This allows to chain several test cases by using
test_new()/test_destroy().
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Pavel Grunt <pgrunt@redhat.com>
Do not assume the device passed as an argument to
spice_server_char_device_remove_interface() is the same as the current
Reds::vdagent instance. This commit adds a check that this is the case,
and returns early if not.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
spice_new is a bit more safe as return a properly typed pointer.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
The usage was removed with commit 7ea1f2c133
("sound: Use RedChannelClient to receive/send data").
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Pavel Grunt <pgrunt@redhat.com>
Following
https://lists.freedesktop.org/archives/spice-devel/2017-February/035800.html
, the consensus was to use 'bool' rather than 'gboolean', and to use
true/false rather than TRUE/FALSE. New code should attempt to follow
the new style.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
Using X == TRUE is fragile, since the input value is a uint8_t. It would be
surprising if the value was set to 2 or 0xFF and the test failed.
Signed-off-by: Christophe de Dinechin <dinechin@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
Some constants/macros/function prototypes are defined in
red-channel-client.h while they are only used by red-channel-client.c.
This commit moves them to the .c file to make them private.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
RedChannelClient is responsible for talking to the client so it knows
if is connected or not.
These vfuncs where used by DummyChannel used by SoundChannel.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
Most channels don't need to do specific settings for the client socket
so avoid the need to do this setting making easier to setup the client
channnel.
Some improvements and commit subject suggested by Christophe Fergeau.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
TCP_NODELAY flag is set by default for all connection inside
reds.c so there's no need to set again for the single
client channel.
Note that there are still some calls to setsockopt to change this
option.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
Since c3d237 "gstreamer: Avoid memory copy if strides are different" is
only needed when zero copy is disabled. This moves the function
definition to an already existing #ifdef block.
Acked-by: Pavel Grunt <pgrunt@redhat.com>
This is a revert of b76e561d.
For a SpicePlaybackInstance, the base interface must be a
SpicePlaybackInterface instance, not a SpiceBaseInterface instance, or
spice-server code will end up reading out of bounds.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
This is a revert of 93b4f4050^ and 93b4f4050.
For a SpiceCharDeviceInstance, the base interface must be a
SpiceCharDeviceInterface instance, not a SpiceBaseInterface instance, or
spice-server code will end up reading out of bounds.
vmc_state/vmc_write/vmc_read implementations also have to be provided.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
This is a revert of 93b4f4050^ and 93b4f4050.
For a SpiceCharDeviceInstance, the base interface must be a
SpiceCharDeviceInterface instance, not a SpiceBaseInterface instance, or
spice-server code will end up reading out of bounds.
vmc_state/vmc_write/vmc_read implementations also have to be provided.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
Use new structures and functions to implement the statistics code.
Use inline functions instead of macros for increased type-safety.
If statistics are disabled, the structures and functions become
empty. This confines the configuration-specific #defines to the
statistics implementation itself and avoids the need for #defines in
the calling functions. This greatly reduces the chance of accidentally
breaking the build for one configuration or the other. The reds option
was removed from stat_inc_counter() as it was not used.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
Non blocking flag is set for all connection inside reds.c so
there's no need to set again for the single client channel.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
Do not make it assume vec contains IOV_MAX elements.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
They no longer contain any vfuncs, so calling them "handler" does not
make a lot of sense. This commit renames them to
OutgoingMessageBuffer/IncomingMessageBuffer.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
Nothing outside of RedChannelClient needs access to data contained in
RedChannelClientPrivate, so we can move all the type definitions to the
.c file to make it fully opaque rather than relying on a private header.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
The methods previously used by OutgoingHandlerInterface and
IncomingHandlerInterface are no longer used as generic callbacks,
but are directly called from RedChannelClient code. We can be explicit
about the type of their first argument (RedChannelClient *) rather than
using a generic void * pointer.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
There is only one implementation of OutgoingHandler which relies
OutgoingHandler::opaque being a RedChannelClient. This commit makes this
explicit in order to get rid of the OutgoingHandler::opaque data member.
This renames red_peer_handle_outgoing() to
red_channel_client_handle_outgoing() as the method is now very much tied
to RedChannelClient.
If we want to keep some genericity, we could return error codes from
red_channel_client_handle_outgoing() and handle RedChannelClient
disconnection/... from the caller rather than directly in the
_handle_outgoing() method. This would probably allow to move the
data reading logic to reds-stream.c
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
There is only one implementation of IncomingHandler which relies
IncomingHandler::opaque to be a RedChannlClient. This commit makes this
explicit. This allows to get rid of the IncomingHandler::opaque data
member.
This renames red_peer_handle_incoming() to
red_channel_client_handle_incoming() as the method is now very much tied
to RedChannelClient.
If we want to keep some genericity, we could return error codes from
red_channel_client_handle_incoming() and handle RedChannelClient
disconnection/... from the caller rather than directly in the
_handle_incoming() method. This would probably allow to move the
data reading logic to reds-stream.c
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
This commit removes what remains of IncomingHandlerInterface. The
remaining function pointers were pointing to RedChannel vfuncs.
Moreover the IncomingHandlerInterface abstraction is unused, ie the
codebase only has a single implementation for it, so we can directly
call the relevant methods and make them static instead.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
This commit removes IncomingHandlerInterface::on_error and
IncomingHandlerInterface::on_input. As with previous commits, these
vfuncs are always calling the method, and RedChannel::init sets them to
point to RedChannelClient methods, which RedChannelClient is then going
to call indirectly through the IncomingHandlerInterface instance.
This commit changes this to direct calls to the corresponding methods.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
Similarly to the previous commits, this removes an indirection level,
IncomingHandlerInterface stores pointers to
alloc_recv_buf/release_recv_buf vfuncs, but these are always set to
RedChannel::alloc_recv_buf/RedChannel::release_recv_buf, which are also
vfuncs which can be overridden if needed. This commit removes the
indirection and directly calls the relevant methods.
These vfuncs really should be in RedChannelClient rather than
RedChannel.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
RedChannel uses OutgoingHandlerInterface to provide constant pointers to
RedChannelClient methods. This OutgoingHandlerInterface structure is
then used in RedChannelClient to indirectly call these methods.
The OutgoingHandlerInterface abstraction is unused, ie the codebase
only has a single implementation for it, so we can directly call the
relevant methods and make them static instead.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <figlio@redhat.com>
Have the RedChannelClient callback call into a RedChannel callback
rather than doing the opposite. This will be useful in some subsequent
refactoring of this code.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
red_channel_client_parse() currently does roughly:
if (klass->parser) {
parsed = klass->parser(msg, msg_size, &parsed_size);
klass->handle_parsed(rcc, parsed_size, msg_type, parsed);
} else {
klass->handle_message(rcc, msg_type, msg, msg_size);
}
The handle_parsed implementation expects a void * 'parsed' argument,
which will then be cast to the correct type. There is not really a need
to provide distinct handle_parsed/handle_message vfuncs, instead we can
say that if a RedChannel subclass provides a 'parser' vfunc, then it's
'handle_message' vfunc will be called with the parsed message, otherwise
it will be called with unparsed data (ie what handle_message currently
expects).
This makes the code slightly easier to follow as messages will always be
handled by the same vfunc.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
smartcard_channel_client_pipe_add_push was just calling
red_channel_client_pipe_add_push without any cast or other
changes.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
This makes the code slightly more readable as this means less local
variables to keep track of when taking a high level view of that code.
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Jonathon Jongsma <jjongsma@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
read_binary() attaches 'mem' to the SpiceReplay::allocated list.
On failure, SpiceReplay::allocated and its content are freed by
spice_replay_free().
SpiceReplay::primary_mem is also freed, which causes a double free
as replay_handle_create_primary() added 'mem' both to
SpiceReplay::primary_mem and SpiceReplay::allocated.
This commit avoids this by ensuring SpiceReplay::primary_mem is not
kept in the SpiceReplay::allocated list.
Note that this double free can happen only on currupted or wrong
record images.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
The Surface and Display channels each have a 'current_list' Ring, and
Surface also has a 'current' Ring. these names are confusing, so at
minimum, add a comment indicating the type of object they hold. The
DisplayChannel::current_list already had a comment, but it was
incorrect.
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
Caller is supposed the function return a buffer able to store
size bytes.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
The limits for capabilities are specified using 32 bit unsigned integers.
This could cause possible integer overflows causing buffer overflows.
For instance the sum of num_common_caps and num_caps can be 0 avoiding
additional checks.
As the link message is now capped to 4096 and the capabilities are
contained in the link message limit the capabilities to 1024
(capabilities are expressed in number of uint32_t items).
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
