Now that we have the buffer synchronization implemented in the
GLAMOR/GBM code, switch to use that code.
At this point, there is still not functional change intended, this is
still preparation work for a fix that is still to come.
Signed-off-by: Olivier Fourdan <ofourdan@redhat.com>
(cherry picked from commit 256cef8b20)
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1569>
Copy the code to deal with synchronization objects from the window
buffers to the GLAMOR/GBM code.
The idea is to deal with synchronizations for all pixmaps, even when
there is no window buffer involved.
This is still preparation work for the following commits, no functional
change intended at this point.
v2: Use a "xwl_window_buffer *" instead of a "void *data" (Michel)
v3: Bail early if there's no xwl_window_buffer (Michel)
v4: Rename xwl_window_submit_pixmap() to
xwl_glamor_gbm_wait_release_fence() (Michel)
Signed-off-by: Olivier Fourdan <ofourdan@redhat.com>
(cherry picked from commit cc021aca99)
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1569>
We want to decorrelate the explicit buffer synchronization from the
window buffers, and move that to the GLAMOR/GBM code instead.
To do that, we need to be able to invoke the xwl_window_buffer's
release_callback() routine from outside the window buffer code.
For that purpose, introduce xwl_window_buffer_release() which calls
xwl_window_buffer_release_callback() for us.
This is preparation work for the following changes, no functional change
intended at this point.
Signed-off-by: Olivier Fourdan <ofourdan@redhat.com>
(cherry picked from commit b5082073b0)
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1569>
If a child window of the same size is unmapped, we should stop walking
the tree looking for the surface window to use.
Whatever lies beneath is not visible anyway.
This also fixes an issue with the Damage list becoming corrupted when
destroying a window, because the first thing that DeleteWindow() does
is to unmap the window and crush the window tree underneath it.
Signed-off-by: Olivier Fourdan <ofourdan@redhat.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Fixes: 3a0fc268 - xwayland: Add xwl_window::surface_window
Closes: https://gitlab.freedesktop.org/xorg/xserver/-/issues/1680
(cherry picked from commit 32e16082c5)
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1569>
Before this change, the xwayland pkgconfig file will always contain an
includedir directive, even though xwayland is not a linkable shared
library:
prefix=/nix/store/3spcjqp5zcyg8arz6dnsj59fal5yk3jy-xwayland-23.2.6
includedir=${prefix}/include
exec_prefix=${prefix}
xwayland=/nix/store/3spcjqp5zcyg8arz6dnsj59fal5yk3jy-xwayland-23.2.6/bin/Xwayland
[…]
Cflags: -I${includedir}
According to a bug reporter this trips up cmake [1], which expects that
the include directory exists, which it does not since xwayland does not
install any header files.
Add the dataonly directive to pkgsconfig.generate() which will remove
the default "." subdir and ensures that includedir is not set inside the
pkgconfig file. Additionally enforce the install directory to
$libdir/pkgconfig, since it otherwise will be installed to
$datadir/pkgconfig, which precludes programs from finding the pkgconfig
because share/pkgconfig is usually not included in the search path.
The resulting pkgconfig does not contain an includedir:
prefix=/nix/store/p7xhdzl65hfhzf36vxykzp2i9cyy7y6c-xwayland-23.2.6
exec_prefix=${prefix}
xwayland=/nix/store/p7xhdzl65hfhzf36vxykzp2i9cyy7y6c-xwayland-23.2.6/bin/Xwayland
have_glamor=true
have_eglstream=true
have_initfd=true
have_listenfd=true
have_verbose=true
have_terminate_delay=true
have_no_touch_pointer_emulation=true
have_force_xrandr_emulation=true
have_geometry=true
have_fullscreen=true
have_host_grab=true
have_decorate=false
have_enable_ei_portal=true
have_byteswappedclients=true
Name: Xwayland
Description: X Server for Wayland
Version: 23.2.6
[1]: https://github.com/NixOS/nixpkgs/pull/309075#issuecomment-2108381428
Signed-off-by: Rouven Czerwinski <rouven@czerwinskis.de>
Reviewed-by: Olivier Fourdan <ofourdan@redhat.com>
(cherry picked from commit 9df084c8d1)
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1569>
This restores the handling of the XRandR emulation for Xwayland rootless
where it was before commit fa7b1c20.
Some compositors may trigger a protocol error if the viewport source is
larger than the actual window size, having that handled in the window
resize hook makes sure we do not regress.
Signed-off-by: Olivier Fourdan <ofourdan@redhat.com>
(cherry picked from commit 702a419c39)
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1535>
In xwl_set_shape(), xwl_window_set_input_region() should be called only
when the input shape of the toplevel window changes.
However, given that xwl_window_from_window() is going to walk the
ancestor tree until it finds an xwl_window, that lookup function cannot
be used. Instead, xwl_window_get() should be used. It's going to return
a valid xwl_window object iff the specified window has one associated
with it.
Fixes: a4ed100c0 - xwayland: Set wl_surface input region
Closes: https://gitlab.freedesktop.org/xorg/xserver/-/issues/1672
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Signed-off-by: Vlad Zahorodnii <vlad.zahorodnii@kde.org>
(cherry picked from commit 8c2b9f4e71)
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1535>
If multiple flips become ready for the same MSC, we would previously
execute them all sequentially, one per MSC for sync flips. This could
result in an unbounded flip queue and corresponding memory consumption.
With implicit sync, leave the mailbox handling to the compositor for
async flips though.
v2:
* Use present_vblank_rec::sync_flip.
Closes: https://gitlab.freedesktop.org/xorg/xserver/-/issues/1664
Fixes: e1f16fb1ac ("xwayland: don't scrap pending present requests")
(cherry picked from commit 0d9a54aa97)
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1494>
The functions xwl_glamor_dmabuf_import_sync_file and
xwl_glamor_dmabuf_export_sync_file are used to ensure proper
synchronization between clients using PresentPixmapSynced and
compositors that do not support the wp_linux_drm_syncobj_v1 protocol
when presenting by flipping. The acquire point's fence will be imported
as the DMA-BUF's implicit fence before handing it off to the compositor,
and then, after the DMA-BUF has been released, its new implicit fence
will be exported and become the release point's fence which the client
is expected to wait for before re-using the buffer.
Both functions currently set the flags arguments of their respective
ioctls to DMA_BUF_SYNC_READ. When importing a sync file, this means that
any subsequent implicitly synchronized reads from the buffer will not
wait for the fence, and when exporting a sync file it means that the
returned fence may be signaled before preceeding reads from the buffer
have completed.
While this is correct for xwl_glamor_dmabuf_export_sync_file since the
compositor will never write to the buffer, it is incorrect for
xwl_glamor_dmabuf_import_sync_file. To avoid corruption, we need any
reads from the buffer by the compositor to wait on the acquire point's
fence.
As a fix, instead of setting the DMA_BUF_SYNC_READ flag in
xwl_glamor_dmabuf_import_sync_file, we set the DMA_BUF_SYNC_WRITE flag.
This *does* provide the necessary guarantees.
Signed-off-by: Erik Kurzinger <ekurzinger@nvidia.com>
(cherry picked from commit d5192ba8eb)
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1494>
Some applications that use client side decorations usually set custom
input shape in order to prevent drop shadows stealing pointer events
from windows below. Currently, the only way to get it is to use some
XFixes APIs.
On the other hand, plenty of wayland compositors use solely the
wl_surface input region to decide what view can receive pointer input,
which results in some pointer input issues around client side drop
shadows because Xwayland doesn't set wl_surface.input_region.
See-also: https://bugs.kde.org/show_bug.cgi?id=448119
Signed-off-by: Vlad Zahorodnii <vlad.zahorodnii@kde.org>
Reviewed-by: Olivier Fourdan <ofourdan@redhat.com>
(cherry picked from commit a4ed100c0c)
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1494>
On DRM lease connector withdrawn event, Xwayland would free the
corresponding xwl_output offered for lease.
However, the pointer is still referenced from the rrLease->outputs[],
meaning that trying to clean up the RANDR DRM leases as done with commit
ef181265 (xwayland: Clean up drm lease when terminating) would cause a
use after free and random crashes.
To avoid that issue, on the connector withdraw event, set the connector
withdrawn flag but do not to remove (i.e. free) the xwayland output if
its is offered for lease.
Then, once the lease is terminated, check for the xwl_outputs with a
withdrawn connector and remove them (once we have no use for them
anymore.
Note that we cannot do that cleanup from xwl_randr_terminate_lease() as
removing the xwl_output will free the RRcrtc resources, which checks for
leases in XRANDR, and calls RRTerminateLease(), which chains back to
xwl_randr_terminate_lease().
v2: Use a "withdrawn_connector" flag to mark outputs to remove (Xaver)
Signed-off-by: Olivier Fourdan <ofourdan@redhat.com>
Reviewed-by: Xaver Hugl <xaver.hugl@kde.org>
fixes: ef181265 - xwayland: Clean up drm lease when terminating
See-also: https://gitlab.freedesktop.org/xorg/xserver/-/issues/946
See-also: https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1130
(cherry picked from commit 4053782443)
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1494>
In xwl_randr_request_lease(), the code checks first for leased device,
and then checks for existing output for lease.
The former assumes there are outputs for lease whereas the latter checks
for the output, connector and lease.
So if there is any existing rrLease->outputs[]->devPrivate unset, the
code would crash on a NULL pointer dereference on the first sanity check
before having a chance to reach the second check that would have caught
the problem.
Invert the sanity checks so that we would catch this first and return a
BadValue instead of possibly segfaulting.
Signed-off-by: Olivier Fourdan <ofourdan@redhat.com>
Reviewed-by: Xaver Hugl <xaver.hugl@kde.org>
(cherry picked from commit 21916ae148)
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1494>
Use the connector name as basis for the Xwayland output name in XRANDR,
similar to what we do for regular outputs, instead of the generic
"XWAYLAND<n>" name which changes every time the output is leased.
Prefix the actual name with "lease-" to distinguish from duplicate names
from the regular outputs.
v2: avoid duplicate names (Simon)
v3: Move the check for duplicates to xwl_output_set_name() (Simon)
Signed-off-by: Olivier Fourdan <ofourdan@redhat.com>
(cherry picked from commit 49b8f131f7)
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1494>
Even though the name provided by either xdg-output or wl_output are
guaranteed to be unique, that might not be the case with output names
between different protocols, such as the one offered for DRM lease.
To avoid running into name conflicts, check that no other existing
output of the same name exists prior to changing the output name.
Signed-off-by: Olivier Fourdan <ofourdan@redhat.com>
(cherry picked from commit d36f66f15d)
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1494>
We're keeping it for unit tests, but we don't want to ship it from this
branch.
Also disable Xvfb in CI for ninja test. It's still built and used for
unit tests as part of ninja dist, but we don't want to run XTS on Xvfb.
(cherry picked from commit 0408fcb329)
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1483>
In xwl_source_validate(), the actual box wasn't updated, so we would
possibly copy several times the same first box.
Signed-off-by: Olivier Fourdan <ofourdan@redhat.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Fixes: aa05f38f3 - xwayland: Add SourceValidate hook
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1485>
xwl_dmabuf_feedback_tranche_target_device always allocates a new
drmDevice for xwl_feedback->tmp_tranche.drm_dev, so the pointers are
never equal here.
Fixes: 6f0b9deed6 ("xwayland: use drmDevice to compare DRM devices")
v2:
* Flip order of checks, so drmDevicesEqual is called only if the
supports_scanout flags match.
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1484>
Fixes leaks:
==13712== 144 bytes in 1 blocks are definitely lost in loss record 4,827 of 7,462
==13712== at 0x48459F3: calloc (vg_replace_malloc.c:1340)
==13712== by 0x49BE94D: drmDeviceAlloc (xf86drm.c:4072)
==13712== by 0x49BFAC9: drmProcessPciDevice (xf86drm.c:4104)
==13712== by 0x49BFAC9: process_device (xf86drm.c:4508)
==13712== by 0x49C35FB: drmGetDeviceFromDevId (xf86drm.c:4670)
==13712== by 0x1AD370: xwl_dmabuf_feedback_main_device (xwayland-dmabuf.c:477)
==13712== by 0x53C03FD: ffi_call_unix64 (unix64.S:104)
==13712== by 0x53BF70C: ffi_call_int (ffi64.c:673)
==13712== by 0x53BFEE2: ffi_call (ffi64.c:710)
==13712== by 0x49AC920: wl_closure_invoke (connection.c:1025)
==13712== by 0x49A8C08: dispatch_event.isra.0 (wayland-client.c:1631)
==13712== by 0x49AA5AB: dispatch_queue (wayland-client.c:1777)
==13712== by 0x49AA5AB: wl_display_dispatch_queue_pending (wayland-client.c:2019)
==13712== by 0x49AAB5E: wl_display_roundtrip_queue (wayland-client.c:1403)
==13712== 576 bytes in 4 blocks are definitely lost in loss record 6,289 of 7,462
==13712== at 0x48459F3: calloc (vg_replace_malloc.c:1340)
==13712== by 0x49BE94D: drmDeviceAlloc (xf86drm.c:4072)
==13712== by 0x49BFAC9: drmProcessPciDevice (xf86drm.c:4104)
==13712== by 0x49BFAC9: process_device (xf86drm.c:4508)
==13712== by 0x49C35FB: drmGetDeviceFromDevId (xf86drm.c:4670)
==13712== by 0x1AD583: xwl_dmabuf_feedback_main_device (xwayland-dmabuf.c:477)
==13712== by 0x1AD583: xwl_window_dmabuf_feedback_main_device (xwayland-dmabuf.c:691)
==13712== by 0x53C03FD: ffi_call_unix64 (unix64.S:104)
==13712== by 0x53BF70C: ffi_call_int (ffi64.c:673)
==13712== by 0x53BFEE2: ffi_call (ffi64.c:710)
==13712== by 0x49AC920: wl_closure_invoke (connection.c:1025)
==13712== by 0x49A8C08: dispatch_event.isra.0 (wayland-client.c:1631)
==13712== by 0x49AA5AB: dispatch_queue (wayland-client.c:1777)
==13712== by 0x49AA5AB: wl_display_dispatch_queue_pending (wayland-client.c:2019)
==13712== by 0x1A1842: xwl_read_events (xwayland-screen.c:566)
==13712== by 0x1A1842: xwl_read_events (xwayland-screen.c:553)
Fixes: 6f0b9deed6 ("xwayland: use drmDevice to compare DRM devices")
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1484>
It's needed when the surface window is a depth 24 descendant of a depth
32 toplevel window.
xwl_source_validate ensures the toplevel window pixmap has valid
contents when a client reads from it, or when the window hierarchy /
geometry changes. It's never called in the normal fullscreen application
case, so there's no GPU copy overhead with that.
v2:
* Don't try to redirect a depth 32 descendant of different-depth
ancestors, the alpha channel wouldn't be handled correctly.
(Olivier Fourdan)
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1300>
A later commit will use it to ensure the toplevel window pixmap has
valid contents.
It's hooked up only while any xwl_window->surface_window_damage points
to a non-empty region. So far it's always NULL, so no functional change
intended.
v2:
* Fix trailing whitespace. (Olivier Fourdan)
v3:
* Use toplevel local variable more in xwl_window_update_surface_window.
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1300>
A later commit will use these to (un)redirect the surface window on
demand.
Not used yet, so no functional change intended.
v2:
* Use "surface_window_damage" instead of "surf_win_damage".
(Olivier Fourdan)
* Slightly simplify logic in xwl_unrealize_window.
v3:
* Add comment in xwl_present_maybe_unredirect_window explaining why we
use a timer. (Olivier Fourdan)
v4:
* Rename unredir_timer field to unredirect_timer.
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1300>
It may track a non-toplevel window which fully covers the area of the
window pixmap / Wayland surface. It is now used instead of
xwl_window::toplevel for updating the Wayland surface contents.
The surface_window can now hit the Present page flip path while it's
automatically redirected.
v2:
* Use "surface_window" instead of "surf_win". (Olivier Fourdan)
* Add comment describing surface_window, and describe what
surface_window/toplevel are useful for respectively. (Olivier Fourdan)
* Use surface_window in xwl_realize_window.
v3:
* Backtrack up to the closest opaque ancestor in
xwl_window_update_surface_window. (Olivier Fourdan)
v4:
* Clean up logic for determining the surface window in
xwl_window_update_surface_window, and document it better.
* Handle window_get_damage(xwl_window->surface_window) returning NULL
in xwl_window_update_surface_window.
* Call xwl_window_update_surface_window after xwl_window_buffers_init
in ensure_surface_for_window, since the former may call
xwl_window_buffers_dispose.
* Rename surf/win_pix to surface/window_pixmap in
xwl_window_update_surface_window.
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1300>
Preparation for next commit.
This might change behaviour for non-InputOutput top-level windows.
ensure_surface_for_window getting called and returning non-NULL for
those would seem like a pre-existing bug though.
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1300>
It's always the toplevel window, i.e. either the root window or a child
of it.
Preparation for later commits, no functional change.
v2: (Olivier Fourdan)
* Fix debug build.
* Add comment describing ::toplevel.
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1300>
When a present request is received, Xwayland will check if there is an
existing request targeting the same window and msc and scrap the older
request if so. Alas, this does not interact well the older fence-based
or newer syncobj-based synchronization features of the Present
extension.
Since execution of a request may be delayed for an unknown length of
time while waiting for a fence to be signaled, the target msc computed
upon receiving a request may not match the actual msc at which the
request is executed. Therefore, we cannot determine in advance whether a
more recently received request will make an older request redundant.
This change removes the code to scrap pending present requests.
We must also ensure requests are executed in the correct order even if
their fences are signaled out of order. To achieve this, whenever
execution of a request needs to wait for a fence, execution of any
later-received requests will be blocked until the earlier request is
ready. The blocked requests will be added to a list tracked in the
xwl_present_window struct. Once the earlier request's fence is signaled,
any blocked requests will be re-executed.
Signed-off-by: Erik Kurzinger <ekurzinger@nvidia.com>
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/967>
This protocol allows for explicit synchronization of GPU operations by
Wayland clients and the compositor. Xwayland can make use of this to
ensure any rendering it initiates has completed before the target image
is accessed by the compositor, without having to rely on kernel-level
implicit synchronization.
Furthermore, for X11 clients that also support explicit synchronization
using the mechanisms exposed in the DRI3 and Present extensions, this
Wayland protocol allows us to simply forward the timeline, acquire, and
release points directly to the compositor, ideally avoiding any
premature stalls in the presentation pipeline.
Signed-off-by: Erik Kurzinger <ekurzinger@nvidia.com>
Part-of: <https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/967>
