xf86RecolorCursor() may be called directly from XRecolorCursor as well
as from xf86ScreenSetCursor(). In the latter case, the input lock is
already held, but not for the former and so we need to add a wrapper
function that acquires the input lock before performing
xf86RecolorCursor()
References: https://bugs.freedesktop.org/show_bug.cgi?id=99358
(cherry picked from commit 7198a6d4e7)
A client which is attended while a grab is blocking execution of its
requests needs to be placed in the saved_ready_clients list so that it
will get scheduled once the grab terminates. Otherwise, if the client
never sends another request, there is no way for it to be placed in
the ready_clients list.
v2: Wrap comment above mark_client_saved_ready.
Remove test for OS_COMM_IGNORED which will always be true.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99333
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
(cherry picked from commit 785053d033)
xrandr --setprovideroutputsource <screen> <gpu screen>
Xorg: ../../../xserver/dix/dispatch.c:4018: AttachOutputGPU:
Assertion `new->isGPU' failed.
GPUScreen is not allowed to be sink output.
Signed-off-by: Qiang Yu <Qiang.Yu@amd.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
(cherry picked from commit 555e0a42d1)
Signed-off-by: Qiang Yu <Qiang.Yu@amd.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
(cherry picked from commit 1012510620)
We are no longer using the present_flip_queue list only for presents
which have already been submitted to the driver for page flipping, but
also for those which we are queueing up to be flipped later, marked
with vblank->queued == TRUE. We were incorrectly calling
present_flip_notify for such entries, failing the assertion in
present_flip_notify (or presumably resulting in other undesirable
behaviour with assertions disabled).
Reproduction recipe: Run the JavaFX test case referenced by
https://bugs.freedesktop.org/show_bug.cgi?id=98831#c6 and alt-tab out
of it while it's fullscreen. May take a few attempts to hit the
assertion failure.
Fixes: bab0f450a7 ("present: Fix presentation of flips out of order")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98854
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit e473b2bc01)
Detailed mode reports 108 mm x 68 mm which is for smaller display.
Maximum image size reports 15 cm x 10 cm which aligns with its physical
size, use this size instead.
Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 9874f73e88)
Fixes a regression introduced in 0b2f30834b. If a driver posts input
events during a timer function (wacom and synaptics do this during tap
timeouts), ProcessInputEvents() is not called for these events. There are no
new events on any fds, so the events just sit in the queue waiting for
something else to happen.
Fix this by simply returning 0 from check_timers if we ran at least one of
them or reset them all. This way the callers ospoll_wait will exit and
continue with normal processing.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Keith Packard <keithp@keithp.com>
If the libEGL we are using has eglGetPlatformDisplayEXT, yet it still
returns NULL, then this very likely means that it does not support the
type (e.g. EGL_PLATFORM_GBM_MESA) passed in, and then returning NULL is
the right thing to do.
This avoids falling back to an eglGetDisplay() implementation which does
not understands the passed in gbm handle, treats it as a pointer to
something else completely, followed by a crash sooner or later.
Specifically this fixes using the nvidia binary driver, with nvidia's
libEGL + the modesetting driver on a secondary GPU crashing inside
glamor_egl_init() sometimes.
[1.19: squash in typo fix from 29a4f3db - ajax]
Cc: Eric Anholt <eric@anholt.net>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
(cherry picked from commit 05e1964425)
The custom os/os.O library reuses *.o files of os/libos.la.
The current rule assumes automake puts all the objects into per-target
am__*_la_OBJECTS variable. At least with AC_REPLACE_FUNCS, this no
longer holds (as wanted objects are put into LTLIBOBJS instead).
Depend on automake's result, the *.la library instead, to express demand
of any its dependencies being built.
Should be fixing randomly occuring "undefined reference to `strlcpy'"
errors when linking Xvfb and other DDX-es that could use os.O.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Mihail Konev <k.mvc@ya.ru>
(cherry picked from commit 5b74e260e0)
In InitOutput, if xf86HandleConfigFile returns CONFIG_NOFILE
(which it does if no config file or directory is present), the
autoconfig flag is set, causing xf86AutoConfig to be called
later on.
xf86AutoConfig calls xf86OutputClassDriverList via the
call tree:
xf86AutoConfig =>
listPossibleVideoDrivers =>
xf86PlatformMatchDriver =>
xf86OutputClassDriverList
and xf86OutputClassDriverList attempts to traverse a linked list
that is a member of the XF86ConfigRec struct pointed to by the
global xf86configptr, which is NULL at this point because the
XF86ConfigRec struct is only allocated (by xf86readConfigFile)
AFTER the config file and directory have been successfully
opened; the CONFIG_NOFILE return from xf86HandleConfigFile
occurs BEFORE the call to xf86readConfigFile which allocates
the XF86ConfigRec struct.
Rx: In read.c (for symmetry with xf86freeConfig, which already
appears in this file), add a new function xf86allocateConfig
which tests the value of xf86configptr and, if it's NULL,
allocates the XF86ConfigRec struct and deposits the pointer
in xf86configptr. In xf86Parser.h, add a prototype for the
new xf86allocateConfig function.
Back in read.c, #include "xf86Config.h". In xf86readConfigFile,
change the open-code call to calloc to a call to the new
xf86allocateConfig function.
In xf86AutoConfig.c, add a call to the new xf86allocateConfig function
to the beginning of xf86AutoConfig to make sure the XF86ConfigRec struct
is allocated.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Ben Crocker <bcrocker@redhat.com>
(cherry picked from commit 8b335d9068)
Fix the following error on 'make distcheck':
make[6]: *** No rule to make target 'scripts/xvfb-piglit.sh', needed by 'scripts/xvfb-piglit.sh.log'. Stop.
make[6]: Leaving directory '/home/pq/git/xserver/xorg-server-1.19.99.1/_build/sub/test'
Makefile:1367: recipe for target 'check-TESTS' failed
The setup to trigger this is:
$ ./configure --prefix=/home/pq/local --disable-docs
--disable-devel-docs --enable-xwayland --disable-xorg --disable-xvfb
--disable-xnest --disable-xquartz --disable-xwin --enable-debug
SCRIPT_TESTS is populated conditionally, but we should distribute the
scripts in any case.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
(cherry picked from commit b365c5d168)
When going from border width zero to a non-zero border width, the
Composite extension is informed via the ConfigNotify callback. The
call-chain looks like this: compConfigNotify -> compReallocPixmap ->
compSetPixmap -> TraverseTree -> compSetPixmapVisitWindow. However, at
this time, pWindow->borderWidth was not yet updated. Thus, HasBorder()
is false and the window border will not be repainted.
To fix this, thread the new bw through to the window visitor, and
inspect that rather than HasBorder(). For the other callers of
compSetPixmap the border does not change size, so we can pass
pWin->borderWidth instead.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98499
Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
(cherry picked from commit f31875510d)
The purpose of rrCheckPixmapBounding is to make sure that the
screen_pixmap is *large* enough for the slave-output which crtc is
being configured.
However until now rrCheckPixmapBounding would also shrink the
screen_pixmap in certain scenarios leading to various problems.
For example: Take a laptop with its internalscreen on a slave-output and
currently disabled and an external monitor at 1920x1080+0+0.
Now lets say that we want to drive the external monitor at its native
resolution of 2560x1440 and have the internal screen mirror the top left
part of the external monitor, so we run:
$ xrandr --output eDP --mode 1920x1080 --pos 0x0 --output HDMI \
--mode 2560x1440 --pos 0x0
Here xrandr utility first calls RRSetScreenSize to 2560x1440, then it
calls RRSetCrtc 1920x1080+0+0 on the eDP, since this is a slave output,
rrCheckPixmapBounding gets called and resizes the screen_pixmap to
1920x1080, undoing the RRSetScreenSize. Then RRSetCrtc 2560x1440+0+0
gets called on the HDMI, depending on crtc->transforms this will
either result in a BadValue error from ProcRRSetCrtcConfig; or
it will succeed, but the monitor ends up running at 2560x1440
while showing a 1920x1080 screen_pixmap + black borders on the right
and bottom. Neither of which is what we want.
This commit removes the troublesome shrinking behavior, fixing this.
Note:
1) One could argue that this will leave us with a too large screen_pixmap
in some cases, but rrCheckPixmapBounding only gets called for slave
outputs, so xrandr clients already must manually shrink the screen_pixmap
after disabling crtcs in normal setups.
2) An alternative approach would be to also call rrCheckPixmapBounding
on RRSetCrtc on normal (non-slave) outputs, but that would result in
2 unnecessary resizes of the screen_pixmap in the above example, which
seems undesirable.
Cc: Nikhil Mahale <nmahale@nvidia.com>
Cc: Dave Airlie <airlied@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
(cherry picked from commit a46afee84d)
The purpose of rrCheckPixmapBounding is to make sure that the
screen_pixmap is large enough for the slave-output which crtc is
being configured.
This should include crtc->x and crtc->y, otherwise the crtc might
still end up scanning out an area outside of the screen-pixmap.
For example: Take a laptop with an external monitor on a slave-output at
1920x1080+0+0 and its internal-screen at 3840x2160+1920+0 and in
gnome-settings-daemon move the external monitor to be on the ri ght of
the internal screen rather then on the left. First g-s-d will do a
RRSetScreenSize to 5760*2160 (which is a nop), then it calls RRSetCrtc
to move the slave output to 1920x1080+3840+0, since this is a slave
output, rrCheckPixmapBounding gets called, since the 2 crtcs now overlap
the code before this commit would shrinks the screen_pixmap to 3180*2160.
Then g-s-d calls RRSetCrtc to move the internal screen to 3180*2160+0+0.
And we end up with the slave-output configured to scan-out an area
which completely falls outside of the screen-pixmap (and end up with
a black display on the external monitor).
This commit fixes this by not substracting the x1 and y1 coordinates
of the union-ed region when determining the new screen_pixmap size.
Cc: Nikhil Mahale <nmahale@nvidia.com>
Cc: Dave Airlie <airlied@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
(cherry picked from commit 3b624aa9a9)
Sometimes, Xwayland will try to use a cursor that has just been freed,
leading to a crash when trying to access that cursor data either in
miPointerUpdateSprite() or AnimCurTimerNotify().
CheckMotion() updates the pointer's cursor based on which xwindow
XYToWindow() returns, and Xwayland implements its own xwl_xy_to_window()
to fake a crossing to the root window when the pointer has left the
Wayland surface but is still within the xwindow.
But after an xwindow is unrealized, the last xwindow used to match the
xwindows is cleared so two consecutive calls to xwl_xy_to_window() may
not return the same xwindow.
To avoid this issue, update the last_xwindow based on enter and leave
notifications instead of xwl_xy_to_window(), and check if the xwindow
found by the regular miXYToWindow() is a child of the known last
xwindow, so that multiple consecutive calls to xwl_xy_to_window()
return the same xwindow, being either the one found by miXYToWindow()
or the root window.
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1385258
Signed-off-by: Olivier Fourdan <ofourdan@redhat.com>
Tested-by: Vít Ondruch <vondruch@redhat.com>
Tested-by: Satish Balay <balay@fastmail.fm>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
(cherry picked from commit 59ad0e6a41)
In glamor_init(), if the minimum requirements are not met, glamor may
fail after setting up its own CloseScreen() and DestroyPixmap()
routines, leading to a crash when either of the two routines is called
if glamor failed to complete its initialization, e.g:
(EE) Backtrace:
(EE) 0: Xwayland (OsSigHandler+0x29)
(EE) 1: /lib64/libpthread.so.0 (__restore_rt+0x0)
(EE) 2: Xwayland (glamor_sync_close+0x2a)
(EE) 3: Xwayland (glamor_close_screen+0x52)
(EE) 4: Xwayland (CursorCloseScreen+0x88)
(EE) 5: Xwayland (AnimCurCloseScreen+0xa4)
(EE) 6: Xwayland (present_close_screen+0x42)
(EE) 7: Xwayland (dix_main+0x4f9)
(EE) 8: /lib64/libc.so.6 (__libc_start_main+0xf1)
(EE) 9: Xwayland (_start+0x2a)
Restore the previous CloseScreen() and DestroyPixmap() vfunc handlers in
case of failure when checking for the minimum requirements, so that if
any of the requirement is not met we don't leave the CloseScreen() and
DestroyPixmap() from glamor handlers in place.
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1390018
Signed-off-by: Olivier Fourdan <ofourdan@redhat.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
(cherry picked from commit f43207c1c4)
The previous code only worked when the barrier was created by the same client
as the one calling XIChangeDeviceHierarchy.
http://bugzilla.redhat.com/show_bug.cgi?id=1384432
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Daniel Stone <daniels@collabora.com>
(cherry picked from commit d6a6e1d6ab)
Commits 816015648f and
fee0827a9a made it so that
wl_keyboard::enter doesn't result in X clients getting KeyPress events
while still updating our internal xkb state to be in sync with the
host compositor.
wl_keyboard::leave needs to be handled in the same way as its
semantics from an X client POV should be the same as an X grab getting
triggered, i.e. X clients shouldn't get KeyRelease events for keys
that are still down at that point.
This patch uses LeaveNotify for these events on wl_keyboard::leave and
changes the current use of KeymapNotify to EnterNotify instead just to
keep some symmetry between both cases.
On ProcessDeviceEvent() we still need to deactivate X grabs if needed
for KeyReleases.
Signed-off-by: Rui Matos <tiagomatos@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
(cherry picked from commit 5611585b87)
Fixes the following warning:
test/Makefile.am:69: warning: variable 'os_LDADD' is defined but no program or
test/Makefile.am:69: library has 'os' as canonical name (possible typo)
Introduced upon the removal of test/os in:
commit 6a5a4e6037
Author: Keith Packard <keithp@keithp.com>
Date: Tue Dec 8 14:39:46 2015 -0800
Remove SIGIO support for input [v5]
This removes all of the SIGIO handling support used for input
throughout the X server, preparing the way for using threads for input
handling instead.
Places calling OsBlockSIGIO and OsReleaseSIGIO are marked with calls
to stub functions input_lock/input_unlock so that we don't lose this
information.
xfree86 SIGIO support is reworked to use internal versions of
OsBlockSIGIO and OsReleaseSIGIO.
v2: Don't change locking order (Peter Hutterer)
v3: Comment weird && FALSE in xf86Helper.c
Leave errno save/restore in xf86ReadInput
Squash with stub adding patch (Peter Hutterer)
v4: Leave UseSIGIO config parameter so that
existing config files don't break (Peter Hutterer)
v5: Split a couple of independent patch bits out
of kinput.c (Peter Hutterer)
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Rhys Kidd <rhyskidd@gmail.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
(cherry picked from commit cf8860786c)
The commit message makes the assertion that the code below damage is not
allowed to change whether there's a damage monitor for the drawable.
That turns out not to be the case! exa's mixed code, at least, will
create and destroy a damage in PrepareAccess. The destroy path can then
be catastrophic, as damageRegionProcessPending will attempt to
RegionEmpty memory from the middle of a freed block.
I'd wanted that invariant for performance, but faster isn't worth
broken, so revert it. I think what exa's doing is reasonable, so the
better way to improve performance for the unmonitored case is to either
revisit dynamically wrapping into the GC, or inline damage into dix.
This reverts commit 4e124203f2.
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1389886
Signed-off-by: Adam Jackson <ajax@redhat.com>
(cherry picked from commit 32e632e858)
I think it is possible that output could get queued to a client during
CloseDownClient. After it is removed from the pending queue, active
grabs are released, the client is awoken if sleeping and any work
queue entries related to the client are processed.
To fix this, move the call removing it from the output_pending chain
until after clientGone has been set and then check clientGone in
output_pending_mark.
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1382444
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
posix_fallocate() does an explicit rollback if it gets EINTR, and
this is a problem on slow systems because when the allocation size
is sufficiently large posix_fallocate() will always be interrupted
by the smart scheduler's SIGALRM.
Changes since v1 - big comment in the code to explain what is going on
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Ian Ray <ian.ray@ge.com>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Acked-by: Daniel Stone <daniels@collabora.com>
[This was originally a workaround for a client-side resource leak:
http://lists.freedesktop.org/archives/xorg-devel/2012-November/034555.html
Obviously that's a broken app, but the performance problem it
illustrates - that walking the linked list ends up burning all your CPU
time - is real enough. - ajax]
v2: Replace with a shorter code sequence which computes the same
results for all but numBits == 7
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
If a work proc wakes up a sleeping client and it is ready to execute,
we need to re-compute the local 'are_ready' value before deciding
what timeout value to use in WaitForSomething.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98030
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
If we're never painting anything in the window, we probably don't need
to map it.
v2: Drop ephyr_glamor_gles2 from hostx.c
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
xf86CheckHWCursor() would dereference sPriv without NULL checking it. If Option
"SWCursor" is specified, sPriv == NULL. In this case we should assume that HW
cursors are not supported.
Signed-off-by: Alex Goins <agoins@nvidia.com>
Reviewed-by: Andy Ritger <aritger@nvidia.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
On Linux, setting the main thread's name changes the program name
(/proc/self/comm). Setting it to MainThread breaks scripts that rely on
the command name, e.g. ps -C Xorg.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
On some random condition, a touch event may trigger a crash in Xwayland
in GetTouchEvents().
The (simplified) backtrace goes as follow:
(gdb) bt
#0 GetTouchEvents() at getevents.c:1892
#1 QueueTouchEvents() at getevents.c:1866
#2 xwl_touch_send_event() at xwayland-input.c:652
#5 wl_closure_invoke() from libwayland-client.so.0
#6 dispatch_event() from libwayland-client.so.0
#7 wl_display_dispatch_queue_pending() from libwayland-client.so.0
#8 xwl_read_events() at xwayland.c:483
#9 ospoll_wait() at ospoll.c:412
#10 WaitForSomething() at WaitFor.c:222
#11 Dispatch() at dispatch.c:412
#12 dix_main() at main.c:287
#13 __libc_start_main() at libc-start.c:289
#14 _start ()
The crash occurs when trying to access the sprite associated with the
touch device, which appears to be NULL. Reason being the device itself
is more a keyboard device than a touch device.
Moreover, it appears the device is neither enabled nor activated
(inited=0, enabled=0) which doesn't seem right, but matches the code in
init_touch() from xwayland-input.c which would enable the device if it
was previously existing and otherwise would create the device but not
activate it.
Make sure we do activate and enable touch devices just like we do for
other input devices such as keyboard and pointer.
Signed-off-by: Olivier Fourdan <ofourdan@redhat.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Pointer enter event coordinates are surface relative and we need them to
be screen relative for pScreen->SetCursorPosition().
https://bugzilla.gnome.org/show_bug.cgi?id=758283
Signed-off-by: Rui Matos <tiagomatos@gmail.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Commit c7e8d4a6ee had already unifdef
MODESETTING_OUTPUT_SLAVE_SUPPORT but commit
9257b1252d didn't notice that.
Signed-off-by: Nikhil Mahale <nmahale@nvidia.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Prior to this commit the Xorg.wrap code to detect if root rights are
necessary checked for DRM_IOCTL_MODE_GETRESOURCES succeeding *and*
reporting more then 0 output connectors.
DRM_IOCTL_MODE_GETRESOURCES succeeding alone is enough to differentiate
between old drm only cards (which need ums and thus root) and kms capable
cards.
Some hybrid gfx laptops have 0 output connectors on one of their 2 GPUs,
resulting in Xorg needlessly running as root. This commits removes the
res.count_connectors > 0 check, fixing this.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Fixes DRI2 client driver name mapping for newer AMD GPUs with the
modesetting driver, allowing the DRI2 extension to initialize.
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Add a missing ifdef needed for --disable-glamor.
Signed-off-by: Mihail Konev <k.mvc@ya.ru>
Reviewed-by: Jon Turney <jon.turney@dronecode.org.uk>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Commit 501d8e2b removed --enable-aiglx, but made xwin always be
--enable-glx.
Signed-off-by: Mihail Konev <k.mvc@ya.ru>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Jon Turney <jon.turney@dronecode.org.uk>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Special case for the systemd-logind case in xfree86: when we're vt-switched
away and a device is plugged in, we get a paused fd from logind. Since we
can't probe the device or do anything with it, we store that device in the
xfree86 and handle it later when we vt-switch back. The device is not added to
inputInfo.devices until that time.
When the device is removed while still vt-switched away, the the config system
never notifies the DDX. It only runs through inputInfo.devices and our device
was never added to that.
When a device is plugged in, removed, and plugged in again while vt-switched
away, we have two entries in the xfree86-specific list that refer to the same
device node, both pending for addition later. On VT switch back, the first one
(the already removed one) will be added successfully, the second one (the
still plugged-in one) fails. Since the fd is correct, the device works until
it is removed again. The removed devices' config_info (i.e. the syspath)
doesn't match the actual device we addded tough (the input number increases
with each plug), it doesn't get removed, the fd remains open and we lose track
of the fd count. Plugging the device in again leads to a dead device.
Fix this by adding a call to notify the DDX to purge any remainders of devices
with the given config_info, that's the only identifiable bit we have at this
point.
https://bugs.freedesktop.org/show_bug.cgi?id=97928
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
No functional changes but it makes it easier to remove elements from the
middle of the list (future patch).
We don't have an init call into this file, so the list is manually
initialized.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
They're identically laid-out structs but let's use the right type to search
for our desired value.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
The extension does not exist in the registry, thus needs to know they're
using EGL 1.5 in order to determine the eglGetPlatformDisplay function
pointer is valid.
Thus brings us into some lovely circular dependency.
Since mesa won't be able (in the foreseeable future) to export the KHR
flavour of extension (another way one could assume that EGL 1.5 is
available) just drop all the heuristics and use the
EGL_EXT_platform_base extension.
In practise (checked with the Mali driver) any EGL 1.5 driver will
advertise support for EGL_EXT_platform_base.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
When putting a device node into a poll-request list, do not overwrite a
"please-remove" element with the same fd, so that a closed device file
is ospoll_remove'd prior to being ospoll_add'ed.
Before, the opposite order was possible, resulting in ospoll_add
considering the newly opened file being already polled, should it have a
fd for which the "please-remove" has not been procesed yet. In this
case, no further events would be seen from the device.
Signed-off-by: Mihail Konev <k.mvc@ya.ru>
Regressed-in: 52d6a1e832
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=97880
Patchwork: https://patchwork.freedesktop.org/patch/113763/
Hit-and-Reduced-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-and-Reduced-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
Prevents the HW cursor from intermittently jumping around when the
cursor image is changed while the cursor is being moved. This is hardly
noticeable in normal operation but can be quite confusing when stepping
through these codepaths in a debugger.
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
xf86CursorScreenRec::HotX/Y contain 0 for PRIME slave screens.
Fixes incorrect HW cursor position on PRIME slave screens.
Also hoist the hotspot translation out from xf86ScreenSet/MoveCursor to
xf86Set/MoveCursor, since the hotspot position is a property of the
cursor, not the screen.
v2:
* Squash patches 1 & 2 of the v1 series, since it's basically the same
problem
* Use the master screen's xf86CursorScreenRec::HotX/Y instead of
CursorRec::bits->x/yhot, since CursorRec::bits can be NULL (Hans de
Goede)
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
They're not needed, and they won't be present on win32.
Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Emil Velikov <emil.l.veikov@gmail.com>
