throttling them is
    not practical as missed or delayed events would cause broken behaviour for the
    client.
   
    This patch thus takes a more forceful approach of setting an absolute upper
    bound on the amount of data we permit to be present in the output buffer at
    any time. The previous patch set a threshold for throttling the output buffer
    by allowing an amount of data equivalent to one complete framebuffer update and
    one seconds worth of audio data. On top of this it allowed for one further
    forced framebuffer update to be queued.
   
    To be conservative, we thus take that throttling threshold and multiply it by
    5 to form an absolute upper bound. If this bound is hit during vnc_write() we
    forceably disconnect the client, refusing to queue further data. This limit is
    high enough that it should never be hit unless a malicious client is trying to
    exploit the sever, or the network is completely saturated preventing any sending
    of data on the socket.
   
    This completes the fix for CVE-2017-15124 started in the previous patches.
   
    Signed-off-by: Daniel P. Berrange <[hidden email]>
    Reviewed-by: Darren Kenny <[hidden email]>
    Reviewed-by: Marc-André Lureau <[hidden email]>
    Message-id: [hidden email]
    Signed-off-by: Gerd Hoffmann <[hidden email]>
    (cherry picked from commit f887cf165db20f405cb8805c716bd363aaadf815)
    Signed-off-by: Michael Roth <[hidden email]>
---
 ui/vnc.c | 29 +++++++++++++++++++++++++++++
 1 file changed, 29 insertions(+)

diff --git a/ui/vnc.c b/ui/vnc.c
index 4021c011..a4f0279 100644
--- a/ui/vnc.c
+++ b/ui/vnc.c
@@ -1521,8 +1521,37 @@ gboolean vnc_client_io(QIOChannel *ioc G_GNUC_UNUSED,
 }
 
 
+/*
+ * Scale factor to apply to vs->throttle_output_offset when checking for
+ * hard limit. Worst case normal usage could be x2, if we have a complete
+ * incremental update and complete forced update in the output buffer.
+ * So x3 should be good enough, but we pick x5 to be conservative and thus
+ * (hopefully) never trigger incorrectly.
+ */
+#define VNC_THROTTLE_OUTPUT_LIMIT_SCALE 5
+
 void vnc_write(VncState *vs, const void *data, size_t len)
 {
+    if (vs->disconnecting) {
+        return;
+    }
+    /* Protection against malicious client/guest to prevent our output
+     * buffer growing without bound if client stops reading data. This
+     * should rarely trigger, because we have earlier throttling code
+     * which stops issuing framebuffer updates and drops audio data
+     * if the throttle_output_offset value is exceeded. So we only reach
+     * this higher level if a huge number of pseudo-encodings get
+     * triggered while data can't be sent on the socket.
+     *
+     * NB throttle_output_offset can be zero during early protocol
+     * handshake, or from the job thread's VncState clone
+     */
+    if (vs->throttle_output_offset != 0 &&
+        vs->output.offset > (vs->throttle_output_offset *
+                             VNC_THROTTLE_OUTPUT_LIMIT_SCALE)) {
+        vnc_disconnect_start(vs);
+        return;
+    }
     buffer_reserve(&vs->output, len);
 
     if (vs->ioc != NULL && buffer_empty(&vs->output)) {
--
generated by git-patchbot for /home/xen/git/qemu-xen.git#master


_______________________________________________
Xen-changelog mailing list
[hidden email]
https://lists.xenproject.org/xen-changelog