| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
An spi_driver does not need to set an owner, it will be populated by the
driver core.
Signed-off-by: Andrew F. Davis <[email protected]>
Acked-by: Jonathan Cameron <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
|
|
Hardware manufacturers group keys in the weirdest way possible. This may
cause a power-key to be grouped together with normal keyboard keys and
thus be reported on the same kernel interface.
However, user-space is often only interested in specific sets of events.
For instance, daemons dealing with system-reboot (like systemd-logind)
listen for KEY_POWER, but are not interested in any main keyboard keys.
Usually, power keys are reported via separate interfaces, however,
some i8042 boards report it in the AT matrix. To avoid waking up those
system daemons on each key-press, we had two ideas:
- split off KEY_POWER into a separate interface unconditionally
- allow filtering a specific set of events on evdev FDs
Splitting of KEY_POWER is a rather weird way to deal with this and may
break backwards-compatibility. It is also specific to KEY_POWER and might
be required for other stuff, too. Moreover, we might end up with a huge
set of input-devices just to have them properly split.
Hence, this patchset implements the second idea: An event-mask to specify
which events you're interested in. Two ioctls allow setting this mask for
each event-type. If not set, all events are reported. The type==0 entry is
used same as in EVIOCGBIT to set the actual EV_* mask of filtered events.
This way, you have a two-level filter.
We are heavily forward-compatible to new event-types and event-codes. So
new user-space will be able to run on an old kernel which doesn't know the
given event-codes or event-types.
Signed-off-by: David Herrmann <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
Remove the unneded semicolon since it is clearly a typo error.
Signed-off-by: Javier Martinez Canillas <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
button bits
commit 92bac83dd79e ("Input: alps - non interleaved V2 dualpoint has
separate stick button bits") assumes that all alps v2 non-interleaved
dual point setups have the separate stick button bits.
Later we limited this to Dell laptops only because of reports that this
broke things on non Dell laptops. Now it turns out that this breaks things
on the Dell Latitude D600 too. So it seems that only the Dell Latitude
D420/430/620/630, which all share the same touchpad / stick combo,
have these separate bits.
This patch limits the checking of the separate bits to only these models
fixing regressions with other models.
Reported-and-tested-by: Larry Finger <[email protected]>
Cc: [email protected]
Tested-by: Hans de Goede <[email protected]>
Signed-off-by: Hans de Goede <[email protected]>
Acked-By: Pali Rohár <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
The driver is now used for the entire USRP e3xx series,
this commit fixes the description that will be displayed in
the menu accordingly.
Signed-off-by: Moritz Fischer <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
The driver handles wakeup irq correctly using device_init_wakeup and
enable_irq_wake. There's no need to use IRQF_NO_SUSPEND while registering
the interrupt.
This patch removes the use of IRQF_NO_SUSPEND flag.
Signed-off-by: Sudeep Holla <[email protected]>
Acked-by: Thierry Reding <[email protected]>
Acked-by: Laxman Dewangan <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
Though the tegra-kbc driver should/will continue to support the legacy
"nvidia,wakeup-source" property to enable keyboard as the wakeup source,
we need to add support for the new standard property "wakeup-source".
This patch adds support for "wakeup-source" property in addition to the
existing "nvidia,wakeup-source" property.
Signed-off-by: Sudeep Holla <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
Internally, xenbus_printf uses memory allocation, so it can fail under
memory pressure, leaving the input device configured as absolute with the
backend supplying relative coordinates.
Signed-off-by: Insu Yun <[email protected]>
Reviewed-by: Julien Grall <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
hp_sdc_rtc_proc_show() use timeval to store the time, which will overflow
in 2038.
This patch fixes this problem by replacing timeval with timespec64.
hp_sdc_rtc_proc_show() only output string, so that userspace will work
normally if we apply this patch.
Not all timer in i8042 have y2038 risk(handshake, match timer, etc),
Replacements in those timer are just for consistency.
Signed-off-by: WEN Pingbo <[email protected]>
Reviewed-by: Arnd Bergmann <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
s/regsiter/register/
Signed-off-by: Geliang Tang <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
Replace struct v4l2_format * with void * to make queue_setup()
for common use.
And then, modify all device drivers related with this change.
Signed-off-by: Junghak Sung <[email protected]>
Signed-off-by: Geunyoung Kim <[email protected]>
Acked-by: Seung-Woo Kim <[email protected]>
Acked-by: Inki Dae <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
[[email protected]: fix missing const in fimc-lite.c]
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
|
|
Do not call xpad_identify_controller at init with wireless devices: it
conflicts with the already sent presence packet and will be called by
xpad360w_process_packet as needed anyway.
Signed-off-by: Pavel Rojtberg <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
Fix build errors due to missing Kconfig dependency.
drivers/built-in.o: In function `sur40_disconnect':
sur40.c:(.text+0x22be6e): undefined reference to `video_unregister_device'
sur40.c:(.text+0x22be77): undefined reference to `v4l2_device_unregister'
drivers/built-in.o: In function `sur40_process_video':
sur40.c:(.text+0x22c1d4): undefined reference to `v4l2_get_timestamp'
drivers/built-in.o: In function `sur40_probe':
sur40.c:(.text+0x22ca82): undefined reference to `v4l2_device_register'
sur40.c:(.text+0x22cb1a): undefined reference to `v4l2_device_unregister'
sur40.c:(.text+0x22cbf7): undefined reference to `video_device_release_empty'
sur40.c:(.text+0x22cc53): undefined reference to `__video_register_device'
sur40.c:(.text+0x22cc90): undefined reference to `video_unregister_device'
drivers/built-in.o: In function `sur40_vidioc_querycap':
sur40.c:(.text+0x22ccb0): undefined reference to `video_devdata'
Signed-off-by: Randy Dunlap <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
If common clock framework is configured, the driver generates a warning,
which is fixed by this change:
root@devkit3250:~# cat /dev/input/touchscreen0
------------[ cut here ]------------
WARNING: CPU: 0 PID: 720 at drivers/clk/clk.c:727 clk_core_enable+0x2c/0xa4()
Modules linked in: sc16is7xx snd_soc_uda1380
CPU: 0 PID: 720 Comm: cat Not tainted 4.3.0-rc2+ #199
Hardware name: LPC32XX SoC (Flattened Device Tree)
Backtrace:
[<>] (dump_backtrace) from [<>] (show_stack+0x18/0x1c)
[<>] (show_stack) from [<>] (dump_stack+0x20/0x28)
[<>] (dump_stack) from [<>] (warn_slowpath_common+0x90/0xb8)
[<>] (warn_slowpath_common) from [<>] (warn_slowpath_null+0x24/0x2c)
[<>] (warn_slowpath_null) from [<>] (clk_core_enable+0x2c/0xa4)
[<>] (clk_core_enable) from [<>] (clk_enable+0x24/0x38)
[<>] (clk_enable) from [<>] (lpc32xx_setup_tsc+0x18/0xa0)
[<>] (lpc32xx_setup_tsc) from [<>] (lpc32xx_ts_open+0x14/0x1c)
[<>] (lpc32xx_ts_open) from [<>] (input_open_device+0x74/0xb0)
[<>] (input_open_device) from [<>] (evdev_open+0x110/0x16c)
[<>] (evdev_open) from [<>] (chrdev_open+0x1b4/0x1dc)
[<>] (chrdev_open) from [<>] (do_dentry_open+0x1dc/0x2f4)
[<>] (do_dentry_open) from [<>] (vfs_open+0x6c/0x70)
[<>] (vfs_open) from [<>] (path_openat+0xb4c/0xddc)
[<>] (path_openat) from [<>] (do_filp_open+0x40/0x8c)
[<>] (do_filp_open) from [<>] (do_sys_open+0x124/0x1c4)
[<>] (do_sys_open) from [<>] (SyS_open+0x2c/0x30)
[<>] (SyS_open) from [<>] (ret_fast_syscall+0x0/0x38)
Signed-off-by: Vladimir Zapolskiy <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Pull input fixes from Dmitry Torokhov:
"Just two small fixups to ads7846 touchscreen controller driver and
Cypress touchpad driver"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
Input: cyapa - fix the copy paste error on electrodes_rx value
Input: ads7846 - correct the value got from SPI
|
|
In the touchscreen controller ISR, reading the tsc starting from
register 0x2 causes the tsc to infrequently update the detected
finger's x and y coordinate. The irq pin toggles at a fast rate to
indicate touch events are happening. However, the tsc on average
updates the touch point's x and y value every ~100 ms which is much
slower than the advertised rate of 100+ Hz. This leads to multiple reads
within this ~100 ms time window returning the same value.
Example:
X: 10 , Y: 30
X: 10 , Y: 30
X: 10, Y: 30
..
// After 100 ms
X: 300, Y: 300
X: 300, y: 300
..
// After 100 ms
X: 1743, Y: 621
X: 1743, Y: 621
For some reason if instead of starting to read at register 0x2 you
start reading at register 0x0 this issue isn't seen. This seems like
a quirk only seen in the EDT FT5506 so to fix this issue simply
adjust the code to start reading from 0x0. Technically this isn't wrong
so no regressions should be seen with other touchscreen controllers
supported by this driver.
Signed-off-by: Franklin S Cooper Jr <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
FT5506 is essentially the same as other FT5x06 devices other than
supporting 10 support points.
Signed-off-by: Franklin S Cooper Jr <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
Update the code so that the maximum supported points aren't hard coded but
can be changed.
Set the maximum support points based on the data passed along side the
compatible field.
Signed-off-by: Franklin S Cooper Jr <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
Calculate the amount of data that needs to be read for the specified max
number of support points. If the maximum number of support points changes
then the amount that is read from the touch screen controller should
reflect this.
Signed-off-by: Franklin S Cooper Jr <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
Some encoders have both outputs low in stable states, others also have
a stable state with both outputs high (half-period mode) and some have
a stable state in all steps (quarter-period mode). The driver used to
support the former states and with this change it can also support the
later.
This commit also deprecates the 'half-period' property and introduces
a new property 'steps-per-period'. This property specifies the
number of steps (stable states) produced by the rotary encoder
for each GPIO period.
Signed-off-by: Guido Martínez <[email protected]>
Signed-off-by: Ezequiel Garcia <[email protected]>
Acked-by: Rob Herring <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
This commit makes uses of_property_read_bool() to read
boolean properties. This is just cosmetic cleanup.
Signed-off-by: Ben Gamari <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
There is an undocumented upper bound for the total number of ff effects:
FF_GAIN (= 96).
This can be found as follows:
- user: write(EV_FF, effect_id, iterations)
calls kernel: ff->playback(effect_id, ...): starts effect "effect_id"
- user: write(EV_FF, FF_GAIN, gain)
calls kernel: ff->set_gain(gain, ...): sets gain
A collision occurs when effect_id equals FF_GAIN.
According to input_ff_event(),
FF_GAIN is the smallest value where a collision occurs.
Therefore the greatest safe value for effect_id is FF_GAIN - 1,
and thus the total number of effects should never exceed FF_GAIN.
Define FF_MAX_EFFECTS as FF_GAIN and check on this limit in ff-core.
Signed-off-by: Elias Vanderstuyft <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
This changAdd support for EV_ABS / EV_REL events to the gpio-keys-polled driver.
The driver already allows specifying what type of events (key / rel / abs)
a button generates when pressed, but for rel / abs axis we also need to
specify which value this specific gpio represents.
One use case is digital joysticks / direction-pads which are hooked up to
gpio, in this case we've left and right buttons which we want to map to
EV_ABS, ABS_X and we want generate events for left with a value of -1 and
for right with a value of +1 (and similar for up / down and ABS_Y).
Signed-off-by: Hans de Goede <[email protected]>
Acked-by: Rob Herring <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
With commit 3fffd1283927 ("i2c: allow specifying separate wakeup
interrupt in device tree") wakeirq is managed by i2c-core, so remove
wakeirq related code from pixcir_i2c_ts driver.
Signed-off-by: Vignesh R <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
This patch adds wake up support to GPIO rotary encoders.
Signed-off-by: Sylvain Rochet <[email protected]>
Reviewed-by: Johan Hovold <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
Add new function input_enable_softrepeat() that allows drivers to
initialize their own values for input_dev->rep[REP_DELAY] and
input_dev->rep[REP_PERIOD], but also use the software autorepeat
functionality from input.c.
For example, a HID driver could do:
static void xyz_input_configured(struct hid_device *hid,
struct hid_input *hidinput)
{
input_enable_softrepeat(hidinput->input, 400, 100);
}
static struct hid_driver xyz_driver = {
.input_configured = xyz_input_configured,
}
Signed-off-by: Petri Gynther <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
Fix the copy paste error on the electrodes_rx value set code which will
cause the electrodes_rx value be always set to the value of electrodes_y.
Reported-by: Sudip Mukherjee <[email protected]>
Signed-off-by: Dudley Du <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
If the upcoming system suspend is not going to be handled by the
platform firmware, like in the suspend-to-idle case, it is not
necessary to reset the controller in i8042_pm_suspend(), so avoid
doing that.
Moreover, if the system resume currently in progress has not been
started by the platform firmware, like in the suspend-to-idle case,
i8042_controller_resume() need not be called by i8042_pm_resume(),
so avoid doing that too in that case.
Additionally, try to catch the event that woke up the system by
calling the interrupt handler early during system resume if it has
not been started by the platform firmware.
Signed-off-by: Rafael J. Wysocki <[email protected]>
Acked-by: Dmitry Torokhov <[email protected]>
|
|
When we initialize the driver/device, we really don't know how many
controllers are connected. So send a "query presence" command to the
base-station. (Command discovered by Zachary Lund)
Presence packet taken from:
https://github.com/computerquip/xpad5
Signed-off-by: Pavel Rojtberg <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
To allow us to later create / destroy the input device from the urb
callback, we need to initialize/ deinitialize the input device from a
separate function. So pull that logic out now to make later patches
more "obvious" as to what they do.
Signed-off-by: "Pierre-Loup A. Griffais" <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>
Signed-off-by: Pavel Rojtberg <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
as discussed here[0], x360w is the only pad that maps dpad_to_button.
This is bad for downstream developers as they have to differ between
x360 and x360w which is not intuitive.
This patch implements the suggested solution of exposing the dpad both
as axes and as buttons. This retains backward compatibility with software
already dealing with the difference while makes new software work as
expected across x360/ x360w pads.
[0] http://www.spinics.net/lists/linux-input/msg34421.html
Signed-off-by: Pavel Rojtberg <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
Move submission logic to a single point at the end of the function.
This makes it easy to add locking/ queuing code later on.
Signed-off-by: Pavel Rojtberg <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
This code was probably wrong ever since and is redundant with
xpad_send_led_command. Both try to send a similar command to the xbox360
controller. However xpad_send_led_command correctly uses the pad_nr instead
of bInterfaceNumber to select the led and re-uses the irq_out URB instead
of creating a new one.
Note that this change only affects the two supported wireless controllers.
Tested using the xbox360 wireless controller (PC).
Signed-off-by: Pavel Rojtberg <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
The pad_nr corresponds to the lit up LED on the controller. Therefore there
should be no gaps when enumerating. Currently a LED is only re-assigned
after a controller is re-connected 4 times.
This patch uses ida to track connected pads - this way we can re-assign
freed up pad number immediately.
Consider the following case:
1. pad A is connected and gets pad_nr = 0
2. pad B is connected and gets pad_nr = 1
3. pad A is disconnected
4. pad A is connected again
using ida_simple_get() controller A now correctly gets pad_nr = 0 again.
Signed-off-by: Pavel Rojtberg <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
Rename led_no -> pad_nr: the number stored there is not the LED number - it
gets translated later on to a LED number in xpad_identify_controller;
Signed-off-by: Pavel Rojtberg <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
The "Razer Atrox Arcade Stick" features 10 buttons, and two of them (LT/RT)
don't work properly. Change its definition in xpad_device[] (mapping
field) to fix.
Signed-off-by: Dario Scarpa <[email protected]>
Signed-off-by: Pavel Rojtberg <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
It is identical to the Xbox One controller but has a different product ID.
Signed-off-by: Pavel Rojtberg <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
The copyright/license notice says that the code is licensed under GPL v2
only (not GPL v2+), so let's use proper string in MODULE_LICENSE().
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
According to the touch controller spec, SPI return a 16 bit value, only 12
bits are valid, they are bit[14-3].
The value of MISO and MOSI can be configured when SPI is in idle mode.
Currently this touch driver assumes the SPI bus sets the MOSI and MISO in
low level when SPI bus is in idle mode. So the bit[15] of the value got
from SPI bus is always 0. But when SPI bus congfigures the MOSI and MISO in
high level during the SPI idle mode, the bit[15] of the value get from SPI
is always 1. If bit[15] is not masked, we may get the wrong value.
Mask the invalid bit to make sure the correct value gets returned.
Regardless of the SPI bus idle configuration.
Signed-off-by: Andrey Gelman <[email protected]>
Signed-off-by: Haibo Chen <[email protected]>
Signed-off-by: Igor Grinberg <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
There is no need to explicitly set .owner for the driver, the core will do
it for us.
Signed-off-by: Fengguang Wu <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
Since reset and wake pin are optional the gpio structure for those
pins may be null. Therefore, they can't be blindly passed to
desc_to_gpio.
Reported-by: Dan Carpenter <[email protected]>
Signed-off-by: Franklin S Cooper Jr <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
Commit 5702222c9a7a ("Input: joydev - use memdup_user() to duplicate
memory from user-space") changed the kmalloc() and copy_from_user()
with a single call to memdup_user() but wrongly used the same error
path than the old code in which the buffer allocated by kmalloc() was
freed if copy_from_user() failed.
This is of course wrong since if memdup_user() fails, no memory was
allocated and the error in the error-valued pointer should be returned.
Fixes: 5702222c9a7a ("Input: joydev - use memdup_user() to duplicate
memory from user-space")
Reported-by: Dan Carpenter <[email protected]>
Signed-off-by: Javier Martinez Canillas <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
This adds support for ROHM BU21023/24 Dual touch resistive touchscreens.
Signed-off-by: Yoichi Yuasa <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
Modify parkbd driver to use the new Parallel Port device model.
Signed-off-by: Sudip Mukherjee <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
Merge with mainline to sync up with changes to parkbd driver.
|
|
The variable i is used to check the port to attach to and we are
supposed to save the reference of struct tgfx in the location given by
tgfx_base[i]. But after finding out the index, i is getting modified
again so we saved in a wrong index.
Fixes: 4de27a638a99 ("Input: turbografx - use parallel port device model")
Signed-off-by: Sudip Mukherjee <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
The variable i is used to check the port to attach to and we are
supposed to save the reference of struct gc in the location given by
gc_base[i]. But after finding out the index, i is getting modified again
so we saved in a wrong index.
Reported-by: Dan Carpenter <[email protected]>
Fixes: a517e87c3dfc ("Input: gamecon - use parallel port device model")
Signed-off-by: Sudip Mukherjee <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
The variable i is used to check the port to attach to and we are
supposed to save the reference of struct db9 in the location given by
db9_base[i]. But after finding out the index, i is getting modified again
so we saved in a wrong index.
While at it mark db9_base[i] as NULL after it is freed.
Fixes: 2260c419b52b ("Input: db9 - use parallel port device model")
Signed-off-by: Sudip Mukherjee <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Pull input layer fixes from Dmitry Torokhov:
"Fixes for two recent regressions (in Synaptics PS/2 and uinput
drivers) and some more driver fixups"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
Revert "Input: synaptics - fix handling of disabling gesture mode"
Input: psmouse - fix data race in __ps2_command
Input: elan_i2c - add all valid ic type for i2c/smbus
Input: zhenhua - ensure we have BITREVERSE
Input: omap4-keypad - fix memory leak
Input: serio - fix blocking of parport
Input: uinput - fix crash when using ABS events
Input: elan_i2c - expand maximum product_id form 0xFF to 0xFFFF
Input: elan_i2c - add ic type 0x03
Input: elan_i2c - don't require known iap version
Input: imx6ul_tsc - fix controller name
Input: imx6ul_tsc - use the preferred method for kzalloc()
Input: imx6ul_tsc - check for negative return value
Input: imx6ul_tsc - propagate the errors
Input: walkera0701 - fix abs() calculations on 64 bit values
Input: mms114 - remove unneded semicolons
Input: pm8941-pwrkey - remove unneded semicolon
Input: fix typo in MT documentation
Input: cyapa - fix address of Gen3 devices in device tree documentation
|
|
This adds support for the FT6x06 and the FT6x36 family of capacitive touch
panel controllers, in particular the FT6236.
Signed-off-by: Noralf Trønnes <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
|