Mesa's eglGetDisplay() chooses the native EGL platform from
EGL_PLATFORM, limited autodetection, or the build-time default. If that
selects Wayland while SDL is using the X11 video backend, Mesa can treat
the X11 Display pointer as a wl_display and crash during eglInitialize().
Probe EGL with the X11 platform explicitly before enabling
SDL_HINT_VIDEO_X11_FORCE_EGL.
Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/3540
Signed-off-by: Akihiko Odaki <odaki@rsg.ci.i.u-tokyo.ac.jp>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-ID: <20260611-sdl-v1-1-93d4a51684bb@rsg.ci.i.u-tokyo.ac.jp>
qemu_create_displaysurface_from() callers such as xlnx_dp.c pass
linesize=0 with data=NULL, relying on pixman to compute the stride.
Since 1ff788db97 ("ui: use a shareable type"), the data=NULL path
goes through qemu_pixman_image_new_shareable() which computes
size = height * rowstride_bytes, resulting in a zero-size allocation
and an abort in qemu_memfd_alloc().
Introduce qemu-pixman-helpers.h with overflow-safe stride and buffer
size computation (matching pixman's create_bits() formula), and use it
from both qemu_pixman_image_new_shareable() and pixman-minimal's
create_bits().
Reported-by: Peter Maydell <peter.maydell@linaro.org>
Fixes: 1ff788db97 ("ui: use a shareable type")
Reviewed-by: Philippe Mathieu-Daudé <philmd@oss.qualcomm.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-ID: <20260611113614.1935094-1-marcandre.lureau@redhat.com>
Add per-handler LED state and a NotifierList for UI backends to
subscribe to LED changes.
Devices call qemu_input_handler_set_led() to store their LED state and
notify backends. Notify also on focus change, or list update.
Note: I considered conflating mouse-mode & led-state changes, but those
are quite different events (from different source kinds etc) and we may
want to improve the internal implementation.
Reviewed-by: Philippe Mathieu-Daudé <philmd@oss.qualcomm.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
With both legacy mouse API consumers converted, the remaining
code in input-legacy.c (LED broadcast, index_from_key, qmp_send_key)
is not legacy-specific. Move it to ui/input.c and delete the file.
Clean up include/ui/console.h by removing the now-unused legacy
mouse API declarations (QEMUPutMouseEvent, QEMUPutMouseEntry,
QEMUPutKBDEvent, QEMUPutKbdEntry) and MOUSE_EVENT_* constants.
Reviewed-by: Philippe Mathieu-Daudé <philmd@oss.qualcomm.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
QemuInputEvent now stores Linux key codes for key events. Use those
codes directly instead of translating between internal key code
representations.
Signed-off-by: Akihiko Odaki <odaki@rsg.ci.i.u-tokyo.ac.jp>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-ID: <20260520-input-v3-7-7c9e4c7abe34@rsg.ci.i.u-tokyo.ac.jp>
Add a text console helper that accepts Linux input key codes and use it
as the common implementation for qemu_text_console_put_qcode(). This lets
callers that already use Linux key codes avoid converting them back to
QKeyCode.
Signed-off-by: Akihiko Odaki <odaki@rsg.ci.i.u-tokyo.ac.jp>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-ID: <20260520-input-v3-6-7c9e4c7abe34@rsg.ci.i.u-tokyo.ac.jp>
Linux input key codes are a better internal representation than
QKeyCode:
- With Linux input key codes as the internal representation, keys
previously lost solely because the middle layer between event sources
and sinks used QKeyCode will be preserved, since Linux key codes
cover all keys that those sources and sinks use. For example,
KEY_KPJPCOMMA cannot be represented with QKeyCode, but it is widely
supported by event sources and sinks since it is included in linux,
atset1, atset2, usb, x11, osx, qnum (derived from atset1),
xorgxquartz, and xorgevdev (derived from linux).
- They make it possible to pass through Linux host key codes to Linux
guests to preserve all key inputs.
- They simplify consumers by avoiding QKeyCode aliases, namely
asterisk/kp_multiply and sysrq/print.
This matches the approach used by virtio and Xen.
Signed-off-by: Akihiko Odaki <odaki@rsg.ci.i.u-tokyo.ac.jp>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-ID: <20260520-input-v3-4-7c9e4c7abe34@rsg.ci.i.u-tokyo.ac.jp>
Since commit af07e5ff02 ("ui: convert key events to QKeyCodes
immediately"), all internal key events are expected to be represented as
QKeyCode. Replace KeyValue in QemuInputKeyEvent with QKeyCode to enforce
that and simplify key code retrieval.
Signed-off-by: Akihiko Odaki <odaki@rsg.ci.i.u-tokyo.ac.jp>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-ID: <20260520-input-v3-3-7c9e4c7abe34@rsg.ci.i.u-tokyo.ac.jp>
QAPI represents union members with wrapper structs and pointer
indirections. They are useful at the QMP boundary, but unnecessary for
QEMU's internal input events and make handlers more verbose.
Define QemuInputEvent as a plain internal tagged union and convert input
handlers, queues, and replay code to access payloads directly.
Signed-off-by: Akihiko Odaki <odaki@rsg.ci.i.u-tokyo.ac.jp>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-ID: <20260520-input-v3-2-7c9e4c7abe34@rsg.ci.i.u-tokyo.ac.jp>
Add QemuInputEvent as the input subsystem's name for InputEvent and use
it in input handler, queue, and replay interfaces.
This prepares for decoupling QEMU's internal input event representation
from the QAPI InputEvent type.
Signed-off-by: Akihiko Odaki <odaki@rsg.ci.i.u-tokyo.ac.jp>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-ID: <20260520-input-v3-1-7c9e4c7abe34@rsg.ci.i.u-tokyo.ac.jp>
Rename the display and graphic console public functions to follow a
consistent qemu_console_ (or qemu_graphic_console_) naming convention.
The previous API used a mix of prefixes: dpy_, graphic_hw_,
graphic_console_, console_has_, and update_displaychangelistener().
Unify them under a common qemu_console_ namespace for better
discoverability and consistency.
The main renames are:
- dpy_gfx_*() / dpy_text_*() / dpy_gl_*() → qemu_console_*()
- dpy_{get,set}_ui_info() → qemu_console_{get,set}_ui_info()
- graphic_hw_*() → qemu_console_hw_*()
- graphic_console_*() → qemu_graphic_console_*()
- console_has_gl() → qemu_console_has_gl()
- update_displaychangelistener() → qemu_console_listener_set_refresh()
No functional changes.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Introduce qemu_console_register_listener() which combines setting
dcl->con, dcl->ops and calling register_displaychangelistener() into a
single call. This removes repetitive boilerplate across all display
backends and makes it harder to forget setting one of the fields.
Also move the early-return check in unregister_displaychangelistener()
before the trace call, so that unregistering a never-registered listener
(e.g. on error paths) does not dereference a NULL ops pointer.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Previously, VNC displays were never torn down on QEMU exit, leaking
resources and leaving connected clients with unclean disconnects.
Add vnc_cleanup() to free all VNC displays during qemu_cleanup().
Make vnc_display_close() initiate disconnection of active clients,
and have vnc_display_free() drain the main loop until all clients
have completed their teardown, instead of asserting the client list
is empty.
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Combine the two-step vnc_display_init()/vnc_display_open() sequence
into a single vnc_display_new() function that returns VncDisplay*.
This simplifies the API by making vnc_display_open() an
internal detail and will allow further code simplification.
vnc_display_new() is moved to vnc.h, since it returns VncDisplay* now.
Add vnc_display_free() for consistency, and it will be later used.
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Since commit e2f82e924d ("console: purge curses bits from
console.h"), console_ch_t is a plain uint32_t typedef and
console_write_ch() is a trivial assignment (*dest = ch). These
abstractions were originally needed because console_ch_t was the
curses chtype when CONFIG_CURSES was enabled, and console_write_ch()
handled VGA-to-curses character translation. That commit moved the
curses logic into curses_update(), making the typedef and helper
dead abstractions.
Replace console_ch_t with uint32_t and console_write_ch() calls
with direct assignments.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
The function uses input.c functions to provide a simpler abstraction for
touch events. Let's move it from the already overloaded console.c, and
to avoid some unnecessary dependency from console.c on input.c.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Replace the two-field design (gfx_update void callback + gfx_update_async
flag) with a single bool return value from gfx_update. Returning true
means the update completed synchronously and graphic_hw_update_done()
should be called by the console layer. Returning false means the update
is deferred and the device will call graphic_hw_update_done() itself
later (as done by QXL/SPICE and Apple GFX).
This simplifies the interface and makes the async contract explicit at
each call site rather than relying on a separate struct field.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Drop the bookkeeping, we can simply afford an empty "foreach".
Notice that dpy_gfx_update_texture() is now called even when there are
no listeners. This is more correct, as the texture is not fully
refreshed when a listener connects, so it may be outdated/garbaged.
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
The function is a trivial wrapper around qemu_default_pixman_format()
and qemu_pixelformat_from_pixman(), so make it static inline in
qemu-pixman.h instead of a standalone function in console.c, allowing to
be easily reused.
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Maintain a list of QemuVT100 instances so the cursor timer can directly
iterate over them and call vt100_refresh(), instead of going through
qemu_invalidate_text_consoles() which iterated over all consoles
(including graphic ones) and called back into the generic display layer.
This removes the qemu_invalidate_text_consoles() function from
console.c, further decoupling VT100 text rendering from the console
core.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Replace the type-checking chain in qemu_console_get_label() (using
QEMU_IS_GRAPHIC_CONSOLE/QEMU_IS_TEXT_CONSOLE) with a QemuConsoleClass
virtual method, allowing each console subclass to provide its own
get_label implementation.
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
include/ui/shader.h and include/ui/surface.h are included by files that
do not depend on Epoxy so they shouldn't include epoxy/gl.h. Otherwise,
compilations of these files can fail because the path to the directory
containing epoxy/gl.h may not be passed to the compiler.
Signed-off-by: Akihiko Odaki <odaki@rsg.ci.i.u-tokyo.ac.jp>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20260303-gl-v1-3-d90f0a237a52@rsg.ci.i.u-tokyo.ac.jp>
Don't change bound GL context when creating new GL context for consistency
with behavior expected by virglrenderer that assumes context-creation doesn't
switch context. eglCreateContext() doesn't require GL context to be bound
when it's invoked. Update qemu_egl_create_context() to spawn GL sub-contexts
from a given shared GL context instead of a currently-bound context.
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Message-ID: <20260303151422.977399-4-dmitry.osipenko@collabora.com>
Message-ID: <20260304165043.1437519-6-alex.bennee@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Neither the VNC or SPICE code for password changes provides error
reporting at source, leading the callers to report a largely useless
generic error message.
Fixing this removes one of the two remaining needs for the undesirable
error_printf_unless_qmp() method.
While fixing this the error message hint is improved to recommend the
'password-secret' option which allows securely passing a password at
startup.
Reported-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Allow user to set a preferred scale (defaulting to 1) of the virtual
display. Along with zoom-to-fix=false, this would be helpful for users
running QEMU on hi-dpi host desktop to achieve pixel to pixel display --
e.g., if the scale factor of a user's host desktop is set to 200%, then
they can set a 0.5 scale for the virtual display to avoid magnification
that might cause blurriness.
Signed-off-by: Weifeng Liu <weifeng.liu.z@gmail.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Tested-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20250601045245.36778-3-weifeng.liu.z@gmail.com>
Add a VMStateDescriptor for QemuClipboardInfo.
Each clipboard owner will have to save its QemuClipboardInfo and
reregister its owned clipboard after loading. (the global cbinfo has
only pointers to owners, so it can't restore the relation with its owner
if it was to handle migration)
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
The set_pixel_conversion() method is responsible for determining whether
the VNC client pixel format matches the server format, and thus whether
we can use the fast path "copy" impl for sending pixels, or must use
the generic impl with bit swizzling.
The VNC server format is set at build time to VNC_SERVER_FB_FORMAT,
which corresponds to PIXMAN_x8r8g8b8.
The qemu_pixman_get_format() method is then responsible for converting
the VNC pixel format into a pixman format.
The VNC client pixel shifts are relative to the associated endianness.
The pixman formats are always relative to the host native endianness.
The qemu_pixman_get_format() method does not take into account the
VNC client endianness, and is thus returning a pixman format that is
only valid with the host endianness matches that of the VNC client.
This has been broken since pixman was introduced to the VNC server:
commit 9f64916da2
Author: Gerd Hoffmann <kraxel@redhat.com>
Date: Wed Oct 10 13:29:43 2012 +0200
pixman/vnc: use pixman images in vnc.
The flaw can be demonstrated using the Tigervnc client by using
vncviewer -AutoSelect=0 -PreferredEncoding=raw server:display
connecting from a LE client to a QEMU on a BE server, or the
reverse.
The bug was masked, however, because almost all VNC clients will
advertize support for the "tight" encoding and the QEMU VNC server
will prefer "tight" if advertized.
The tight_pack24 method is responsible for taking a set of pixels
which have already been converted into client endianness and then
repacking them into the TPIXEL format which the RFB spec defines
as
"TPIXEL is only 3 bytes long, where the first byte is the
red component, the second byte is the green component,
and the third byte is the blue component of the pixel
color value"
IOW, the TPIXEL format is fixed on the wire, regardless of what
the VNC client declare as its endianness.
Since the VNC pixel encoding code was failing to honour the endian
flag of the client, the tight_pack24 method was always operating
on data in native endianness. Its impl cancelled out the VNC pixel
encoding bug.
With the VNC pixel encoding code now fixed, the tight_pack24 method
needs to take into account that it is operating on data in client
endianness, not native endianness. It thus may need to invert the
pixel shifts.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
