As the indexes are used to compute the index inside an array
using modulo operation when a signed value overflows, the
modulo becames negative, causing a buffer underflow.
Unlikely to happens (take lot of time) but is safer that way.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
Cursor resources (basically the shape of it) was retained till
it was used however it was copied so there were no reason to not release
this resource.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
You could easily trigger this issue using multiple monitors and
a modified spice-gtk client with this patch:
--- a/src/channel-main.c
+++ b/src/channel-main.c
@@ -1699,6 +1699,7 @@ static gboolean _channel_new(channel_new_t *c)
{
g_return_val_if_fail(c != NULL, FALSE);
+ if (c->type == SPICE_CHANNEL_DISPLAY) c->id = 0;
spice_channel_new(c->session, c->type, c->id);
g_object_unref(c->session);
This as g_initable_new in this case returns NULL (dcc.c).
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
Although dispatcher_send_message() does not allow you to send a message
type that is invalid for a dispatcher, it still makes sense to verify
that the type is valid in the receiver. This should only be possible in
the case of some severe problem such as memory corruption, so if it is
invalid, we simply abort.
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
Is possible to have a leak processing update commands if
the update command is synchronous and the rectangle list
is empty. Note that Qemu always pass an empty list.
If the list is empty display_channel_update fill the list.
This is used to send back the list in case of asynchronous
requests. But in handle_dev_update_async (the callback that
handle the asynchronous case) the list is correctly freed.
This was discovered by accident looking at the code.
Reproduced with a Windows recording file using GCC address
sanitizer and this patch to spice-server-replay:
--- a/server/red-replay-qxl.c
+++ b/server/red-replay-qxl.c
@@ -1280,7 +1280,13 @@ static void replay_handle_dev_input(QXLWorker *worker, SpiceReplay *replay,
replay->created_primary = FALSE;
worker->destroy_surfaces(worker);
break;
- case RED_WORKER_MESSAGE_UPDATE:
+ case RED_WORKER_MESSAGE_UPDATE: {
+ static uint8_t count = 0;
+ QXLRect dummy;
+ QXLRect update = { 0, 0, 100, 100 };
+ count ^= 1;
+ worker->update_area(worker, 0, &update, count ? &dummy : NULL, count ? 1 : 0, 0);
+ } break;
// XXX do anything? we record the correct bitmaps already.
case RED_WORKER_MESSAGE_DISPLAY_CONNECT:
// we want to ignore this one - it is sent on client connection, we
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
RedPipeItem already implements reference counting so
this avoid duplicating code to handle a object with reference
counting that points to another object with reference counting
that holds a RedCursorCmd.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
Now the push is done automatically when a PipeItem is added
(cfr commit 5c460de1a3
"worker: push data when clients can receive them"),
forcing a push cause only network fragmentation and is required
only if you are handling data in a loop instead of using the
default loop.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
Implements red_channel_pipes_add_type and
red_channel_pipes_add_empty_msg using red_channel_pipes_add.
This avoid duplicating items for each client.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
Considering that now RedPipeItem have reference counting
and that lot of items are just used to store constant
data to send, using reference counting instead of creating
different items for each client is easier to do.
So this new red_channel_pipes_add allows to add a single item
to all clients.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
Now the push is done automatically when a PipeItem is added
(cfr commit 5c460de1a3
"worker: push data when clients can receive them"),
forcing a push cause only network fragmentation and is required only if
you are handling data in a polling loop (and thus, you are preventing
the default event loop from running).
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
display-channel.h contains lots of information used by different
DisplayChannel components.
In the past all RedWorker, CursorChannel and DisplayChannel code was in
a single file. Since lots of code to handle DisplayChannel is still in
RedWorker, display-channel.h contains a lot of declarations so that they
can be accessed from RedWorker.
Moving declarations that are not needed by RedWorker and other external
class components helps to reduce dependencies between RedWorker and
DisplayChannel.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
This patch allocates VMC IDs by finding the first ID not used
instead of using a global variable and incrementing the value
for each channel created.
This solves some potential issues:
- remove the global state potentially making possible
to use multiple SpiceServer on the same process;
- don't potentially overflow the variable. This can happen if
channels are allocated/deallocated multiple times
(currently not done by Qemu).
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
This can happen as the connection is asynchronous so (MT main thread,
CT channel thread):
- MT you get a new connection;
- MT a connection is sent to CT;
- MT you get a disconnection of main channel;
- MT red_client_destroy is called;
- CT you attempt to add the RCC to RedClient.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
A RedChannelClient is always attached to a valid RedChannel.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
Name will be visible in debugger and /proc filesystem
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
The message is asynchronous so to avoid the object to potentially
been released before being processed keep a strong reference to
it.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
red_disconnect_display() is duplicating what red_channel_disconnect()
already does, so red_disconnect_display() and red_disconnect_cursor()
are actually identical code-wise. We can directly call
red_channel_disconnect() from flush_commands() rather than passing a
'red_disconnect_t disconnect' argument to that function.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
cursor_channel_disconnect() calls
cursor_channel_client_reset_cursor_cache() on all CursorChannelClient
associated with the current CursorChannel before calling
red_channel_disconnect().
red_channel_disconnect() will iterate over all CursorChannelClient
calling red_channel_client_disconnect(), which will eventually call
CursorChannelClient::on_disconnect. This will in turn
cursor_channel_client_reset_cursor_cache(), so calling it in
cursor_channel_disconnect() before calling red_channel_disconnect() is
redundant.
cursor_channel_disconnect() can thus be replaced by a direct call to
red_channel_disconnect().
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
SoundChannelClient has a stub implementation of
RedChannelClient::on_disconnect(), this commit removes the need for it.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
This vfunc only has a RedChannelClient * argument, and most of the time,
it operates on RedChannelClient, not on RedChannel. Moreover, the only
time it's used is from RedChannelClient. This commit moves the vfunc to
RedChannelClient, which seems like a better fit for it.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
ORC library is used internally by GStreamer to generate code
dynamically.
If ORC cannot allocate executable memory, the failure causes
an abort(3) to be called.
This happens on some SELinux configurations that disable executable
memory allocation (execmem boolean).
Check that ORC could work before attempting to use GStreamer to
avoid crashes.
While this check is done, the ORC library outputs an error which will
be well visible in Qemu output.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
If you are testing for NULL data this means that variable could be
NULL so avoid to access before the check to make sure the check is hit.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
The multimedia time is defined as uint32_t.
Use the proper type instead of int.
Currently no arithmetic is done on this value but
just copies so considering that on the architectures
we support sizeof(int) == sizeof(uint32_t) there's
no change in the resulting machine code.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
A RedClient can be freed from the main thread following a main channel
disconnection (reds_client_disconnect). This can happen while another
thread is allocating a new channel client for that client.
To prevent the usage of a pointer which can be invalid
take ownership of the pointer.
Note that we don't need this when disconnecting as disconnection is
done synchronously (the dispatch messages are registered with
DISPATCH_ACK).
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
Trace the number of loops done processing display commands
and the number of loops in which the queue was full.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
If a DisplayChannelClient cannot be instantiated capabilities
are not released correctly.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
display variable already contains the DCC_TO_DC(dcc) value so
reuse it.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
Some RedChannelClient data members were marked as int when they only
hold booleans.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
reds_get_n_clients is a single line and is used only by
spice_server_get_num_clients.
The 2 functions have very similar names so inlining
reds_get_n_clients does not make code less readable.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
The leak detector we use currently is not enough to detect
some kind of leak in DisplayChannel so manually test.
These tests are enabled only when --enable-extra-checks is passed
to configure.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
Mostly of red_channel_destroy calls were preceded by
a call to unregister the channel.
The only exception was the main channel as this channel is
always present and its initialisation is a bit different.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
RedWorker should not handle directly to client but
defer the job to DisplayChannel.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
Make easier to understant the value to use in the code.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
From spice_server_port_event API you can send port events to
any char device. Although currently this is used only for "port"
devices implemented in spicevmc.c this will allow to support
such events using different objects.
This will be used for instance for a streaming device which
will be a specific SpicePort implementation.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
When a RedCursorCmd is passed to cursor_channel_process_cmd(), it
constructs a new CursorItem which takes ownership of that command. If
the cursor_cmd->type falls through to the default case of the switch
statement, we will print a warning and return without freeing the
CursorItem (and thus the RedCursorCmd).
Acked-by: Frediano Ziglio <fziglio@redhat.com>
Now the push is done automatically when a PipeItem is added
(cfr commit 5c460de1a3
"worker: push data when clients can receive them"),
forcing a push cause only network fragmentation and is required
only if you are handling data in a loop instead of using the
default loop.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Victor Toso <victortoso@redhat.com>
Under error: 'link' fields are being accessed, so it's
wrong to goto error with link == NULL.
Instead, return immediately.
Found by coverity.
Signed-off-by: Uri Lublin <uril@redhat.com>
Acked-by: Pavel Grunt <pgrunt@redhat.com>
If qxl->descriptor.type is QUIC, red_replay_data_chunks_free
frees qxl (data), so no need to free it again at the bottom
of the function.
Found by coverity.
Signed-off-by: Uri Lublin <uril@redhat.com>
Acked-by: Pavel Grunt <pgrunt@redhat.com>
It was also possible for a malicious client to set
VDAgentMonitorsConfig::num_of_monitors to a number larger
than the actual size of VDAgentMOnitorsConfig::monitors.
This would lead to buffer overflows, which could allow the guest to
read part of the host memory. This might cause write overflows in the
host as well, but controlling the content of such buffers seems
complicated.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Total message size received from the client was unlimited. There is
a 2kiB size check on individual agent messages, but the MonitorsConfig
message can be split in multiple chunks, and the size of the
non-chunked MonitorsConfig message was never checked. This could easily
lead to memory exhaustion on the host.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
To help avoid stream.c and dcc.c to access display-channel private
structure to get the nth Stream structure pointer.
Signed-off-by: Victor Toso <victortoso@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
As we have a function for that, don't do the math elsewhere.
Signed-off-by: Victor Toso <victortoso@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
When compiling, -Werror=missing-field-initializers is enabled.
However, some gcc versions (like Gentoo 4.9.4 one) fail to see
that all the members of the SpiceBaseInterface struct are
initialized:
test-display-base.c:844:5: error: missing initializer for field
'description' of 'SpiceBaseInterface'
[-Werror=missing-field-initializers] .base.description = "test
spice virtual channel char device",
The solution is to initialize .base member as a structure at once
instead of multiple times per each member.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
As discussed recently the usage of domain for logging has
different issues (they are not filtered and handled coherently)
and are not widely used in the code.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
This was not done until now, and it's only going to be needed if we receive
a partial ClientMonitorsConfig message.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
The function name is always prepended by the spice_log macro, so we
don't need to explicitly add it in debug messages.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
RedsClientMonitorsConfig duplicates what SpiceBuffer does,
so using we can replace it with SpiceBuffer and make
reds_on_main_agent_monitors_config() simpler.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
This is causing issues with potential improvements to the logging
system, and I've always found this usage a bit odd anyway.
Using spice_debug(""); was not possible as this triggers
-Wformat-zero-length warnings from our use of -Wall.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
In the past CursorItem structure was stored in
RedCursorCmd::device_data field so the check was there to check if the
structure fit into that field.
Since 2ba69f9f88
("libspice: add surface 0 support") the structure is no more stored in
the field so there's no reason for this check causing only confusion.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Victor Toso <victortoso@redhat.com>
The maximum value for a 32 bit variable is UINT32_MAX and not
UINT_MAX. Currently all supported platforms have these two
constants having the same value so this patch don't change
nothing in the generated code but potentially this could
change in a future.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Victor Toso <victortoso@redhat.com>
File transfer and Copy & Paste can be disabled on the server even when
they're supported by the guest agent. Tell it the client by adjusting
the agent capabilities.
Related: rhbz#1373725
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
When a guest is rebooted, the QXL driver gets unloaded at some point in
the reboot process. When the driver is unloaded, the spice server sets a
single flag to FALSE: RedWorker::driver_cap_monitors_config. This flag
indicates whether the driver is capable of sending its own monitors config
messages to the client.
The only place this flag is used is when a new primary surface is created. If
this flag is true, the server assumes that the driver will send its own
monitors config very soon after the surface is created. If it's false, the
server directly sends its own temporary monitors config message to the client
based on the size of the just-created primary surface. This temporary monitors
config message always has a maximum monitor count of 1.
This flag is set to false at startup so that in early boot (e.g. vga text
mode), the server will send out these 'temporary' monitor config messages to
the client whenever the primary surface is destroyed and re-created. This
causes the client to resize its window to match the size of the guest. When the
QXL driver is loaded and starts taking over the monitors config
responsibilities, we set this flag to true and the server stops sending
monitors config messages out on its own.
If we reboot and set this flag to false, it will result in the server sending a
monitors config message to the client indicating that the guest now supports a
maximum of 1 monitor. If the guest happens to have more than one display window
open, it will destroy those extra windows because they exceed the maximum
allowed number of monitors. This means that if you reboot a guest with 4
monitors, after reboot it will only have 1 monitor.
To avoid this, we assume that if we had the ability to support multiple
monitors at some point, that will return at some point. So when the server
constructs its own monitors config message to send to the client (when the
driver_cap_monitors_config flag is false), we send the previous maximum monitor
count instead of always sending a maximum of 1.
Resolves: rhbz#1274447
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
The variable is used to store a boolean type.
Update style for this boolean variable.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Pavel Grunt <pgrunt@redhat.com>
Acked-by: Victor Toso <victortoso@redhat.com>
Use a #define rather than a magic number to make the code a bit more
readable.
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
rcc is already deferenced in red_channel_client_get_client so
checking for NULL after that is uselss.
Also this call is generated from red_channel_client_disconnect
which requires the rcc pointer to be valid.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
The base RedChannel already keeps a list of channel clients, so there's
no need for the SndChannel to also keep track of the client itself.
Since the SndChannel only supports a single client (whereas other
channels may have some partial support for multiple clients), I've
provided a convenience function for getting the client and warning if
there is more than one.
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
The content of these functions almost exclusively deals with channel
client functionality except one line where the channel's active state is
set to TRUE.
These functions are called in two different places.
The first place is from the public API spice_server_record_start() and
spice_server_playback_start(). These functions should alter the
channel's active state, and then set the associated channel client to
active.
The second place is when a new channel client is created. In this
case, it is only called if the channel is already active, so it doesn't
make much sense to set the channel's active state inside of the
function.
To simplify things (and enable some future refactoring), this function
now only deals with the SndChannelClient. The functions have also been
renamed to reflect this fact. The SndChannel's active state is now only
modified from the public API functions.
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Acked-by: Pavel Grunt <pgrunt@redhat.com>
Instead of putting a 'next' link within the channel structure itself,
just use a generic GList structure to keep a list of active sound
channels.
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Acked-by: Pavel Grunt <pgrunt@redhat.com>
When a new PlaybackChannelClient or RecordChannelClient is created,
there are several places where we make decisions based on whether the
client is active or not. But these checks are done before the 'active'
flag is ever set, so this code is effectively dead. This has been the
case since commit 6fdcb931 and d351bb35, so this code has not been
executed for approximately 7 years now.
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
It is only called from the constructor, so move all of the code into
that function.
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
This function is only called from the constructor, so move all of that
code into the constructor.
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
Reverse return values of the various bool methods so that 'true' means
success, and 'false' failure rather than the opposite.
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
The red_get_* methods in red-parse-qxl.c return a boolean, even though
their return type is an int, and they return 1/0. This commit changes
this to the more explicit bool/true/false.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
This use to be an anonymous enum, used as an int in
RedCharDeviceWriteBufferPrivate::origin.
Introducing a typedef clarifies what kind of value it's holding.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
This is intended to hold the fields that only char-device.c has a use
for, but for now this only adds the boilerplate for it, the next commit
will move the relevant field there.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
red_channel_client_clear_sent_item() will clear send_data.blocked, so no
need to do it in red_channel_client_msg_sent() as well.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
Now that red_channel_client_set_blocked() is no longer changing the
events we are watching for, we can call it from
red_channel_client_push().
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
Since 5c460d, we need to watch for WATCH_EVENT_WRITE as long as there are
items queued waiting to be sent, this does not need to be done only when
the network queue is full.
When red_channel_client_set_blocked() is called, as a message is being
sent, WATCH_EVENT_WRITE will already be set, so it does not need to set
it again.
red_channel_client_msg_sent() removes WATCH_EVENT_WRITE, but this will
be done later anyway by red_channel_client_push() if needed.
Since it's redundant, we can remove this.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
red_channel_client_msg_sent() calls red_channel_client_clear_sent_item()
which will clear the rcc->priv->send_data.blocked flag. Later on, it
contains a check for !red_channel_client_is_blocked(), which will always
be true as nothing in between could have set it again. This commit
removes this unneeded check.
This check was already redundant when it was introduced in
9a62a9a809
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
Before this commit, red_channel_client_start_ping_timer() and
red_channel_client_cancel_ping_timer() had checks to be no-ops when
rcc->priv->latency_monitor.timer is NULL.
This commit adds a similar check to
red_channel_client_restart_ping_timer(), and removes explicit NULL
checks before calls to red_channel_client_{cancel,restart,start}_ping_timer()
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
rcc is just used on the next line so cannot be NULL.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
Many function already used bool as boolean type.
Change last gboolean occurrencies and values.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Victor Toso <victortoso@redhat.com>
Commit 590acf3c55 introduced a regression after migration:
PlaybackChannel::connection is never set even if a client is connected,
which means the playback channel is non-functional until a client
reconnects as all the snd_channel_set_xxx() method checks for a non-NULL
PlaybackChannel::connection before sending data to the client.
This commit slightly changes the code flow in
on_new_playback_channel_client()/playback_channel_client_constructed()
so that PlaybackChannel::connection is set regardless of what
red_client_during_migrate_at_target() returns. This is what was done
prior to 590acf3c55.
This resolves https://bugs.freedesktop.org/show_bug.cgi?id=100136
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
This is a particularly opaque part of the code for managing pending
Drawable operations. This patch adds documentation atempting to explain
these functions.
It can be useful for debug to know what the codec of the playback and
record channels are, so add a debug-level print statement indicating
this.
Resolves: rhbz#1436251
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Acked-by: Pavel Grunt <pgrunt@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
Main-channel-client.c has code to send pings when the statistics code is
enabled. This is done through a timer calling a ping_timer_cb callback.
However, this timer is created but never started, so this code is
actually dead-code.
Since this code does not seem to be doing much apart from sending the
pings (which can be achieved by simply setting monitor-latency to TRUE
in main_channel_client_create()), this commit removes it.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
Their uint32_t value is never used, all that matters is whether we
received data or not.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
in main_channel_push_agent_connected(), it used the convenience function
red_channel_test_remote_cap() to determine whether to send the
AGENT_CONNECTED_TOKENS message, or the AGENT_CONNECTED message. However,
this function returns false if *any* one client doesn't support the
capability. We should instead check each individual client separately to
see if they support the capability.
This was a convenience function around
red_channel_client_test_remote_common_cap() that simply iterated over
each client (currently we only really support a single client anyway)
and returned TRUE only if all clients supported the capability. But
nobody ever called this wrapper function.
Both _test_remote_cap() and _test_remote_common_cap() are used as
boolean values, so change the return type from int to bool to follow our
coding standard.
This is a setting for determining whether to compress the audio playback
channel or not. It is variously typed as int or uint32_t. Convert it to
a 'bool' to make it more clear that it is a true/false value rather than
an enumeration or something like that.
The DUMP_BITMAP compile option enable some debugging code to
output image bitmaps in a subdirectory of /tmp.
However if this directory does not exists the server will crash
as not able to create a file in it.
Try to create directory before creating the file.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
- glz_enc_dictionary_reset
- monitors_config_new
- red_channel_any_blocked
- red_channel_no_item_being_sent
- red_client_get_channel
are only used in the file where they are defined, so they can as well be
static.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
This is only available in newer FreeBSD releases (9.1 and later), and
will cause build errors or older versions
This fixes https://bugs.freedesktop.org/show_bug.cgi?id=99213
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
This new function removes one place outside of RedsStream which needs to
access RedsStream::socket
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
This allows to move some low-level code out of reds.c
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
This allows to move some low-level code out of reds.c
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
The code to enable/disable on a TCP socket is duplicated in multiple
places in the code base, this commit replaces this duplicated code with
a helper in RedsStream.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
Doing a memset(0) on a SpiceStatNode does not create an empty/unattached
node, but instead creates a node '0'. This node could be non-existing,
or totally unrelated.
This patch creates a default '0' node as soon as the file is created.
This will make the behaviour of 0-filled nodes more in line with what
one would expect.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
This changes the header guards in all .h files to follow this format:
/*
* Licensing block
*/
/* ... */
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
This is the default value, and the property is marked as
_CONSTRUCT_ONLY, so we don't need to explicitly force the default when
instantiating a RedChannelClient.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
InputsChannelClient::new and SmartcardChannelClient::new both accept a
"monitor_latency" argument, which is always FALSE. It can be removed.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
Before GLib 2.36 you should call g_type_init before attempting
any GLib type usage.
As constructor function are called before even main we need
to call g_type_init much before do_spice_init.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
Stress a bit video encoders.
This check different combination of
- encoder type;
- image formats;
- image clipping (encoding partial frames);
- handling frames split into chunks.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Pavel Grunt <pgrunt@redhat.com>
Handle single chunk writev as normal write.
From some test more than 60% of the times writev is called with 1 as
counter. We can easily and very cheaply turn this call to a simpler
write avoiding the need to pass the array to the kernel.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
As pointed out by danpb, tests/pki/server-cert.pem is only valid until
Fri Nov 29 15:51:44 UTC 2019
This changes tests/pki/server-cert.pem and tests/pki/ca-cert.pem to be
valid until 2048. These certificates were generated using the
instructions on https://www.spice-space.org/page/SSLConnection
The -subj args were omitted, and the defaults suggested by openssl used.
The -days parameter was changed to -days 10950.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
This allows to remove a small hack in server/Makefile.am where we were
using make check-valgrind-memcheck rather than make check-valgrind to
make sure we get a non-0 exit code on failures.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
This allows to run automatically our test-suite with valgrind to test
for memory leaks.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
test-qxl-parsing is really a series of several tests. Porting it to
GTest makes this more obvious. This also has the side-effect of making
it more friendly to 'make check-valgrind' (which would fail on SIGALRM,
and on unexpected g_warning()).
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
On my 64 bit Fedora 25, CMSG_SPACE() adds 4 bytes of padding after the
file descriptor in the control data. This causes warnings when ran under
valgrind as we set msg_controllen to CMSG_SPACE().
This commit fills the control data to 0 to avoid these warnings.
==30301== Syscall param sendmsg(msg.msg_control) points to uninitialised byte(s)
==30301== at 0x8127367: sendmsg (sendmsg.c:28)
==30301== by 0x41880B: reds_stream_send_msgfd (reds-stream.c:295)
==30301== by 0x40953F: main (test-stream.c:121)
==30301== Address 0xffefff1b4 is on thread 1's stack
==30301== in frame #1, created by reds_stream_send_msgfd (reds-stream.c:263)
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Pavel Grunt <pgrunt@redhat.com>
This reverts commit c3d237075b.
When you call gst_buffer_add_video_meta_full GStreamer assumes that
buffer is contiguous. Specifically GStreamer assumes that the first
chunk is big enough to hold a whole frame. This results usually in
some pixel shifts in the video. The pixel shifts you can see are
artifacts due to how the guest sends the frames.
Assuming you allocate the 2 chunks with 2 malloc:
p1 = malloc(size);
...
p2 = malloc(size);
Usually the memory allocator tend to allocate linearly if there are
space adding a prefix to describe next block leading to a memory
arrangement as:
+-------------------+
| p1 prefix |
+-------------------+
| p1 buffer |
+-------------------+
| p2 prefix |
+-------------------+
| p2 buffer |
+-------------------+
now if you take p1 pointer and you assume it points to a 2 * size
buffer you will get p2 prefix in the middle, this prefix is the pixel
shifts.
Problems happens specifically in gst_video_frame_map_id.
This bug is reported in https://bugzilla.gnome.org/show_bug.cgi?id=779524.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
A few dispose/finalize implementations were not chaining up to their
parent classes, causing memory leaks.
This fixes leaks like the following:
==16240== Memcheck, a memory error detector
==16240== Copyright (C) 2002-2015, and GNU GPL'd, by Julian Seward et al.
==16240== Using Valgrind-3.12.0 and LibVEX; rerun with -h for copyright info
==16240== Command: ./test-leaks
==16240==
==16240==
==16240== HEAP SUMMARY:
==16240== in use at exit: 245,490 bytes in 3,638 blocks
==16240== total heap usage: 5,418 allocs, 1,780 frees, 17,676,968 bytes allocated
==16240==
==16240== 16 bytes in 1 blocks are definitely lost in loss record 203 of 771
==16240== at 0x4C2DADE: malloc (vg_replace_malloc.c:298)
==16240== by 0x4C2FC91: realloc (vg_replace_malloc.c:785)
==16240== by 0x52864E: dispatcher_register_handler (dispatcher.c:374)
==16240== by 0x5293E0: main_dispatcher_constructed (main-dispatcher.c:315)
==16240== by 0x7F873DB: g_object_new_internal (gobject.c:1823)
==16240== by 0x7F87EE4: g_object_new_valist (gobject.c:2042)
==16240== by 0x7F86E90: g_object_new (gobject.c:1626)
==16240== by 0x5292A5: main_dispatcher_new (main-dispatcher.c:295)
==16240== by 0x429A0A: do_spice_init (reds.c:3416)
==16240== by 0x42A3F5: spice_server_init (reds.c:3663)
==16240== by 0x4095B1: server_leaks (test-leaks.c:45)
==16240== by 0x844C60A: test_case_run (gtestutils.c:2161)
==16240== by 0x844C9CA: g_test_run_suite_internal (gtestutils.c:2244)
==16240== by 0x844CA73: g_test_run_suite_internal (gtestutils.c:2256)
==16240== by 0x844CC8A: g_test_run_suite (gtestutils.c:2332)
==16240== by 0x844B92C: g_test_run (gtestutils.c:1599)
==16240== by 0x409A0B: main (test-leaks.c:126)
==16240==
==16240== 16 bytes in 1 blocks are definitely lost in loss record 204 of 771
==16240== at 0x4C2DADE: malloc (vg_replace_malloc.c:298)
==16240== by 0x4C2FC91: realloc (vg_replace_malloc.c:785)
==16240== by 0x52864E: dispatcher_register_handler (dispatcher.c:374)
==16240== by 0x5293E0: main_dispatcher_constructed (main-dispatcher.c:315)
==16240== by 0x7F873DB: g_object_new_internal (gobject.c:1823)
==16240== by 0x7F87EE4: g_object_new_valist (gobject.c:2042)
==16240== by 0x7F86E90: g_object_new (gobject.c:1626)
==16240== by 0x5292A5: main_dispatcher_new (main-dispatcher.c:295)
==16240== by 0x429A0A: do_spice_init (reds.c:3416)
==16240== by 0x42A3F5: spice_server_init (reds.c:3663)
==16240== by 0x40BFD4: test_new (test-display-base.c:902)
==16240== by 0x40979D: vmc_leaks (test-leaks.c:92)
==16240== by 0x844C60A: test_case_run (gtestutils.c:2161)
==16240== by 0x844C9CA: g_test_run_suite_internal (gtestutils.c:2244)
==16240== by 0x844CA73: g_test_run_suite_internal (gtestutils.c:2256)
==16240== by 0x844CC8A: g_test_run_suite (gtestutils.c:2332)
==16240== by 0x844B92C: g_test_run (gtestutils.c:1599)
==16240== by 0x409A0B: main (test-leaks.c:126)
==16240==
==16240== 96 bytes in 1 blocks are definitely lost in loss record 638 of 771
==16240== at 0x4C2FA50: calloc (vg_replace_malloc.c:711)
==16240== by 0x8427D3C: g_malloc0 (gmem.c:124)
==16240== by 0x842801F: g_malloc0_n (gmem.c:355)
==16240== by 0x527B44: dispatcher_constructed (dispatcher.c:141)
==16240== by 0x529321: main_dispatcher_constructed (main-dispatcher.c:307)
==16240== by 0x7F873DB: g_object_new_internal (gobject.c:1823)
==16240== by 0x7F87EE4: g_object_new_valist (gobject.c:2042)
==16240== by 0x7F86E90: g_object_new (gobject.c:1626)
==16240== by 0x5292A5: main_dispatcher_new (main-dispatcher.c:295)
==16240== by 0x429A0A: do_spice_init (reds.c:3416)
==16240== by 0x42A3F5: spice_server_init (reds.c:3663)
==16240== by 0x4095B1: server_leaks (test-leaks.c:45)
==16240== by 0x844C60A: test_case_run (gtestutils.c:2161)
==16240== by 0x844C9CA: g_test_run_suite_internal (gtestutils.c:2244)
==16240== by 0x844CA73: g_test_run_suite_internal (gtestutils.c:2256)
==16240== by 0x844CC8A: g_test_run_suite (gtestutils.c:2332)
==16240== by 0x844B92C: g_test_run (gtestutils.c:1599)
==16240== by 0x409A0B: main (test-leaks.c:126)
==16240==
==16240== 96 bytes in 1 blocks are definitely lost in loss record 639 of 771
==16240== at 0x4C2FA50: calloc (vg_replace_malloc.c:711)
==16240== by 0x8427D3C: g_malloc0 (gmem.c:124)
==16240== by 0x842801F: g_malloc0_n (gmem.c:355)
==16240== by 0x527B44: dispatcher_constructed (dispatcher.c:141)
==16240== by 0x529321: main_dispatcher_constructed (main-dispatcher.c:307)
==16240== by 0x7F873DB: g_object_new_internal (gobject.c:1823)
==16240== by 0x7F87EE4: g_object_new_valist (gobject.c:2042)
==16240== by 0x7F86E90: g_object_new (gobject.c:1626)
==16240== by 0x5292A5: main_dispatcher_new (main-dispatcher.c:295)
==16240== by 0x429A0A: do_spice_init (reds.c:3416)
==16240== by 0x42A3F5: spice_server_init (reds.c:3663)
==16240== by 0x40BFD4: test_new (test-display-base.c:902)
==16240== by 0x40979D: vmc_leaks (test-leaks.c:92)
==16240== by 0x844C60A: test_case_run (gtestutils.c:2161)
==16240== by 0x844C9CA: g_test_run_suite_internal (gtestutils.c:2244)
==16240== by 0x844CA73: g_test_run_suite_internal (gtestutils.c:2256)
==16240== by 0x844CC8A: g_test_run_suite (gtestutils.c:2332)
==16240== by 0x844B92C: g_test_run (gtestutils.c:1599)
==16240== by 0x409A0B: main (test-leaks.c:126)
==16240==
==16240== LEAK SUMMARY:
==16240== definitely lost: 224 bytes in 4 blocks
==16240== indirectly lost: 0 bytes in 0 blocks
==16240== possibly lost: 0 bytes in 0 blocks
==16240== still reachable: 207,718 bytes in 3,312 blocks
==16240== of which reachable via heuristic:
==16240== newarray : 1,536 bytes in 16 blocks
==16240== suppressed: 34,548 bytes in 302 blocks
==16240== Reachable blocks (those to which a pointer was found) are not shown.
==16240== To see them, rerun with: --leak-check=full --show-leak-kinds=all
==16240==
==16240== For counts of detected and suppressed errors, rerun with: -v
==16240== ERROR SUMMARY: 4 errors from 4 contexts (suppressed: 20 from 20)
FAIL test-leaks (exit status: 1)
Acked-by: Frediano Ziglio <fziglio@redhat.com>
The code that manages pending QXL Drawable operations is fairly complex
and difficult to understand. This is an attempt to start documenting
that code to save time when we have to work on this code in the future.
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
This commit changes all functions returning TRUE/FALSE from having an
'int' return value to 'bool'.
This way it's obvious that such a function is not going to return
anything else than TRUE or FALSE.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
