The "XVP" (Xen VNC Proxy) extension defines a mechanism for a VNC client
to issue power control requests to trigger graceful shutdown, reboot, or
hard reset.
This option is not enabled by default, since we cannot assume that users
with VNC access implicitly have administrator access to the guest OS.
Thus is it enabled with a boolean "power-control" option e.g.
-vnc :1,power-control=on
While, QEMU can easily support shutdown and reset, there's no easy way
to wire up reboot support at this time. In theory it could be done by
issuing a shutdown, followed by a reset, but there's no convenient
wiring for such a pairing in QEMU. It also isn't possible to have the
VNC server directly talk to QEMU guest agent, since the agent chardev is
typically owned by an external mgmt app.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
[ kraxel: rebase to master ]
[ kraxel: add missing break ]
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
There is a new vnc extension for cursors with an alpha channel. Use
it if supported by the vnc client, prefer it over the "rich cursor"
extension which supports only a bitmask for transparency.
This is a visible improvement especially on modern desktops which
actually use the alpha channel when defining cursors.
https://github.com/rfbproto/rfbproto/blob/master/rfbproto.rst#cursor-with-alpha-pseudo-encoding
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Tested-by: Daniel P. Berrangé <berrange@redhat.com>
Message-id: 20201208115737.18581-7-kraxel@redhat.com
vnc stopped using the copyrect pseudo encoding in 2017, in commit
50628d3479 ("cirrus/vnc: zap bitblit support from console code.")
So we can drop the now unused copyrect feature bit.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-id: 20201208115737.18581-5-kraxel@redhat.com
Use an enum for the vnc feature bits. That way they are enumerated
automatically and we don't have to do that manually when adding or
removing features.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-id: 20201208115737.18581-4-kraxel@redhat.com
We don't need Error **, as all callers pass local Error object, which
isn't used after the call, or NULL. Use Error * instead.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20191205174635.18758-6-vsementsov@virtuozzo.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Currently when qemu receives a vnc connect, it creates a 'VncState' to
represent this connection. In 'vnc_worker_thread_loop' it creates a
local 'VncState'. The connection 'VcnState' and local 'VncState' exchange
data in 'vnc_async_encoding_start' and 'vnc_async_encoding_end'.
In 'zrle_compress_data' it calls 'deflateInit2' to allocate the libz library
opaque data. The 'VncState' used in 'zrle_compress_data' is the local
'VncState'. In 'vnc_zrle_clear' it calls 'deflateEnd' to free the libz
library opaque data. The 'VncState' used in 'vnc_zrle_clear' is the connection
'VncState'. In currently implementation there will be a memory leak when the
vnc disconnect. Following is the asan output backtrack:
Direct leak of 29760 byte(s) in 5 object(s) allocated from:
0 0xffffa67ef3c3 in __interceptor_calloc (/lib64/libasan.so.4+0xd33c3)
1 0xffffa65071cb in g_malloc0 (/lib64/libglib-2.0.so.0+0x571cb)
2 0xffffa5e968f7 in deflateInit2_ (/lib64/libz.so.1+0x78f7)
3 0xaaaacec58613 in zrle_compress_data ui/vnc-enc-zrle.c:87
4 0xaaaacec58613 in zrle_send_framebuffer_update ui/vnc-enc-zrle.c:344
5 0xaaaacec34e77 in vnc_send_framebuffer_update ui/vnc.c:919
6 0xaaaacec5e023 in vnc_worker_thread_loop ui/vnc-jobs.c:271
7 0xaaaacec5e5e7 in vnc_worker_thread ui/vnc-jobs.c:340
8 0xaaaacee4d3c3 in qemu_thread_start util/qemu-thread-posix.c:502
9 0xffffa544e8bb in start_thread (/lib64/libpthread.so.0+0x78bb)
10 0xffffa53965cb in thread_start (/lib64/libc.so.6+0xd55cb)
This is because the opaque allocated in 'deflateInit2' is not freed in
'deflateEnd'. The reason is that the 'deflateEnd' calls 'deflateStateCheck'
and in the latter will check whether 's->strm != strm'(libz's data structure).
This check will be true so in 'deflateEnd' it just return 'Z_STREAM_ERROR' and
not free the data allocated in 'deflateInit2'.
The reason this happens is that the 'VncState' contains the whole 'VncZrle',
so when calling 'deflateInit2', the 's->strm' will be the local address.
So 's->strm != strm' will be true.
To fix this issue, we need to make 'zrle' of 'VncState' to be a pointer.
Then the connection 'VncState' and local 'VncState' exchange mechanism will
work as expection. The 'tight' of 'VncState' has the same issue, let's also turn
it to a pointer.
Reported-by: Ying Fang <fangying1@huawei.com>
Signed-off-by: Li Qiang <liq3ea@163.com>
Message-id: 20190831153922.121308-1-liq3ea@163.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Some of the generated qapi-types-MODULE.h are included all over the
place. Changing a QAPI type can trigger massive recompiling. Top
scorers recompile more than 1000 out of some 6600 objects (not
counting tests and objects that don't depend on qemu/osdep.h):
6300 qapi/qapi-builtin-types.h
5700 qapi/qapi-types-run-state.h
3900 qapi/qapi-types-common.h
3300 qapi/qapi-types-sockets.h
3000 qapi/qapi-types-misc.h
3000 qapi/qapi-types-crypto.h
3000 qapi/qapi-types-job.h
3000 qapi/qapi-types-block-core.h
2800 qapi/qapi-types-block.h
1300 qapi/qapi-types-net.h
Clean up headers to include generated QAPI headers only where needed.
Impact is negligible except for hw/qdev-properties.h.
This header includes qapi/qapi-types-block.h and
qapi/qapi-types-misc.h. They are used only in expansions of property
definition macros such as DEFINE_PROP_BLOCKDEV_ON_ERROR() and
DEFINE_PROP_OFF_AUTO(). Moving their inclusion from
hw/qdev-properties.h to the users of these macros avoids pointless
recompiles. This is how other property definition macros, such as
DEFINE_PROP_NETDEV(), already work.
Improves things for some of the top scorers:
3600 qapi/qapi-types-common.h
2800 qapi/qapi-types-sockets.h
900 qapi/qapi-types-misc.h
2200 qapi/qapi-types-crypto.h
2100 qapi/qapi-types-job.h
2100 qapi/qapi-types-block-core.h
270 qapi/qapi-types-block.h
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190812052359.30071-3-armbru@redhat.com>
No header includes qemu-common.h after this commit, as prescribed by
qemu-common.h's file comment.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20190523143508.25387-5-armbru@redhat.com>
[Rebased with conflicts resolved automatically, except for
include/hw/arm/xlnx-zynqmp.h hw/arm/nrf51_soc.c hw/arm/msf2-soc.c
block/qcow2-refcount.c block/qcow2-cluster.c block/qcow2-cache.c
target/arm/cpu.h target/lm32/cpu.h target/m68k/cpu.h target/mips/cpu.h
target/moxie/cpu.h target/nios2/cpu.h target/openrisc/cpu.h
target/riscv/cpu.h target/tilegx/cpu.h target/tricore/cpu.h
target/unicore32/cpu.h target/xtensa/cpu.h; bsd-user/main.c and
net/tap-bsd.c fixed up]
The 'qemu_acl' type was a previous non-QOM based attempt to provide an
authorization facility in QEMU. Because it is non-QOM based it cannot be
created via the command line and requires special monitor commands to
manipulate it.
The new QAuthZ subclasses provide a superset of the functionality in
qemu_acl, so the latter can now be deleted. The HMP 'acl_*' monitor
commands are converted to use the new QAuthZSimple data type instead
in order to provide temporary backwards compatibility.
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Use the new keyboard state tracked for vnc. Allows to drop the
vnc-specific modifier state tracking code.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-id: 20190122092814.14919-7-kraxel@redhat.com
Add #if defined(CONFIG_VNC) in generated code, and adjust the
qmp/hmp code accordingly.
query-qmp-schema no longer reports the command/events etc as
available when disabled at compile.
Commands made conditional:
* query-vnc, query-vnc-servers, change-vnc-password
Before the patch, the commands for !CONFIG_VNC are stubs that fail
like this:
{"error": {"class": "GenericError",
"desc": "The feature 'vnc' is not enabled"}}
Afterwards, they fail like this:
{"error": {"class": "CommandNotFound",
"desc": "The command FOO has not been found"}}
I call that an improvement, because it lets clients distinguish
between command unavailable (class CommandNotFound) and command failed
(class GenericError).
Events made conditional:
* VNC_CONNECTED, VNC_INITIALIZED, VNC_DISCONNECTED
HMP change:
* info vnc
Will return "unknown command: 'info vnc'" when VNC is compiled
out (same as error for spice when --disable-spice)
Occurrences of VNC (case insensitive) in the schema that aren't
covered by this change:
* add_client
Command has other uses, including "socket bases character devices".
These are unconditional as far as I can tell.
* set_password, expire_password
In theory, these commands could be used for managing any service's
password. In practice, they're used for VNC and SPICE services.
They're documented for "remote display session" / "remote display
server".
The service is selected by argument @protocol. The code special-cases
protocol-specific argument checking, then calls a protocol-specific
function to do the work. If it fails, the command fails with "Could
not set password". It does when the service isn't compiled in (it's a
stub then).
We could make these commands conditional on the conjunction of all
services [currently: defined(CONFIG_VNC) || defined(CONFIG_SPICE)],
but I doubt it's worthwhile.
* change
Command has other uses, namely changing media.
This patch inlines a stub; no functional change.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20180703155648.11933-14-marcandre.lureau@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Set magic cookie on initialization. Clear on cleanup. Sprinkle a bunch
of assert()s checking the cookie, to verify the pointer is valid.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-id: 20180507102254.12107-1-kraxel@redhat.com
In my "build everything" tree, a change to the types in
qapi-schema.json triggers a recompile of about 4800 out of 5100
objects.
The previous commit split up qmp-commands.h, qmp-event.h, qmp-visit.h,
qapi-types.h. Each of these headers still includes all its shards.
Reduce compile time by including just the shards we actually need.
To illustrate the benefits: adding a type to qapi/migration.json now
recompiles some 2300 instead of 4800 objects. The next commit will
improve it further.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20180211093607.27351-24-armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
[eblake: rebase to master]
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20180201111846.21846-4-armbru@redhat.com>
The VNC server already has the ability to listen on multiple sockets.
Converting it to use the QIONetListener APIs though, will reduce the
amount of code in the VNC server and improve the clarity of what is
left.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Message-id: 20180201164514.10330-1-berrange@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Pixman returns a signed int for the image width/height, but the VNC
protocol only permits a unsigned int16. Effective framebuffer size
is determined by the guest, limited by the video RAM size, so the
dimensions are unlikely to exceed the range of an unsigned int16,
but this is not currently validated.
With the current use of 'int' for client width/height, the calculation
of offsets in vnc_update_throttle_offset() suffers from integer size
promotion and sign extension, causing coverity warnings
*** CID 1385147: Integer handling issues (SIGN_EXTENSION)
/ui/vnc.c: 979 in vnc_update_throttle_offset()
973 * than that the client would already suffering awful audio
974 * glitches, so dropping samples is no worse really).
975 */
976 static void vnc_update_throttle_offset(VncState *vs)
977 {
978 size_t offset =
>>> CID 1385147: Integer handling issues (SIGN_EXTENSION)
>>> Suspicious implicit sign extension:
"vs->client_pf.bytes_per_pixel" with type "unsigned char" (8 bits,
unsigned) is promoted in "vs->client_width * vs->client_height *
vs->client_pf.bytes_per_pixel" to type "int" (32 bits, signed), then
sign-extended to type "unsigned long" (64 bits, unsigned). If
"vs->client_width * vs->client_height * vs->client_pf.bytes_per_pixel"
is greater than 0x7FFFFFFF, the upper bits of the result will all be 1.
979 vs->client_width * vs->client_height * vs->client_pf.bytes_per_pixel;
Change client_width / client_height to be a size_t to avoid sign
extension and integer promotion. Then validate that dimensions are in
range wrt the RFB protocol u16 limits.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Message-id: 20180118155254.17053-1-berrange@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
While the QIOChannel APIs for reading/writing data return ssize_t, with negative
value indicating an error, the VNC code passes this return value through the
vnc_client_io_error() method. This detects the error condition, disconnects the
client and returns 0 to indicate error. Thus all the VNC helper methods should
return size_t (unsigned), and misleading comments which refer to the possibility
of negative return values need fixing.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-id: 20171218191228.31018-14-berrange@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The VNC server must throttle data sent to the client to prevent the 'output'
buffer size growing without bound, if the client stops reading data off the
socket (either maliciously or due to stalled/slow network connection).
The current throttling is very crude because it simply checks whether the
output buffer offset is zero. This check is disabled if the client has requested
a forced update, because we want to send these as soon as possible.
As a result, the VNC client can cause QEMU to allocate arbitrary amounts of RAM.
They can first start something in the guest that triggers lots of framebuffer
updates eg play a youtube video. Then repeatedly send full framebuffer update
requests, but never read data back from the server. This can easily make QEMU's
VNC server send buffer consume 100MB of RAM per second, until the OOM killer
starts reaping processes (hopefully the rogue QEMU process, but it might pick
others...).
To address this we make the throttling more intelligent, so we can throttle
full updates. When we get a forced update request, we keep track of exactly how
much data we put on the output buffer. We will not process a subsequent forced
update request until this data has been fully sent on the wire. We always allow
one forced update request to be in flight, regardless of what data is queued
for incremental updates or audio data. The slight complication is that we do
not initially know how much data an update will send, as this is done in the
background by the VNC job thread. So we must track the fact that the job thread
has an update pending, and not process any further updates until this job is
has been completed & put data on the output buffer.
This unbounded memory growth affects all VNC server configurations supported by
QEMU, with no workaround possible. The mitigating factor is that it can only be
triggered by a client that has authenticated with the VNC server, and who is
able to trigger a large quantity of framebuffer updates or audio samples from
the guest OS. Mostly they'll just succeed in getting the OOM killer to kill
their own QEMU process, but its possible other processes can get taken out as
collateral damage.
This is a more general variant of the similar unbounded memory usage flaw in
the websockets server, that was previously assigned CVE-2017-15268, and fixed
in 2.11 by:
commit a7b20a8efa
Author: Daniel P. Berrange <berrange@redhat.com>
Date: Mon Oct 9 14:43:42 2017 +0100
io: monitor encoutput buffer size from websocket GSource
This new general memory usage flaw has been assigned CVE-2017-15124, and is
partially fixed by this patch.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-id: 20171218191228.31018-11-berrange@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The VNC server must throttle data sent to the client to prevent the 'output'
buffer size growing without bound, if the client stops reading data off the
socket (either maliciously or due to stalled/slow network connection).
The current throttling is very crude because it simply checks whether the
output buffer offset is zero. This check must be disabled if audio capture is
enabled, because when streaming audio the output buffer offset will rarely be
zero due to queued audio data, and so this would starve framebuffer updates.
As a result, the VNC client can cause QEMU to allocate arbitrary amounts of RAM.
They can first start something in the guest that triggers lots of framebuffer
updates eg play a youtube video. Then enable audio capture, and simply never
read data back from the server. This can easily make QEMU's VNC server send
buffer consume 100MB of RAM per second, until the OOM killer starts reaping
processes (hopefully the rogue QEMU process, but it might pick others...).
To address this we make the throttling more intelligent, so we can throttle
when audio capture is active too. To determine how to throttle incremental
updates or audio data, we calculate a size threshold. Normally the threshold is
the approximate number of bytes associated with a single complete framebuffer
update. ie width * height * bytes per pixel. We'll send incremental updates
until we hit this threshold, at which point we'll stop sending updates until
data has been written to the wire, causing the output buffer offset to fall
back below the threshold.
If audio capture is enabled, we increase the size of the threshold to also
allow for upto 1 seconds worth of audio data samples. ie nchannels * bytes
per sample * frequency. This allows the output buffer to have a mixture of
incremental framebuffer updates and audio data queued, but once the threshold
is exceeded, audio data will be dropped and incremental updates will be
throttled.
This unbounded memory growth affects all VNC server configurations supported by
QEMU, with no workaround possible. The mitigating factor is that it can only be
triggered by a client that has authenticated with the VNC server, and who is
able to trigger a large quantity of framebuffer updates or audio samples from
the guest OS. Mostly they'll just succeed in getting the OOM killer to kill
their own QEMU process, but its possible other processes can get taken out as
collateral damage.
This is a more general variant of the similar unbounded memory usage flaw in
the websockets server, that was previously assigned CVE-2017-15268, and fixed
in 2.11 by:
commit a7b20a8efa
Author: Daniel P. Berrange <berrange@redhat.com>
Date: Mon Oct 9 14:43:42 2017 +0100
io: monitor encoutput buffer size from websocket GSource
This new general memory usage flaw has been assigned CVE-2017-15124, and is
partially fixed by this patch.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-id: 20171218191228.31018-10-berrange@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Currently the VNC servers tracks whether a client has requested an incremental
or forced update with two boolean flags. There are only really 3 distinct
states to track, so create an enum to more accurately reflect permitted states.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-id: 20171218191228.31018-7-berrange@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Currently there is only a single listener for plain VNC and
a single listener for websockets VNC. This means that if
getaddrinfo() returns multiple IP addresses, for a hostname,
the VNC server can only listen on one of them. This is
just bearable if listening on wildcard interface, or if
the host only has a single network interface to listen on,
but if there are multiple NICs and the VNC server needs
to listen on 2 or more specific IP addresses, it can't be
done.
This refactors the VncDisplay state so that it holds an
array of listening sockets, but still only listens on
one socket.
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Message-id: 20170203120649.15637-4-berrange@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Rename the vnc_init_state method to reflect what its actual
purpose is, to discourage future devs from using it for more
general state initialization.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Message-id: 1475163940-26094-10-git-send-email-berrange@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Most of the fields in VncState are initialized in the
vnc_connect() method, but some are done in vnc_init_state()
instead.
The purpose of having vnc_init_state() is to delay starting
of the VNC wire protocol until after the websockets handshake
has completed. As such the vnc_init_state() method only needs
to be used for initialization that is dependant on the wire
protocol running.
This also lets us get rid of the initialized boolean flag
from the VncState struct.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Message-id: 1475163940-26094-9-git-send-email-berrange@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
There is a lot of repeated code in the auth type setup method,
particularly around checking TLS credential types. Refactor
it to reduce duplication and instead of having one method
do both plain and websockets at once, call it separately
for each.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Message-id: 1475163940-26094-6-git-send-email-berrange@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The 'ws_tls' field in VncState is only ever representing
the result of 'tlscreds != NULL' and is thus pointless.
Replace use of 'ws_tls' with a direct check against
'tlscreds'
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Message-id: 1475163940-26094-4-git-send-email-berrange@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The 'ws_enabled' field is never used outside of the
vnc_display_open method, so can be a local variable.
The 'enabled' field is easily replaced by a check
for whether 'lsock' is non-NULL.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Message-id: 1475163940-26094-3-git-send-email-berrange@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Limits the rate kbd events from the vnc server are forwarded to the
guest, so input devices which are typically low-bandwidth can keep
up even on bulky input.
v2: update documentation too.
v3: spell fixes.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Tested-by: Yang Hongyang <hongyang.yang@easystack.cn>
Message-id: 1464762150-25817-1-git-send-email-kraxel@redhat.com
Clean up includes so that osdep.h is included first and headers
which it implies are not included manually.
This commit was created with scripts/clean-includes.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Eric Blake <eblake@redhat.com>
Remove custom websock handling code from the VNC server and use
the QIOChannelWebsock class instead.
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Switch VNC server over to using the QIOChannelTLS object for
the TLS session. This removes all remaining VNC specific code
for dealing with TLS handshakes.
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The minimal first step conversion to use QIOChannelSocket
classes instead of directly using POSIX sockets API. This
will later be extended to also cover the TLS, SASL and
websockets code.
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The Buffer code in the VNC server is useful for the IO channel
code, so pull it out into a shared module, QIOBuffer.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Switch VNC server over to using the QCryptoTLSSession object
for the TLS session. This removes the direct use of gnutls
from the VNC server code. It also removes most knowledge
about TLS certificate handling from the VNC server code.
This has the nice effect that all the CONFIG_VNC_TLS
conditionals go away and the user gets an actual error
message when requesting TLS instead of it being silently
ignored.
With this change, the existing configuration options for
enabling TLS with -vnc are deprecated.
Old syntax for anon-DH credentials:
-vnc hostname:0,tls
New syntax:
-object tls-creds-anon,id=tls0,endpoint=server \
-vnc hostname:0,tls-creds=tls0
Old syntax for x509 credentials, no client certs:
-vnc hostname:0,tls,x509=/path/to/certs
New syntax:
-object tls-creds-x509,id=tls0,dir=/path/to/certs,endpoint=server,verify-peer=no \
-vnc hostname:0,tls-creds=tls0
Old syntax for x509 credentials, requiring client certs:
-vnc hostname:0,tls,x509verify=/path/to/certs
New syntax:
-object tls-creds-x509,id=tls0,dir=/path/to/certs,endpoint=server,verify-peer=yes \
-vnc hostname:0,tls-creds=tls0
This aligns VNC with the way TLS credentials are to be
configured in the future for chardev, nbd and migration
backends. It also has the benefit that the same TLS
credentials can be shared across multiple VNC server
instances, if desired.
If someone uses the deprecated syntax, it will internally
result in the creation of a 'tls-creds' object with an ID
based on the VNC server ID. This allows backwards compat
with the CLI syntax, while still deleting all the original
TLS code from the VNC server.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Various VNC server I/O functions return 'long' and then
also pass this to a method accepting 'int'. All these
should be ssize_t to match the signature of read/write
APIs and thus avoid potential for integer truncation /
wraparound.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Remove the direct use of gnutls for hash processing in the
websockets code, in favour of using the crypto APIs. This
allows the websockets code to be built unconditionally
removing countless conditional checks from the VNC code.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Message-Id: <1435770638-25715-9-git-send-email-berrange@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The logic for decoding websocket frames wants to fully
decode the frame header and payload, before allowing the
VNC server to see any of the payload data. There is no
size limit on websocket payloads, so this allows a
malicious network client to consume 2^64 bytes in memory
in QEMU. It can trigger this denial of service before
the VNC server even performs any authentication.
The fix is to decode the header, and then incrementally
decode the payload data as it is needed. With this fix
the websocket decoder will allow at most 4k of data to
be buffered before decoding and processing payload.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
[ kraxel: fix frequent spurious disconnects, suggested by Peter Maydell ]
@@ -361,7 +361,7 @@ int vncws_decode_frame_payload(Buffer *input,
- *payload_size = input->offset;
+ *payload_size = *payload_remain;
[ kraxel: fix 32bit build ]
@@ -306,7 +306,7 @@ struct VncState
- uint64_t ws_payload_remain;
+ size_t ws_payload_remain;
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The previous change to the auth scheme handling guarantees we
can never have nested TLS sessions in the VNC websockets server.
Thus we can remove the separate gnutls_session instance.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The way the websockets TLS code was integrated into the VNC server
made it essentially useless. The only time that the websockets TLS
support could be used is if the primary VNC server had its existing
TLS support disabled. ie QEMU had to be launched with:
# qemu -vnc localhost:1,websockets=5902,x509=/path/to/certs
Note the absence of the 'tls' flag. This is already a bug, because
the docs indicate that 'x509' is ignored unless 'tls' is given.
If the primary VNC server had TLS turned on via the 'tls' flag,
then this prevented the websockets TLS support from being used,
because it activates the VeNCrypt auth which would have resulted
in TLS being run over a TLS session. Of course no websockets VNC
client supported VeNCrypt so in practice, since the browser clients
cannot setup a nested TLS session over the main HTTPS connection,
so it would not even get past auth.
This patch causes us to decide our auth scheme separately for the
main VNC server vs the websockets VNC server. We take account of
the fact that if TLS is enabled, then the websockets client will
use https, so setting up VeNCrypt is thus redundant as it would
lead to nested TLS sessions.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
If the VNC server is built without tls, sasl or websocket support
and the user requests one of these features, they are just silently
ignored. This is bad because it means the VNC server ends up running
in a configuration that is less secure than the user asked for.
It also leads to an tangled mass of preprocessor conditionals when
configuring the VNC server.
This ensures that the tls, sasl & websocket options are always
processed and an error is reported back to the user if any of
them were disabled at build time.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Nobody cares about those strings, they are only used to check whenever
the vnc server / websocket support is enabled or not. Add bools for
this and drop the strings.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Gonglei <arei.gonglei@huawei.com>
Also track the number of connections in "connecting" and "shared" state
(in addition to the "exclusive" state). Apply a configurable limit to
these connections.
The logic to apply the limit to connections in "shared" state is pretty
simple: When the limit is reached no new connections are allowed.
The logic to apply the limit to connections in "connecting" state (this
is the state you are in *before* successful authentication) is
slightly different: A new connect kicks out the oldest client which is
still in "connecting" state. This avoids a easy DoS by unauthenticated
users by simply opening connections until the limit is reached.
Cc: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Replace with a vnc_displays list, so we can have multiple vnc server
instances. Add vnc_server_find function to lookup a display by id.
With no id supplied return the first vnc server, for backward
compatibility reasons.
It is not possible (yet) to actually create multiple vnc server
instances.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Gonglei <arei.gonglei@huawei.com>
We need to remember has_updates for each vnc client. Otherwise it might
happen that vnc_update_client(has_dirty=1) takes the first exit due to
output buffers not being flushed yet and subsequent calls with
has_dirty=0 take the second exit, wrongly assuming there is nothing to
do because the work defered in the first call is ignored.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Peter Lieven <pl@kamp.de>
this patch makes the VNC server work correctly if the
server surface and the guest surface have different sizes.
Basically the server surface is adjusted to not exceed VNC_MAX_WIDTH
x VNC_MAX_HEIGHT and additionally the width is rounded up to multiple of
VNC_DIRTY_PIXELS_PER_BIT.
If we have a resolution whose width is not dividable by VNC_DIRTY_PIXELS_PER_BIT
we now get a small black bar on the right of the screen.
If the surface is too big to fit the limits only the upper left area is shown.
On top of that this fixes 2 memory corruption issues:
The first was actually discovered during playing
around with a Windows 7 vServer. During resolution
change in Windows 7 it happens sometimes that Windows
changes to an intermediate resolution where
server_stride % cmp_bytes != 0 (in vnc_refresh_server_surface).
This happens only if width % VNC_DIRTY_PIXELS_PER_BIT != 0.
The second is a theoretical issue, but is maybe exploitable
by the guest. If for some reason the guest surface size is bigger
than VNC_MAX_WIDTH x VNC_MAX_HEIGHT we end up in severe corruption since
this limit is nowhere enforced.
Signed-off-by: Peter Lieven <pl@kamp.de>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Since VNC_CONNECTED, VNC_DISCONNECTED, VNC_INITIALIZED share some
common functions, convert them in one patch.
Signed-off-by: Wenchao Xia <wenchaoqemu@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
vnc_update_client currently scans the dirty bitmap of each client
bitwise which is a very costly operation if only few bits are dirty.
vnc_refresh_server_surface does almost the same.
this patch optimizes both by utilizing the heavily optimized
function find_next_bit to find the offset of the next dirty
bit in the dirty bitmaps.
The following artifical test (just the bitmap operation part) running
vnc_update_client 65536 times on a 2560x2048 surface illustrates the
performance difference:
All bits clean - vnc_update_client_new: 0.07 secs
vnc_update_client_old: 10.98 secs
All bits dirty - vnc_update_client_new: 11.26 secs
vnc_update_client_old: 20.19 secs
Few bits dirty - vnc_update_client_new: 0.08 secs
vnc_update_client_old: 10.98 secs
The case for all bits dirty is still rather slow, this
is due to the implementation of find_and_clear_dirty_height.
This will be addresses in a separate patch.
Signed-off-by: Peter Lieven <pl@kamp.de>
Reviewed-by: Wenchao Xia <xiawenc@linux.vnet.ibm.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
