|
|
5d360b |
From 0f463c3d8c457d6a0bf3b91e48fc4e9162061cf6 Mon Sep 17 00:00:00 2001
|
|
|
5d360b |
From: "Daniel P. Berrange" <berrange@redhat.com>
|
|
|
5d360b |
Date: Thu, 8 Feb 2018 17:50:37 +0100
|
|
|
5d360b |
Subject: [PATCH 23/27] ui: fix VNC client throttling when forced update is
|
|
|
5d360b |
requested
|
|
|
5d360b |
MIME-Version: 1.0
|
|
|
5d360b |
Content-Type: text/plain; charset=UTF-8
|
|
|
5d360b |
Content-Transfer-Encoding: 8bit
|
|
|
5d360b |
|
|
|
5d360b |
RH-Author: Daniel P. Berrange <berrange@redhat.com>
|
|
|
5d360b |
Message-id: <20180208175041.5634-24-berrange@redhat.com>
|
|
|
5d360b |
Patchwork-id: 78949
|
|
|
5d360b |
O-Subject: [RHEL-7.5 qemu-kvm PATCH v1 23/27] ui: fix VNC client throttling when forced update is requested
|
|
|
5d360b |
Bugzilla: 1527405
|
|
|
5d360b |
RH-Acked-by: Laszlo Ersek <lersek@redhat.com>
|
|
|
5d360b |
RH-Acked-by: Gerd Hoffmann <kraxel@redhat.com>
|
|
|
5d360b |
RH-Acked-by: Miroslav Rezanina <mrezanin@redhat.com>
|
|
|
5d360b |
|
|
|
5d360b |
From: "Daniel P. Berrange" <berrange@redhat.com>
|
|
|
5d360b |
|
|
|
5d360b |
The VNC server must throttle data sent to the client to prevent the 'output'
|
|
|
5d360b |
buffer size growing without bound, if the client stops reading data off the
|
|
|
5d360b |
socket (either maliciously or due to stalled/slow network connection).
|
|
|
5d360b |
|
|
|
5d360b |
The current throttling is very crude because it simply checks whether the
|
|
|
5d360b |
output buffer offset is zero. This check is disabled if the client has requested
|
|
|
5d360b |
a forced update, because we want to send these as soon as possible.
|
|
|
5d360b |
|
|
|
5d360b |
As a result, the VNC client can cause QEMU to allocate arbitrary amounts of RAM.
|
|
|
5d360b |
They can first start something in the guest that triggers lots of framebuffer
|
|
|
5d360b |
updates eg play a youtube video. Then repeatedly send full framebuffer update
|
|
|
5d360b |
requests, but never read data back from the server. This can easily make QEMU's
|
|
|
5d360b |
VNC server send buffer consume 100MB of RAM per second, until the OOM killer
|
|
|
5d360b |
starts reaping processes (hopefully the rogue QEMU process, but it might pick
|
|
|
5d360b |
others...).
|
|
|
5d360b |
|
|
|
5d360b |
To address this we make the throttling more intelligent, so we can throttle
|
|
|
5d360b |
full updates. When we get a forced update request, we keep track of exactly how
|
|
|
5d360b |
much data we put on the output buffer. We will not process a subsequent forced
|
|
|
5d360b |
update request until this data has been fully sent on the wire. We always allow
|
|
|
5d360b |
one forced update request to be in flight, regardless of what data is queued
|
|
|
5d360b |
for incremental updates or audio data. The slight complication is that we do
|
|
|
5d360b |
not initially know how much data an update will send, as this is done in the
|
|
|
5d360b |
background by the VNC job thread. So we must track the fact that the job thread
|
|
|
5d360b |
has an update pending, and not process any further updates until this job is
|
|
|
5d360b |
has been completed & put data on the output buffer.
|
|
|
5d360b |
|
|
|
5d360b |
This unbounded memory growth affects all VNC server configurations supported by
|
|
|
5d360b |
QEMU, with no workaround possible. The mitigating factor is that it can only be
|
|
|
5d360b |
triggered by a client that has authenticated with the VNC server, and who is
|
|
|
5d360b |
able to trigger a large quantity of framebuffer updates or audio samples from
|
|
|
5d360b |
the guest OS. Mostly they'll just succeed in getting the OOM killer to kill
|
|
|
5d360b |
their own QEMU process, but its possible other processes can get taken out as
|
|
|
5d360b |
collateral damage.
|
|
|
5d360b |
|
|
|
5d360b |
This is a more general variant of the similar unbounded memory usage flaw in
|
|
|
5d360b |
the websockets server, that was previously assigned CVE-2017-15268, and fixed
|
|
|
5d360b |
in 2.11 by:
|
|
|
5d360b |
|
|
|
5d360b |
commit a7b20a8efa28e5f22c26c06cd06c2f12bc863493
|
|
|
5d360b |
Author: Daniel P. Berrange <berrange@redhat.com>
|
|
|
5d360b |
Date: Mon Oct 9 14:43:42 2017 +0100
|
|
|
5d360b |
|
|
|
5d360b |
io: monitor encoutput buffer size from websocket GSource
|
|
|
5d360b |
|
|
|
5d360b |
This new general memory usage flaw has been assigned CVE-2017-15124, and is
|
|
|
5d360b |
partially fixed by this patch.
|
|
|
5d360b |
|
|
|
5d360b |
RHEL-7 note: context differences in "struct VncState" and
|
|
|
5d360b |
vnc_jobs_consume_buffer() due to downstream lacking (a) commit
|
|
|
5d360b |
fb6ba0d5256c ("qapi event: convert VNC events", 2014-06-23), part of
|
|
|
5d360b |
v2.1.0, and (b) commit 04d2529da27d ("ui: convert VNC server to use
|
|
|
5d360b |
QIOChannelSocket", 2015-12-18), part of v2.6.0.
|
|
|
5d360b |
|
|
|
5d360b |
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
|
|
|
5d360b |
Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
|
|
|
5d360b |
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
|
|
|
5d360b |
Message-id: 20171218191228.31018-11-berrange@redhat.com
|
|
|
5d360b |
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
|
|
|
5d360b |
(cherry picked from commit ada8d2e4369ea49677d8672ac81bce73eefd5b54)
|
|
|
5d360b |
Signed-off-by: Miroslav Rezanina <mrezanin@redhat.com>
|
|
|
5d360b |
---
|
|
|
5d360b |
ui/vnc-auth-sasl.c | 5 +++++
|
|
|
5d360b |
ui/vnc-jobs.c | 5 +++++
|
|
|
5d360b |
ui/vnc.c | 28 ++++++++++++++++++++++++----
|
|
|
5d360b |
ui/vnc.h | 7 +++++++
|
|
|
5d360b |
4 files changed, 41 insertions(+), 4 deletions(-)
|
|
|
5d360b |
|
|
|
5d360b |
diff --git a/ui/vnc-auth-sasl.c b/ui/vnc-auth-sasl.c
|
|
|
5d360b |
index 804b8e7..8188081 100644
|
|
|
5d360b |
--- a/ui/vnc-auth-sasl.c
|
|
|
5d360b |
+++ b/ui/vnc-auth-sasl.c
|
|
|
5d360b |
@@ -76,6 +76,11 @@ long vnc_client_write_sasl(VncState *vs)
|
|
|
5d360b |
|
|
|
5d360b |
vs->sasl.encodedOffset += ret;
|
|
|
5d360b |
if (vs->sasl.encodedOffset == vs->sasl.encodedLength) {
|
|
|
5d360b |
+ if (vs->sasl.encodedRawLength >= vs->force_update_offset) {
|
|
|
5d360b |
+ vs->force_update_offset = 0;
|
|
|
5d360b |
+ } else {
|
|
|
5d360b |
+ vs->force_update_offset -= vs->sasl.encodedRawLength;
|
|
|
5d360b |
+ }
|
|
|
5d360b |
vs->output.offset -= vs->sasl.encodedRawLength;
|
|
|
5d360b |
vs->sasl.encoded = NULL;
|
|
|
5d360b |
vs->sasl.encodedOffset = vs->sasl.encodedLength = 0;
|
|
|
5d360b |
diff --git a/ui/vnc-jobs.c b/ui/vnc-jobs.c
|
|
|
5d360b |
index e553bd9..9705899 100644
|
|
|
5d360b |
--- a/ui/vnc-jobs.c
|
|
|
5d360b |
+++ b/ui/vnc-jobs.c
|
|
|
5d360b |
@@ -170,6 +170,11 @@ void vnc_jobs_consume_buffer(VncState *vs)
|
|
|
5d360b |
vnc_client_write, vs);
|
|
|
5d360b |
}
|
|
|
5d360b |
buffer_move(&vs->output, &vs->jobs_buffer);
|
|
|
5d360b |
+
|
|
|
5d360b |
+ if (vs->job_update == VNC_STATE_UPDATE_FORCE) {
|
|
|
5d360b |
+ vs->force_update_offset = vs->output.offset;
|
|
|
5d360b |
+ }
|
|
|
5d360b |
+ vs->job_update = VNC_STATE_UPDATE_NONE;
|
|
|
5d360b |
}
|
|
|
5d360b |
flush = vs->csock != -1 && vs->abort != true;
|
|
|
5d360b |
vnc_unlock_output(vs);
|
|
|
5d360b |
diff --git a/ui/vnc.c b/ui/vnc.c
|
|
|
5d360b |
index 952a051..96b6caf 100644
|
|
|
5d360b |
--- a/ui/vnc.c
|
|
|
5d360b |
+++ b/ui/vnc.c
|
|
|
5d360b |
@@ -906,14 +906,28 @@ static bool vnc_should_update(VncState *vs)
|
|
|
5d360b |
break;
|
|
|
5d360b |
case VNC_STATE_UPDATE_INCREMENTAL:
|
|
|
5d360b |
/* Only allow incremental updates if the pending send queue
|
|
|
5d360b |
- * is less than the permitted threshold
|
|
|
5d360b |
+ * is less than the permitted threshold, and the job worker
|
|
|
5d360b |
+ * is completely idle.
|
|
|
5d360b |
*/
|
|
|
5d360b |
- if (vs->output.offset < vs->throttle_output_offset) {
|
|
|
5d360b |
+ if (vs->output.offset < vs->throttle_output_offset &&
|
|
|
5d360b |
+ vs->job_update == VNC_STATE_UPDATE_NONE) {
|
|
|
5d360b |
return true;
|
|
|
5d360b |
}
|
|
|
5d360b |
break;
|
|
|
5d360b |
case VNC_STATE_UPDATE_FORCE:
|
|
|
5d360b |
- return true;
|
|
|
5d360b |
+ /* Only allow forced updates if the pending send queue
|
|
|
5d360b |
+ * does not contain a previous forced update, and the
|
|
|
5d360b |
+ * job worker is completely idle.
|
|
|
5d360b |
+ *
|
|
|
5d360b |
+ * Note this means we'll queue a forced update, even if
|
|
|
5d360b |
+ * the output buffer size is otherwise over the throttle
|
|
|
5d360b |
+ * output limit.
|
|
|
5d360b |
+ */
|
|
|
5d360b |
+ if (vs->force_update_offset == 0 &&
|
|
|
5d360b |
+ vs->job_update == VNC_STATE_UPDATE_NONE) {
|
|
|
5d360b |
+ return true;
|
|
|
5d360b |
+ }
|
|
|
5d360b |
+ break;
|
|
|
5d360b |
}
|
|
|
5d360b |
return false;
|
|
|
5d360b |
}
|
|
|
5d360b |
@@ -975,8 +989,9 @@ static int vnc_update_client(VncState *vs, int has_dirty)
|
|
|
5d360b |
}
|
|
|
5d360b |
}
|
|
|
5d360b |
|
|
|
5d360b |
- vnc_job_push(job);
|
|
|
5d360b |
+ vs->job_update = vs->update;
|
|
|
5d360b |
vs->update = VNC_STATE_UPDATE_NONE;
|
|
|
5d360b |
+ vnc_job_push(job);
|
|
|
5d360b |
vs->has_dirty = 0;
|
|
|
5d360b |
return n;
|
|
|
5d360b |
}
|
|
|
5d360b |
@@ -1241,6 +1256,11 @@ static long vnc_client_write_plain(VncState *vs)
|
|
|
5d360b |
if (!ret)
|
|
|
5d360b |
return 0;
|
|
|
5d360b |
|
|
|
5d360b |
+ if (ret >= vs->force_update_offset) {
|
|
|
5d360b |
+ vs->force_update_offset = 0;
|
|
|
5d360b |
+ } else {
|
|
|
5d360b |
+ vs->force_update_offset -= ret;
|
|
|
5d360b |
+ }
|
|
|
5d360b |
buffer_advance(&vs->output, ret);
|
|
|
5d360b |
|
|
|
5d360b |
if (vs->output.offset == 0) {
|
|
|
5d360b |
diff --git a/ui/vnc.h b/ui/vnc.h
|
|
|
5d360b |
index d7eede3..70316ba 100644
|
|
|
5d360b |
--- a/ui/vnc.h
|
|
|
5d360b |
+++ b/ui/vnc.h
|
|
|
5d360b |
@@ -268,6 +268,7 @@ struct VncState
|
|
|
5d360b |
|
|
|
5d360b |
VncDisplay *vd;
|
|
|
5d360b |
VncStateUpdate update; /* Most recent pending request from client */
|
|
|
5d360b |
+ VncStateUpdate job_update; /* Currently processed by job thread */
|
|
|
5d360b |
int has_dirty;
|
|
|
5d360b |
uint32_t features;
|
|
|
5d360b |
int absolute;
|
|
|
5d360b |
@@ -301,6 +302,12 @@ struct VncState
|
|
|
5d360b |
|
|
|
5d360b |
QObject *info;
|
|
|
5d360b |
|
|
|
5d360b |
+ /* Job thread bottom half has put data for a forced update
|
|
|
5d360b |
+ * into the output buffer. This offset points to the end of
|
|
|
5d360b |
+ * the update data in the output buffer. This lets us determine
|
|
|
5d360b |
+ * when a force update is fully sent to the client, allowing
|
|
|
5d360b |
+ * us to process further forced updates. */
|
|
|
5d360b |
+ size_t force_update_offset;
|
|
|
5d360b |
/* We allow multiple incremental updates or audio capture
|
|
|
5d360b |
* samples to be queued in output buffer, provided the
|
|
|
5d360b |
* buffer size doesn't exceed this threshold. The value
|
|
|
5d360b |
--
|
|
|
5d360b |
1.8.3.1
|
|
|
5d360b |
|