A disabled device doesn't have a sprite (less so a sprite->win) and triggers
a NULL-pointer dereference on shutdown when all active grabs are released as
part of the cleanup.
Fix this by checking for sprite being non-null and setting the focus window
to the NullWindow if it is. The rest of the patch just attempts to make
things more readable.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Keith Packard <keithp@keithp.com>
If the device is disabled, the sprite window is NULL and dereferencing
crashes the server.
This is only triggered for XI 1.x grabs (ProcXGrabDevice) as XI2 grabs would
trigger another code path, creating a sprite for the disabled device as if
detaching it (which is wrong and fixed with this patch too).
Grabbing a disabled device doesn't make sense as it won't send events
anyway. However, the protocol specs do not prohibit it, so we need to keep
it working.
Luckily, oldWin is only used for focus out events, which aren't necessary
given that the device is disabled.
X.Org Bug 54934 <http://bugs.freedesktop.org/show_bug.cgi?id=54934>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Chase Douglas <chase.douglas@ubuntu.com>
Scale_to_desktop() converts ABS events from device coordinates
to screen coordinates:
[dev_X_min, dev_X_max] -> [screen_X_min, screen_X_max]
[dev_Y_min, dev_Y_max] -> [screen_Y_min, screen_Y_max]
An edge ABS event with X = dev_X_max (e.g., generated from the
edge of a touchscreen) will be converted to have screen X value
= screen_X_max, which, however, will be filterd out when xserver
tries to find proper Window to receive the event, because the
range check for a Window to receive events is
window_X_min <= event_screen_X < window_X_max
Events with event_screen_X = screen_X_max will fail the test get
and rejected by the Window.
To fix this, we change the device to screen coordinates mapping to
[dev_X_min, dev_X_max] -> [screen_X_min, screen_X_max-1]
[dev_Y_min, dev_Y_max] -> [screen_Y_min, screen_Y_max-1]
Reviewed-by: Chase Douglas <chase.douglas@canonical.com>
Reviewed-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
Signed-off-by: Yufeng Shen <miletus@chromium.org>
Signed-off-by: Keith Packard <keithp@keithp.com>
This hack was added to suppress events generated by Composite's internal
unmap/map cycle on redirection state change. Since that cycle was
removed in 193ecc8b4, these can go.
Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Ville Syrjälä <syrjala@sci.fi>
Signed-off-by: Keith Packard <keithp@keithp.com>
The property handler is registered after setting the property, so
dev->transform remains as all-zeros. That causes pixman_f_transform_invert()
to fail (in transformAbsolute()) and invert remains as garbage. This
may then cause a cursor jump to 0,0.
Since the axes are not yet initialized here and we need to allow for drivers
changing the matrix, we cannot use the property handler for matrix
initialization, essentially duplicating the code.
Triggered by the fix to (#49347) in 749a593e49https://bugzilla.redhat.com/show_bug.cgi?id=852841
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Chase Douglas <chase.douglas@ubuntu.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
Forwarding proxies like sshd will appear to be local, even though they
aren't really. This leads to weird behaviour for extensions that truly
require running under the same OS services as the client, like MIT-SHM
and DRI2.
Add two new legal values for the initial connection's byteOrder field,
'r' and 'R'. These act like 'l' and 'B' respectively, but have the side
effect of forcing the client to be treated as non-local. Forwarding
proxies should attempt to munge the first packet of the connection
accordingly; older servers will reject connections thusly munged, so the
proxy should fall back to passthrough if the munged connection attempt
fails.
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Adam Jackson <ajax@redhat.com>
If we don't free this here, it gets freed later in the resource
cleanups, however it then looks up up pmap->pScreen, which we
freed already in this function. So free the default colormap
when we should.
This fixes a bug after a couple of hotplug cycles when you try
to exit the X server and it crashes.
Reviewed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
In WarpPointer calls, we get input in screen coordinates. They must be
scaled to device coordinates, and then back to screen coordinates for screen
crossing and root coordinates in events.
The rounding errors introduced (and clipping in core/XI 1.x events) can lead
to the actual position being different to the requested input coordinates.
e.g. 200 scales to 199.9999, truncated to 199 in the event.
Avoid this by simply overwriting the scaled screen coordinates with the
input coordinates for the POINTER_SCREEN case.
X.Org Bug 53037 <http://bugs.freedesktop.org/show_bug.cgi?id=53037>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Keith Packard <keithp@keithp.com>
This should make cleaning up the GPU screens easier as the core
screens they are associated with will still be around.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
If a device has smooth scrolling axes, but submits scroll button events, we
convert those to motion events and update the valuators. For legacy button
events, the valuator mask is likely unset though, causing
add_to_scroll_valuator() to return early, leaving us with an empty mask.
That again skipped the rest of the code and no events were generated.
Fix it by making sure that the scroll valuator in the mask is at least
initialized to 0.
Broke evdev wheel emulation, introduced by
54476b5e44.
X.Org Bug 52508 <http://bugs.freedesktop.org/show_bug.cgi?id=52508>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Chase Douglas <chase.douglas@canonical.com>
Since the free routines free the master pixmap then the slave, we should
be taking a reference when we bind them together.
Fixes a use-after-free when resizing a primed gears.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
The indenter seems to have gotten confused by initializing arrays of
structs with the struct defined inline - for predefined structs it did
a better job, so match that.
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Since we call directly into XKB and may be doing so before the extension
has been initialised, make sure its privates are set up first. XTest
had a hack to do this itself, but seems cleaner to just make sure we do
it in AllocDevicePair.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Keith Packard <keithp@keithp.com>
In:
commit d792ac125a
Author: Alan Coopersmith <alan.coopersmith@oracle.com>
Date: Mon Jul 9 19:12:43 2012 -0700
Use C99 designated initializers in dix Replies
the initializer for the .length element of the xGetPointerMappingReply
structure uses the value of rep.nElts, but that won't be set until
after this initializer runs, so we get garbage in the length element
and clients using it will generally wedge.
Easy to verify:
$ xmodmap -pp
Fixed by creating a local nElts variable and using that.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Adds new function padding_for_int32() and uses existing pad_to_int32()
depending on required results.
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Tested-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Tested-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Tested-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Tested-by: Daniel Stone <daniel@fooishbar.org>
Let the compiler worry about 0-filling the rest of the fields,
instead of memsetting the whole struct and then going back to
overwrite some of the fields.
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Tested-by: Daniel Stone <daniel@fooishbar.org>
When passing variable pointers to functions or otherwise doing long
sequences to compute values for replies, create & use some new
temporary variables, to allow for simpler initialization of reply
structures in the following patches.
Move memsets & other initializations to group with the rest of the
filling in of the reply structure, now that they're not needed so
early in the code path.
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Tested-by: Daniel Stone <daniel@fooishbar.org>
Always initialize to zero, and then if permission is granted, copy
the current key state maps, instead of always copying and then
zeroing out if permission was denied.
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Tested-by: Daniel Stone <daniel@fooishbar.org>
Always initialize to zero, and then if permission is granted, copy
the current key state maps. Use memcpy instead of memmove for the
copy, since we're always copying to a newly allocated event on the
stack, so guaranteed not to overlap with the device map structure.
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Tested-by: Daniel Stone <daniel@fooishbar.org>
Don't need to pass an empty reply to NullPropertyReply, let it make
it's own. Move reply initialization code in remaining replies in
ProcGetProperty to group with the rest of the fields. (Prepares
for coming C99 designated initializer conversion.)
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Tested-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Tested-by: Daniel Stone <daniel@fooishbar.org>
Casting return to (void) was used to tell lint that you intended
to ignore the return value, so it didn't warn you about it.
Casting the third argument to (char *) was used as the most generic
pointer type in the days before compilers supported C89 (void *)
(except for a couple places it's used for byte-sized pointer math).
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Tested-by: Daniel Stone <daniel@fooishbar.org>
When allocating new global privates, make sure the gpu
screens get their private offsets updated.
This only affects GPU screens that enumerate before the non-GPU
screens, which generally requires that the related device be present
when the system boots so that it can get an earlier DRM filename.
Signed-off-by: Keith Packard <keithp@keithp.com>
add the linked list and provider hooks.
v1.1: add another assert in the add path.
Reviewed-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This adds two functions for drivers to use directly to keep a
linked list of slave pixmaps to do damage tracking on and keep
updated. It also adds a helper function that drivers may optionally
call to do a simple copy area damage update.
v2: use damage.h not damagestr.h, fixes ephyr build.
v3: address ajax review: use slave_dst, drop unused dst member.
v4: check DamageCreate return, add SourceValidate comment,
add a comment addressing possible optimisation possibility
Reviewed-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
When randr notices a crtc configuration request for a slave device,
it checks if the slave allocated pixmap exists and is suitable,
if not it allocates a new shared pixmap from the master, shares
it to the slave, and starts the master tracking damage to it,
to keep it updated from the current front pixmap.
If the resize means the crtc is no longer used it will destroy
the slave pixmap.
This adds the concept of a scanout_pixmap to the randr_crtc object,
and also adds a master pixmap pointer to the pixmap object, along
with defining some pixmap helper functions for getting pixmap box/regions.
v2: split out pixmap sharing to a separate function.
v3: update for void *
Reviewed-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Just add the interfaces to attach/detach output slaves, and
a linked list to keep track of them. Hook up the randr providers
list to include these slaves.
v1.1: add another assert to the add path.
Reviewed-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This list is meant for attaching unbound gpu screens to initially,
before the client side rebinds them.
v1.1: add another assert in the add path.
Reviewed-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This patch introduces gpu screens into screenInfo. It adds interfaces
for adding and removing gpu screens, along with adding private fixup,
block handler support, and scratch pixmap init.
GPU screens have a myNum that is offset by GPU_SCREEN_OFFSET (256),
this is used for logging etc.
RemoveGPUScreen isn't used until "xfree86: add platform bus hotplug support".
v2: no glyph pictures for GPU screens for now.
v3: introduce MAXGPUSCREENS, fix return value check
v4: fixup myNum when renumbering screens (ajax)
v5: drop cursor privates for now.
Reviewed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This is a precursor for reusing this code to init gpu screens.
v2: fixup int check as per Keith's review.
Reviewed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This will permit midispcur to allocate its privates for hotplug outputs
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Screen-specific privates areas are only allocated for objects related
to the target screen; objects allocated for other screens will not
have the private space reserved. This saves memory in these objects
while also allowing hot-plug screens to have additional private
allocation space beyond what the core screens are using.
Drivers are encouraged to switch to this mechanism as it will reduce
memory usage in multi-GPU environments, but it is only required for
drivers which will be loaded after the server starts, like
modesetting.
Objects providing screen-specific privates *must* be managed by the
screen-specific private API when allocating or initializing privates
so that the per-screen area can be initialized properly.
The objects which support screen-specific privates are:
Windows
Pixmaps
GCs
Pictures
Extending this list to include Colormaps would be possible, but
require slightly more work as the default colormap is created before
all colormap privates are allocated during server startup, and hence
gets a bunch of special treatment.
Of particular note, glyphs are *not* capable of supporting
screen-specific privates as they are global objects, not allocated on
a screen-specific basis, and so each driver must be able to see their
privates within the glyph.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Don't leak if ti->history is NULL.
Found by coverity.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Dave Airlie <airlied@redhat.com>
This merge includes a minor fixup for '%p' arguments; must cast to
uintptr_t instead of uint64_t as we use -Werror=pointer-to-int-cast
which complains when doing a cast (even explicitly) from a pointer
to an integer of different size.
