If any part of the stack calls XIGetKnownProperty during device shutdown
the property is re-initialized before the server generation resets, leaving
the value invalid again.
Move the reset to the extension init which happens before input devices are
initialized before the first property is requested.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Julien Cristau <jcristau@debian.org>
(cherry picked from commit 5cd11d2356)
Properties allocated through XIGetKnownProperty() aren't reset on the second
server generation but keep the old value. As a result, wrong Atoms are
supplied to the driver, resulting in potential data corruption or weird
error message.
Reproducible by running "xlsatom | grep FLOAT" twice on a plain X server.
The second X server generation won't have the FLOAT atom defined anymore,
despite the users of this atom not noticing any errors.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Julien Cristau <jcristau@debian.org>
Signed-off-by: Keith Packard <keithp@keithp.com>
(cherry picked from commit 9802839d35)
The modifier key count is maintained by the XKB layer and
increased/decreased for all modifiers that set state.
Test case, MD/SD modifier key count in comment:
1. keyboard 1: press and hold Shift_L # SD:1 MD:1
2. keyboard 2: press and release Shift_L # SD:1,0 MD:1,0
<class copy happens> # SD:1 MD:1
3. keyboard 1: release Shift_L # SD:0 MD:1
4. keyboard 1: press and release Shift_L # SD:1,0 MD:2,1
The modifier is now logically down on the MD but not on keyboard 1 or
keyboard 2.
XKB is layered in before the DIX, it increases/decreases the modifier key
count accordingly. In the above example, during (2), the MD gets the key
release and thus clears the modifier bit. (3) doesn't forward the release to
the MD because it is already cleared. The copy of modifierKeysDown when the
lastSlave changes however increases the counter for the held key. On (4),
the press and release are both forwarded to the MD, causing a offset by 1
and thus do not clear the logical modifier state.
X.Org Bug 25480 <http://bugs.freedesktop.org/show_bug.cgi?id=25480>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Acked-by: Daniel Stone <daniel@fooishbar.org>
(cherry picked from commit dc614484f9)
Reintroduce a check which used to be there in the old
ProcessKeyboardEvent/ProcessPointerEvent codepath, which avoids us
recording events subject to a grab twice: once when it's first processed
in EnqueueEvent, and then again when it's thawed and being replayed.
This required a tiny amount of code motion to expose syncEvents.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
(cherry picked from commit a2ea8c2f2c)
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The XI protocol spec only allows for two errors on the SetDeviceMode
requests: BadMatch or BadMode. BadMode however is a dynamically assigned
extension error and the driver doesn't have access to the actual error
number. Hence, if a SetDeviceMode driver returns an error other than
BadMatch, assume BadMode.
The two exceptions are BadAlloc and BadImplementations, pass these on to the
client (any request is allowed to return either of those).
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Dan Nicholson <dbn.lists@gmail.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
(cherry picked from commit e40ba5798c)
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
RECORD was disabled during the switch to internal events. This patch
modifies the record callback to work with internal events instead of
xEvents. The InternalEvents are converted to core/Xi events as needed.
Since record is a loadable extension, the EventTo* calls must be externed.
Signed-off-by: Chris Dekter <cdekter@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
(cherry picked from commit 122fc0e7a0)
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
If the indicator flags have the XkbSLI_IsDefault bit set, the indicator map
and names aren't their own bit of memory but rather point into the
device->key->xkbInfo->desc structure. XkbCopySrvLedInfo knows about this and
leaves the pointers alone.
When copying the classes from the slave to the master, these pointers are
copied and still point to the dev->key class of the slave device. If the
slave device is removed, the memory becomes invalid and a call to modify
this data (e.g. XkbSetIndicators) may cause a deadlock.
The copying of dev->key relies on dev->kbdfeed to be already set up. Hence
the pointers need to be reset once _both_ kbdfeed and key have been copied
into the master device.
X.Org Bug 25640 <http://bugs.freedesktop.org/show_bug.cgi?id=25640>
Fedora Bug 540584 <https://bugzilla.redhat.com/show_bug.cgi?id=540584>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Keith Packard <keithp@keithp.com>
(cherry picked from commit 48f7298657)
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Commit 0e6cee853d introduced cleanup code to
remove the accel properties when switching schemes. The same code is
triggered by the default closedown code but only after unconditionally
removing all device properties (as part of the cleanup). The properties,
although deleted never got reset to NULL.
X.Org Bug 25374 <http://bugs.freedesktop.org/show_bug.cgi?id=25374>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Tested-by: Andy Furniss <lists@andyfurniss.entadsl.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
(cherry picked from commit ea14a41821)
Trying to unregister property handlers during the device closure process
leads to invalid memory accesses.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Keith Packard <keithp@keithp.com>
(cherry picked from commit 1b127ab842)
Deleting a property that was not set on a device leads to a null-pointer
reference. The protocol allows deleting those properties - it has to be a
noop.
Reproducible:
xinput --set-prop "My device" --type=int --format=8 "my property" 1
xinput --delete-prop "My other device" "my property"
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Keith Packard <keithp@keithp.com>
(cherry picked from commit a30e739a14)
Instead of returning BadAccess when "read" permission is denied
on a device, falsify the device state (buttons down, keys pressed).
This is nicer to applications, but may still have undesired side
effects. The long-term solution is not to use these requests in
event-driven code!
Requests affected: QueryPointer, QueryKeymap, XiQueryDevice.
Signed-off-by: Eamon Walsh <ewalsh@tycho.nsa.gov>
(cherry picked from commit 8502c06e19)
Subpixel data in data_frac is stored as FP32.32, hence we need to get that
down again before adding it to the current value.
Reported-by: Thomas Jaeger
Tested-by: Thomas Jaeger
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
(cherry picked from commit 9bd08c690f)
The length field needs to include the bytes required for the valuators
(INT32) as well.
The reply length has the right value and since the valuator state is always
last, clients didn't notice the wrong offset.
Tested-by: Thomas Jaeger
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
(cherry picked from commit 6ee796e9bb)
include/protocol-versions.h specifies each extension version as supported by
the server and sent back on the wire to the client.
This fixes up several issues with the server potentially reporting a higher
version of the protocol if recompiled against a newer version of the
protocol.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Acked-by: Rémi Cardona <remi@gentoo.org>
Acked-by: Julien Cristau <jcristau@debian.org>
For core and XI1 events, store the key_repeat flag in the sequence number
until TryClientEvents. The sequenceNumber is unset until TryClientEvents.
[Also thrown in, some random indentation changes. Thanks]
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The pointer advanced 12 bytes too short. Rather unfortunate if both the code
and the test have the same bug.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Bits above XI2LASTEVENT are invalid and cause in BadValues. These checks
must be performed anywhere where a mask_len parameter is given.
This patch also adds the missing checks to grab masks.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
A grab mask provided in the request may be larger than the one used in the
server. Cut down to size before memcopying.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Fixes crash if the first XISelectEvents has a zero sized event mask.
Fixes crash if the mask provided is larger than others->xi2mask[].
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
masks can be of arbitrary length. If the client did not initialize mask_len,
some sort of boundary check is needed to avoid running over memory.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Note that the Xi events are critical and should thus cause a flush to
the client when an input event is pending.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
A missing break meant that ButtonPress would fall through into
ButtonRelease, but luckily it appears to have been completely harmless.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
This patch corrects a misnaming of XTest-related functions.
The extension itself announces itself as XTEST. Xtst is the library name
itself, but all library functions are prefixed by XTest. Same with the
naming in the server.
- Rename all *Xtst* functions to *XTest* for consistency with the library
and in-server API.
- Rename the "Xtst device" property to "XTEST device" for consistency with
the extension naming.
- Rename the device naming to "<master device name> XTEST device". The
default xtest devices become "Virtual core XTEST pointer" and "Virtual
core XTEST keyboard".
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The extern declaration in xichangehierarchy.c was broken anyway.
This fixes a crash on creating a new master device.
Reported-by: Maxim Levitsky
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
buttons_size is necessary as WriteToClient swaps the buttons_len field,
resulting in the wrong number of bytes being written later.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
DCEs are now processed when sent throught the master device, not when sent
through the slave device. This includes a removal of some un-used (or partly
used) fields in the DCE itself to something more self-explanatory.
TODO: if a device has events queued and its attachment is changed, the DCE
is silently dropped now. Instead, it should be generated as soon as the
first event after the attachment is sent.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The version in eventconvert.c was half broken and for some reason we ended
up with a second version in exevents.c (which works). Move it over to where
it belongs and call EventToXI2 instad of having a custom function for it.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
If a new device posts an event while the DCE is in the queue, getting the
data from the device may result in invalid memory access.
X.Org Bug 23100 <http://bugs.freedesktop.org/show_bug.cgi?id=23100>
Unlike Enter/Leave events generated by a device pushing the pointer around,
a device doesn't change focus all by itself. It's a result of a
SetInputFocus call, a window becoming unviewable or a grab activating. As
such, the sourceid for focus events is always the deviceid itself.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Use enum EventType instead of ints. This requires a load of default
cases in various switch statements to silence compiler warnings.
Reported-by: Aaron Plattner
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
