Refactor qemu_xkeymap_mapping_table() to have a single exit point,
so we can easily free the memory allocated by XGetAtomName().
This fixes when running a binary configured with --enable-sanitizers:
Direct leak of 22 byte(s) in 1 object(s) allocated from:
#0 0x561344a7473f in malloc (qemu-system-x86_64+0x1dab73f)
#1 0x7fa4d9dc08aa in XGetAtomName (/lib64/libX11.so.6+0x2a8aa)
Fixes: 2ec78706d1 ("ui: convert GTK and SDL1 frontends to keycodemapdb")
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210430155009.259755-1-philmd@redhat.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
This patch add vnc_display_reload_certs() to support
update x509 certificates.
Signed-off-by: Zihao Chang <changzihao1@huawei.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20210316075845.1476-3-changzihao1@huawei.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The VNC ACL concept has been replaced by the pluggable "authz" framework
which does not use monitor commands.
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
A long time ago the VNC server code had some memory corruption
fixes done in:
commit bea60dd767
Author: Peter Lieven <pl@kamp.de>
Date: Mon Jun 30 10:57:51 2014 +0200
ui/vnc: fix potential memory corruption issues
One of the implications of the fix was that the VNC server would have a
thin black bad down the right hand side if the guest desktop width was
not a multiple of 16. In practice this was a non-issue since the VNC
server was always honouring a guest specified resolution and guests
essentially always pick from a small set of sane resolutions likely in
real world hardware.
We recently introduced support for the extended desktop resize extension
and as a result the VNC client has ability to specify an arbitrary
desktop size and the guest OS may well honour it exactly. As a result we
no longer have any guarantee that the width will be a multiple of 16,
and so when resizing the desktop we have a 93% chance of getting the
black bar on the right hand size.
The VNC server maintains three different desktop dimensions
1. The guest surface
2. The server surface
3. The client desktop
The requirement for the width to be a multiple of 16 only applies to
item 2, the server surface, for the purpose of doing dirty bitmap
tracking.
Normally we will set the client desktop size to always match the server
surface size, but that's not a strict requirement. In order to cope with
clients that don't support the desktop size encoding, we already allow
for the client desktop to be a different size that the server surface.
Thus we can trivially eliminate the black bar, but setting the client
desktop size to be the un-rounded server surface size - the so called
"true width".
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20210311182957.486939-5-berrange@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The WMVi message is supposed to provide the same width/height
information as the regular desktop resize and extended desktop
resize messages. There can be times where the client width and
height are different from the pixman surface dimensions.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20210311182957.486939-4-berrange@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
We plan framebuffer update rects based on the VNC server surface. If the
client doesn't support desktop resize, then the client bounds may differ
from the server surface bounds. VNC clients may become upset if we then
send an update message outside the bounds of the client desktop.
This takes the approach of clamping the rectangles from the worker
thread immediately before sending them. This may sometimes results in
sending a framebuffer update message with zero rectangles.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20210311182957.486939-3-berrange@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
This adds trace points for desktop size and audio related messages.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20210311182957.486939-2-berrange@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
ui/cocoa used to call exit immediately after calling
qemu_system_shutdown_request, which prevents QEMU from actually
perfoming system shutdown. Just sleep forever, and wait QEMU to call
exit and kill the Cocoa thread.
Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Message-Id: <20210219111652.20623-1-akihiko.odaki@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
OpenGL ES does not support conversion from the given data format
to the internal format with glTexImage2D.
Use the given data format as the internal format, and ignore
the given alpha channels with GL_TEXTURE_SWIZZLE_A in case the
format contains alpha channels.
Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Message-Id: <20210219094803.90860-1-akihiko.odaki@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
With the new "password-secret" option, there is no reason to use the old
inecure "password" option with -spice, so it can be deprecated.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20210311114343.439820-4-berrange@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Currently when using SPICE the "password" option provides the password
in plain text on the command line. This is insecure as it is visible
to all processes on the host. As an alternative, the password can be
provided separately via the monitor.
This introduces a "password-secret" option which lets the password be
provided up front.
$QEMU --object secret,id=vncsec0,file=passwd.txt \
--spice port=5901,password-secret=vncsec0
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20210311114343.439820-3-berrange@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Currently when using VNC the "password" flag turns on password based
authentication. The actual password has to be provided separately via
the monitor.
This introduces a "password-secret" option which lets the password be
provided up front.
$QEMU --object secret,id=vncsec0,file=passwd.txt \
--vnc localhost:0,password-secret=vncsec0
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20210311114343.439820-2-berrange@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
ui/cocoa deassociates the mouse input and the mouse cursor
position only when relative movement inputs are expected. Such
inputs may let the mouse cursor leave the view and cause undesired
side effects if they are associated. On the other hand, the
problem does not occur when inputting absolute points, and the
association allows seamless cursor movement across views.
However, the synchronization of the association and the expected
input type was only done when grabbing the mouse. In reality, the
state whether the emulated input device expects absolute pointing
inputs or relative movement inputs can vary dynamically due to
USB device hot-plugging, for example.
This change adds association state updates according to input type
expectation changes. It also removes an internal flag representing
the association state because the state can now be determined with
the current input type expectation and it only adds the
complexity of the state tracking.
Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Message-Id: <20210222150714.21766-1-akihiko.odaki@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
ui/cocoa does not receive NSEventTypeFlagsChanged when it is not active,
and the modifier state can be desynchronized in such a situation.
[NSEvent -modifierFlags] tells whether a modifier is *not* pressed, so
check it whenever receiving an event and clear the modifier if it is not
pressed.
Note that [NSEvent -modifierFlags] does not tell if a certain modifier
*is* pressed because the documented mask for [NSEvent -modifierFlags]
generalizes left shift and right shift, for example. CapsLock is the
only exception. The pressed state is synchronized only with
NSEventTypeFlagsChanged.
This change also removes modifier keys from keycode map. If they
are input with NSEventTypeKeyDown or NSEventTypeKeyUp, it leads to
desynchronization. Although such a situation is not observed, they are
removed just in case.
Moreover, QKbdState is introduced for automatic key state tracking.
Thanks to Konstantin Nazarov for testing and finding a bug in this
change:
https://gist.github.com/akihikodaki/87df4149e7ca87f18dc56807ec5a1bc5#gistcomment-3659419
Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Message-Id: <20210310144602.58528-1-akihiko.odaki@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The first argument of the executable was used to get its path, but it is
not reliable because the executer can specify any arbitrary string. Use the
interfaces provided by QEMU and the platform to get those paths.
Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Message-Id: <20210309122226.23117-2-akihiko.odaki@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
kCGColorSpaceGenericRGB | Apple Developer Documentation
https://developer.apple.com/documentation/coregraphics/kcgcolorspacegenericrgb
> Deprecated
> Use kCGColorSpaceSRGB instead.
This change also removes the legacy color space specification for
PowerPC.
Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Message-Id: <20210305121304.65096-1-akihiko.odaki@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
c821a58ee7 ("ui/console: Pass placeholder surface to display")
eliminated the possibility that NULL is passed as surface to
dpy_gfx_switch and removed some NULL checks from gd_switch, but the
removal was not thoroughly. Remaining NULL checks were confusing for
Coverity and probably also for humans. This change removes those NULL
checks.
Reported-by: Coverity (CID 1448421)
Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-Id: <20210308140713.17901-1-akihiko.odaki@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The 'running' argument from VMChangeStateHandler does not require
other value than 0 / 1. Make it a plain boolean.
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Acked-by: David Gibson <david@gibson.dropbear.id.au>
Message-Id: <20210111152020.1422021-3-philmd@redhat.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Follow the inclusive terminology from the "Conscious Language in your
Open Source Projects" guidelines [*] and replace the words "whitelist"
appropriately.
[*] https://github.com/conscious-lang/conscious-lang-docs/blob/main/faq.md
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210303184644.1639691-2-philmd@redhat.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
ui/console used to accept NULL as graphic console surface, but its
semantics was inconsistent among displays:
- cocoa and gtk-egl perform NULL dereference.
- egl-headless, spice and spice-egl do nothing.
- gtk releases underlying resources.
- sdl2-2d and sdl2-gl destroys the window.
- vnc shows a message, "Display output is not active."
Fortunately, only virtio-gpu and virtio-gpu-3d assign NULL so
we can study them to figure out the desired behavior. They assign
NULL *except* for the primary display when the device is realized,
reset, or its scanout is disabled. This effectively destroys
windows for the (uninitialized) secondary displays.
To implement the consistent behavior of display device
realization/reset, this change embeds it to the operation
switching the surface. When NULL was given as a new surface when
switching, ui/console will instead passes a placeholder down
to each display listeners.
sdl destroys the window for a secondary console if its surface is a
placeholder. The other displays simply shows the placeholder.
Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Message-Id: <20210225101316.83940-2-akihiko.odaki@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The surfaces created with former qemu_create_message_surface
did not display the content from the guest and always contained
simple messages describing the reason.
A display backend may want to hide the window showing such a
surface. This change renames the function to
qemu_create_placeholder_surface, and adds "placeholder" flag; the
display can check the flag to decide to do anything special like
hiding the window.
Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Message-Id: <20210225101316.83940-1-akihiko.odaki@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
This has the following visible changes:
- GBM is required only for OpenGL dma-buf.
- X11 is explicitly required by gtk-egl.
- EGL is now mandatory for the OpenGL displays.
The last one needs some detailed description. Before this change,
EGL was tested only for OpenGL dma-buf with the check of
EGL_MESA_image_dma_buf_export. However, all of the OpenGL
displays depend on EGL and EGL_MESA_image_dma_buf_export is always
defined by epoxy's EGL interface.
Therefore, it makes more sense to always check the presence of EGL
and say the OpenGL displays are available along with OpenGL dma-buf
if it is present.
Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Message-Id: <20210223060307.87736-1-akihiko.odaki@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
A display can receive an image which its stride is greater than its
width. In fact, when a guest requests virtio-gpu to scan out a
smaller part of an image, virtio-gpu passes it to a display as an
image which its width represents the one of the part and its stride
equals to the one of the whole image.
This change makes ui/cocoa to cover such cases.
Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Message-Id: <20210222144012.21486-1-akihiko.odaki@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
When using the GTK UI with libvte, multicharacter keystrokes are not
sent correctly (such as arrow keys). gd_vc_in should check the
CharBackend's can_receive instead of assuming multiple characters can be
received. This is not an issue for e.g. the SDL UI because
qemu_chr_be_write is called with len=1 for each character (SDL sends
more than once keystroke).
Modify gd_vc_in to call qemu_chr_be_write multiple times if necessary.
Buglink: https://bugs.launchpad.net/qemu/+bug/1407808
Signed-off-by: Zack Marvel <zpmarvel@gmail.com>
Message-Id: <20210221170613.13183-2-zpmarvel@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The detections of [NSView -enterFullScreen:] and
[NSView -exitFullScreen:] were wrong. A detection is coded as:
[NSView respondsToSelector:@selector(exitFullScreenModeWithOptions:)]
but it should be:
[NSView instancesRespondToSelector:@selector(exitFullScreenModeWithOptions:)]
Because of those APIs were not detected, ui/cocoa always falled
back to a borderless window whose frame matches the screen to
implement fullscreen behavior.
The code using [NSView -enterFullScreen:] and
[NSView -exitFullScreen:] will be used if you fix the detections,
but its behavior is undesirable; the full screen view stretches
the video, changing the aspect ratio, even if zooming is disabled.
This change removes the code as it does nothing good.
Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Message-Id: <20210220013138.51437-1-akihiko.odaki@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
It is not used, and it is unlikely that a new use case will emerge
anytime soon because the scope of OpenGL contexts are limited due to
the nature of the frontend, VirGL, processing simple commands from the
guest.
Remove the function and ease implementing a new OpenGL backend a little.
Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Message-Id: <20210219094702.90789-1-akihiko.odaki@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
There is no need of dynamic allocation as dcl is a small singleton.
Static allocation reduces code size and makes hacking with ui/cocoa a
bit easier.
Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Message-Id: <20210219084419.90181-1-akihiko.odaki@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Old Macs were not equipped with mice with an ability to perform
"right clicks" and ui/cocoa interpreted left button down with
left command key pressed as right button down as a workaround.
The workaround has an obvious downside: you cannot tell the guest
that the left button is down while the left command key is
pressed.
Today, Macs has trackpads, Apple Mice, or Magic Mice. They are
capable to emulate right clicks with gestures, which also allows
to perform right clicks on "BootCamp" OSes like Windows.
By removing the workaround, we overcome its downside, and provide
a behavior consistent with BootCamp.
Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Message-Id: <20210212000706.28616-1-akihiko.odaki@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
When qemu is built with modules, but a given module doesn't load
qemu should handle that gracefully. When ui-spice-core.so isn't
able to be loaded and qemu is invoked with -display spice-app or
-spice, qemu will dereference a null pointer. With this change we
check the pointer before dereferencing and error out in a normal
way.
Signed-off-by: Bruce Rogers <brogers@suse.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210213032318.346093-1-brogers@suse.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The old CocoaView had an idea of synchronizing the host window
configuration and the guest screen configuration. Here, the guest screen
actually means pixman image given ui/cocoa display implementation.
However, [CocoaView -drawRect:] directly interacts with the pixman
image buffer in reality. There is no such distinction of "host" and
"guest." This change removes the "host" configuration and let drawRect
consistently have the direct reference to pixman image. It allows to
get rid of the error-prone "sync" and reduce code size a bit.
Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Message-Id: <20210212000629.28551-1-akihiko.odaki@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
This solves the client having slow/outdated VGA/2D console. It's a
regression introduced when the code was switched to render it via opengl
in commit 4423184376 ("spice/gl: render DisplaySurface via opengl")
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20210216092056.2301293-2-marcandre.lureau@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Since the introduction of spice/virgl support in commit
474114b7 ("spice: add opengl/virgl/dmabuf support"), the drawing isn't
being flushed before notifying the client. This results in
outdated/sluggish drawing on client side, in particular when using the
Linux console.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20210216092056.2301293-1-marcandre.lureau@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Displaying rendered resources requires blocking qemu GPU to avoid extra
framebuffer copies. For an external display, via Spice currently, there
is a callback to block/unblock the rendering in the same thread.
But with the vhost-user-gpu backend, the qemu process doesn't handle
the rendering itself, and the blocking callback isn't effective.
Instead, the backend must be notified when the display code is done.
Fix this by adding a new GraphicHwOps callback to indicate the GL state
is flushed, and we are done manipulating the shared GL resources. Call
it from gtk and spice display.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20210204105232.834642-19-marcandre.lureau@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
GtkGLArea is used on wayland, where EGL is usually available.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20210204105232.834642-17-marcandre.lureau@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
This check is currently limited. It only is used by vhost-user-gpu (not
by vfio-display), and will print an error repeatedly during run-time.
We are going to dissociate the GL context from the
DisplayChangeListener, and listeners may come and go. The following
patches will address this differently.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20210204105232.834642-10-marcandre.lureau@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
There are no users left.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20210204105232.834642-7-marcandre.lureau@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
QEMU used to run qemu_spice.display_init() before vm_start(), and
QXL/display interfaces where started then. Now, vm_start() happens
before QXL/display interfaces are added and Spice server doesn't
automatically start them in this case (fixed in spice git)
Fixes Spice regression introduced after 5.2, with refactoring commits
b4e1a34211 ("vl: remove separate preconfig main_loop") and
facf7c60ee ("vl: initialize displays _after_ exiting preconfiguration"),
probably others.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20210129152351.161971-1-marcandre.lureau@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
