platform/kernel/linux-starfive.git
8 years agoleds: bcm6328: simplify duplicated unlock in bcm6328_blink_set
Álvaro Fernández Rojas [Tue, 15 Dec 2015 20:03:51 +0000 (21:03 +0100)]
leds: bcm6328: simplify duplicated unlock in bcm6328_blink_set

Instead of duplicating the unlock we can store the return value and use a
single unlock before returning that value, which simplifies the code.

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: bcm6358: add little endian support
Álvaro Fernández Rojas [Tue, 15 Dec 2015 20:03:50 +0000 (21:03 +0100)]
leds: bcm6358: add little endian support

This adds little endian support while keeping big endian support.

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: bcm6328: add little endian support
Álvaro Fernández Rojas [Tue, 15 Dec 2015 20:03:49 +0000 (21:03 +0100)]
leds: bcm6328: add little endian support

This adds little endian support while keeping big endian support.

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: bcm6358: remove unneded lock when checking initial LED status
Álvaro Fernández Rojas [Tue, 15 Dec 2015 20:03:48 +0000 (21:03 +0100)]
leds: bcm6358: remove unneded lock when checking initial LED status

This lock isn't really needed, since we're only reading the register and
changes made to other LEDs aren't relevant.

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: bcm6358: Use bcm6358_led_set() in order to get rid of the lock
Álvaro Fernández Rojas [Tue, 15 Dec 2015 20:03:47 +0000 (21:03 +0100)]
leds: bcm6358: Use bcm6358_led_set() in order to get rid of the lock

Replaces bcm6358_led_mode with bcm6358_led_set so we don't need to handle
the lock.

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: bcm6328: remove unneded lock when checking initial LED
Álvaro Fernández Rojas [Tue, 15 Dec 2015 20:03:46 +0000 (21:03 +0100)]
leds: bcm6328: remove unneded lock when checking initial LED

This lock isn't really needed, since we're only reading the register and
changes made to other LEDs aren't relevant.

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: bcm6328: code cleaning
Álvaro Fernández Rojas [Tue, 15 Dec 2015 20:03:45 +0000 (21:03 +0100)]
leds: bcm6328: code cleaning

- Remove double whitespace
- Add missing brackets
- Add spinlock info

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: syscon: Make the driver explicitly non-modular
Paul Gortmaker [Sun, 13 Dec 2015 21:45:51 +0000 (16:45 -0500)]
leds: syscon: Make the driver explicitly non-modular

The Kconfig currently controlling compilation of this code is:

drivers/leds/Kconfig:config LEDS_SYSCON
drivers/leds/Kconfig:   bool "LED support for LEDs on system controllers"

...meaning that it currently is not being built as a module by anyone.

Lets remove the modular code that is essentially orphaned, so that
when reading the driver there is no doubt it is builtin-only.

We explicitly disallow a driver unbind, since that doesn't have a
sensible use case anyway, and it allows us to drop the ".remove"
code for non-modular drivers.

Since module_platform_driver() uses the same init level priority as
builtin_platform_driver() the init ordering remains unchanged with
this commit.

Also note that MODULE_DEVICE_TABLE is a no-op for non-modular code.

Cc: Richard Purdie <rpurdie@rpsys.net>
Cc: linux-leds@vger.kernel.org
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: ledtrig-ide-disk: Make the driver explicitly non-modular
Paul Gortmaker [Sun, 13 Dec 2015 21:45:50 +0000 (16:45 -0500)]
leds: ledtrig-ide-disk: Make the driver explicitly non-modular

The Kconfig for this driver is currently:

config LEDS_TRIGGER_IDE_DISK
        bool "LED IDE Disk Trigger"

...meaning that it currently is not being built as a module by anyone.
Lets remove the modular code that is essentially orphaned, so that
when reading the driver there is no doubt it is builtin-only.

Since module_init translates to device_initcall in the non-modular
case, the init ordering remains unchanged with this commit.

We also delete the MODULE_LICENSE tag etc. since all that information
is already contained at the top of the file in the comments.

Cc: Bryan Wu <cooloney@gmail.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Cc: linux-leds@vger.kernel.org
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: ledtrig-cpu: Make the driver explicitly non-modular
Paul Gortmaker [Sun, 13 Dec 2015 21:45:49 +0000 (16:45 -0500)]
leds: ledtrig-cpu: Make the driver explicitly non-modular

The Kconfig for this driver is currently:

config LEDS_TRIGGER_CPU
        bool "LED CPU Trigger"

...meaning that it currently is not being built as a module by anyone.
Lets remove the modular code that is essentially orphaned, so that
when reading the driver there is no doubt it is builtin-only.

Since module_init translates to device_initcall in the non-modular
case, the init ordering remains unchanged with this commit.

We also delete the MODULE_LICENSE tag etc. since all that information
is already contained at the top of the file in the comments.

Cc: Bryan Wu <cooloney@gmail.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Cc: linux-leds@vger.kernel.org
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: sunfire: Use platform_register/unregister_drivers()
Thierry Reding [Wed, 2 Dec 2015 16:27:01 +0000 (17:27 +0100)]
leds: sunfire: Use platform_register/unregister_drivers()

These new helpers simplify implementing multi-driver modules and
properly handle failure to register one driver by unregistering all
previously registered drivers.

Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: max77693: Add missing of_node_put
Jacek Anaszewski [Wed, 25 Nov 2015 08:34:10 +0000 (09:34 +0100)]
leds: max77693: Add missing of_node_put

Each break out of the for_each_available_child_of_node loop requires
of_node_put().

This patch adds missing of_node_put() when loop breaks on conflicting
values of led-sources DT property.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Cc: Rob Herring <robh@kernel.org>
Cc: Sakari Ailus <sakari.ailus@linux.intel.com>
8 years agoleds: aat1290: Add missing of_node_put
Jacek Anaszewski [Wed, 25 Nov 2015 08:22:19 +0000 (09:22 +0100)]
leds: aat1290: Add missing of_node_put

The refcount of device_node increases after of_get_next_available_child()
is called. Make sure it is decremented also in case of DT parsing error.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Cc: Rob Herring <robh@kernel.org>
Cc: Sakari Ailus <sakari.ailus@linux.intel.com>
8 years agoleds: powernv: Implement brightness_set_blocking op
Jacek Anaszewski [Fri, 20 Nov 2015 15:38:50 +0000 (16:38 +0100)]
leds: powernv: Implement brightness_set_blocking op

Since brightness setting can sleep for this driver, implement
brightness_set_blocking op, instead of brightness_set.
It makes this driver compatible with LED triggers.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Cc: Vasant Hegde <hegdevasant@linux.vnet.ibm.com>
8 years agoleds: ipaq-micro: Implement brightness_set_blocking op
Jacek Anaszewski [Fri, 20 Nov 2015 15:47:51 +0000 (16:47 +0100)]
leds: ipaq-micro: Implement brightness_set_blocking op

Since brightness setting can sleep for this driver, implement
brightness_set_blocking op, instead of brightness_set.
It makes this driver compatible with LED triggers.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
8 years agoleds: powernv: removing NULL check
Saurabh Sengar [Mon, 23 Nov 2015 09:28:16 +0000 (14:58 +0530)]
leds: powernv: removing NULL check

No need to explicitly check for pointer to be null,
of_prop_next_string anyways returns NULL, if passed pointer is NULL
and hence loop will continue.

Signed-off-by: Saurabh Sengar <saurabh.truth@gmail.com>
Acked-by: Vasant Hegde <hegdevasant@linux.vnet.ibm.com>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: bcm6328: Swap LED ON and OFF definitions
Simon Arlott [Sun, 15 Nov 2015 13:34:37 +0000 (13:34 +0000)]
leds: bcm6328: Swap LED ON and OFF definitions

The values of BCM6328_LED_MODE_ON and BCM6328_LED_MODE_OFF were named
for active low LEDs. These should be swapped so that they are named for
the default case of active high LEDs.

Signed-off-by: Simon Arlott <simon@fire.lp0.eu>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: bcm6328: Reuse bcm6328_led_set() instead of copying its functionality
Simon Arlott [Mon, 16 Nov 2015 20:24:59 +0000 (20:24 +0000)]
leds: bcm6328: Reuse bcm6328_led_set() instead of copying its functionality

When ensuring a consistent initial LED state in bcm6328_led (as they may
be blinking instead of on/off), the LED register is set using an inverted
copy of bcm6328_led_set(). To avoid further errors relating to active low
handling, call this function directly instead.

As bcm6328_led_set() acquires the same spinlock again when updating the
register, it is called after unlocking.

Signed-off-by: Simon Arlott <simon@fire.lp0.eu>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: ns2: Remove work queue
Jacek Anaszewski [Fri, 20 Nov 2015 10:39:41 +0000 (11:39 +0100)]
leds: ns2: Remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Tested-by: Simon Guinot <simon.guinot@sequanux.org>
8 years agoleds: pwm: fix driver description and make license match the header
Uwe Kleine-König [Mon, 25 Nov 2013 20:43:45 +0000 (21:43 +0100)]
leds: pwm: fix driver description and make license match the header

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: turn off the LED and wait for completion on unregistering LED class device
Milo Kim [Fri, 20 Nov 2015 08:03:00 +0000 (17:03 +0900)]
leds: turn off the LED and wait for completion on unregistering LED class device

Workqueue, 'set_brightness_work' is used for scheduling brightness control.
This workqueue is canceled when the LED class device is unregistered.
Currently, LED subsystem handles like below.

  cancel_work_sync(&led_cdev->set_brightness_work)
  led_set_brightness(led_cdev, LED_OFF)

However, this could be a problem.
Workqueue is going to be canceled but LED device needs to be off.
The worst case is null pointer access due to scheduling a workqueue.

LED module is loaded.
  LED driver private data is allocated by using devm_zalloc().

LED module is unloaded.
  led_classdev_unregister() is called.
    cancel_work_sync()
      led_set_brightness(led_cdev, LED_OFF)
        schedule_work() if LED driver uses brightness_set_blocking()
        In the meantime, driver private data will be freed.

        ..scheduling..

        brightness_set_blocking() callback is invoked.
          For the brightness control, LED driver tries to access private
          data but resource is removed!

To avoid this problem, LED subsystem should turn off the brightness first
and wait for completion.

  led_set_brightness(led_cdev, LED_OFF)
  flush_work(&led_cdev->set_brightness_work)

It guarantees that LED driver turns off the brightness prior to
resource management.

Cc: linux-leds@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Milo Kim <milo.kim@ti.com>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: flash: Remove checking for state < 1 in flash_strobe_store()
Jacek Anaszewski [Wed, 18 Nov 2015 15:32:37 +0000 (16:32 +0100)]
leds: flash: Remove checking for state < 1 in flash_strobe_store()

Strobe state variable is declared as unsigned long, remove
the check for values less than zero then.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Reported-by: David Binderman <dcb314@hotmail.com>
8 years agoleds: ktd2692: Add missing of_node_put
Ingi Kim [Mon, 16 Nov 2015 14:19:55 +0000 (15:19 +0100)]
leds: ktd2692: Add missing of_node_put

The refcount of device_node increases after of_node_get() is called.
So, a break out of the loop requires of_node_put().

This patch adds missing of_node_put() when loop breaks.

Signed-off-by: Ingi Kim <ingi2.kim@samsung.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: call led_pwm_set() in leds-pwm to enforce default LED_OFF
Markus Hofstaetter [Wed, 11 Nov 2015 11:40:29 +0000 (12:40 +0100)]
leds: call led_pwm_set() in leds-pwm to enforce default LED_OFF

Some PWMs are disabled by default or the default pin setting
does not match the LED_OFF state (e.g., active-low leds).
Hence, the driver may end up reporting 0 brightness, but
the leds are actually on using full brightness, because
it never enforces its default configuration.
So enforce it by calling led_pwm_set() after successfully
registering the device.

Tested on a Phytec phyFLEX i.MX6Q board based on kernel
v3.19.5.

Signed-off-by: Markus Hofstaetter <markus.hofstaetter@ait.ac.at>
Tested-by: Markus Hofstaetter <markus.hofstaetter@ait.ac.at>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: triggers: oneshot: fix module license specification
Uwe Kleine-König [Thu, 12 Nov 2015 09:20:39 +0000 (10:20 +0100)]
leds: triggers: oneshot: fix module license specification

The header of this file fixes the license to GPL 2 only without the
option to use later version. So use the string "GPL v2" that is to be
used in this case.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Acked-by: Fabio Baltieri <fabio.baltieri@gmail.com>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: ledtrig-transient: fix duration to be msec instead of jiffies
Rob Herring [Tue, 10 Nov 2015 23:10:17 +0000 (17:10 -0600)]
leds: ledtrig-transient: fix duration to be msec instead of jiffies

The transient trigger duration is documented to be in msec units, but is
actually in jiffies units. Other time based triggers are in msec units
as well. Fix the timer setup to convert from msec.

This could break an existing userspace that worked around this problem,
but exposing jiffies to userspace is just wrong and would break anyway
if HZ is changed.

Signed-off-by: Rob Herring <robh@kernel.org>
Cc: Shuah Khan <shuahkhan@gmail.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Cc: linux-leds@vger.kernel.org
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: lm355x: Remove work queue
Jacek Anaszewski [Thu, 20 Aug 2015 13:53:10 +0000 (15:53 +0200)]
leds: lm355x: Remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Cc: Daniel Jeong <daniel.jeong@ti.com>
Cc: G.Shark Jeong <gshark.jeong@gmail.com>
8 years agoleds: pwm: remove work queue
Jacek Anaszewski [Thu, 20 Aug 2015 13:52:51 +0000 (15:52 +0200)]
leds: pwm: remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Cc: Raphael Assenat <raph@8d.com>
8 years agoleds: gpio: Remove work queue
Jacek Anaszewski [Thu, 20 Aug 2015 13:52:32 +0000 (15:52 +0200)]
leds: gpio: Remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Cc: Raphael Assenat <raph@8d.com>
8 years agoleds: wm8350: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 10:59:45 +0000 (12:59 +0200)]
leds: wm8350: Remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Acked-by: Antonio Ospite <ao2@ao2.it>
Reviewed-by: Mark Brown <broonie@kernel.org>
8 years agoleds: regulator: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 10:55:39 +0000 (12:55 +0200)]
leds: regulator: Remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Acked-by: Antonio Ospite <ao2@ao2.it>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: mc13783: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 10:52:08 +0000 (12:52 +0200)]
leds: mc13783: Remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: max8997: Remove unneeded workqueue include
Andrew Lunn [Tue, 11 Aug 2015 07:58:47 +0000 (09:58 +0200)]
leds: max8997: Remove unneeded workqueue include

Work queues are not used in this driver, so remove the include.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
8 years agoleds: lt3593: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 10:50:05 +0000 (12:50 +0200)]
leds: lt3593: Remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Cc: Daniel Mack <daniel@caiaq.de>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: dac124d085: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 10:47:25 +0000 (12:47 +0200)]
leds: dac124d085: Remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Cc: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: da9052: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 10:45:23 +0000 (12:45 +0200)]
leds: da9052: Remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: da903x: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 10:43:49 +0000 (12:43 +0200)]
leds: da903x: Remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: wm831x: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 10:40:45 +0000 (12:40 +0200)]
leds: wm831x: Remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: pca963x: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 10:38:44 +0000 (12:38 +0200)]
leds: pca963x: Remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Acked-by: Dan Murphy <dmurphy@ti.com>
Cc: Ricardo Ribalda <ricardo.ribalda@gmail.com>
8 years agoleds: pca955x: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 10:32:35 +0000 (12:32 +0200)]
leds: pca955x: Remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Cc: Nate Case <ncase@xes-inc.com>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: lp8860: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 10:28:26 +0000 (12:28 +0200)]
leds: lp8860: Remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Acked-by: Dan Murphy <dmurphy@ti.com>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: lp8788: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 10:27:05 +0000 (12:27 +0200)]
leds: lp8788: Remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Cc: Milo Kim <milo.kim@ti.com>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: lp55xx: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 10:22:57 +0000 (12:22 +0200)]
leds: lp55xx: Remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Cc: Milo Kim <milo.kim@ti.com>
8 years agoleds: lp3944: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 10:09:57 +0000 (12:09 +0200)]
leds: lp3944: Remove work queue

Now the core implements the work queue, remove it from the driver,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Acked-by: Antonio Ospite <ao2@ao2.it>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: pca9532: Remove work queue for LEDs.
Andrew Lunn [Thu, 20 Aug 2015 10:08:10 +0000 (12:08 +0200)]
leds: pca9532: Remove work queue for LEDs.

Now the core implements the work queue, remove it from the driver,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: lm3642: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 10:05:54 +0000 (12:05 +0200)]
leds: lm3642: Remove work queue

Now the core implements the work queue, remove it from the driver,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Cc: Daniel Jeong <daniel.jeong@ti.com>
Cc: G.Shark Jeong <gshark.jeong@gmail.com>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: lm3533: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 10:02:17 +0000 (12:02 +0200)]
leds: lm3533: Remove work queue

Now the core implements the work queue, remove it from the driver,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Cc: Johan Hovold <johan@kernel.org>
8 years agoleds: blinkm: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 09:59:26 +0000 (11:59 +0200)]
leds: blinkm: Remove work queue

Now the core implements the work queue, remove it from the driver,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Cc: Jan-Simon Moeller <dl9pf@gmx.de>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: bd2802: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 09:57:07 +0000 (11:57 +0200)]
leds: bd2802: Remove work queue

Now the core implements the work queue, remove it from the driver,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Cc: Kim Kyuwon <q1.kim@samsung.com>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: adp5520: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 13:54:57 +0000 (15:54 +0200)]
leds: adp5520: Remove work queue

Now the core implements the work queue, remove it from the driver,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Cc: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: 88pm860x: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 09:50:34 +0000 (11:50 +0200)]
leds: 88pm860x: Remove work queue

Now the core implements the work queue, remove it from the driver,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agoleds: tlc591xx: Remove work queue
Andrew Lunn [Thu, 20 Aug 2015 09:41:01 +0000 (11:41 +0200)]
leds: tlc591xx: Remove work queue

Now the core implements the work queue, remove it from the driver,
and switch to using brightness_set_blocking op.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
8 years agomedia: flash: use led_set_brightness_sync for torch brightness
Jacek Anaszewski [Wed, 7 Oct 2015 09:10:48 +0000 (11:10 +0200)]
media: flash: use led_set_brightness_sync for torch brightness

LED subsystem shifted responsibility for choosing between SYNC or ASYNC
way of setting brightness from drivers to the caller. Adapt the wrapper
to those changes.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Cc: linux-media@vger.kernel.org
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Acked-by: Pavel Machek <pavel@ucw.cz>
Acked-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
8 years agoleds: ktd2692: Remove work queue
Jacek Anaszewski [Wed, 7 Oct 2015 09:10:47 +0000 (11:10 +0200)]
leds: ktd2692: Remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Cc: Ingi Kim <ingi2.kim@samsung.com>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Acked-by: Pavel Machek <pavel@ucw.cz>
8 years agoleds: aat1290: Remove work queue
Jacek Anaszewski [Wed, 7 Oct 2015 09:10:46 +0000 (11:10 +0200)]
leds: aat1290: Remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Acked-by: Pavel Machek <pavel@ucw.cz>
8 years agoleds: max77693: Remove work queue
Jacek Anaszewski [Wed, 7 Oct 2015 09:10:45 +0000 (11:10 +0200)]
leds: max77693: Remove work queue

Now the core implements the work queue, remove it from the drivers,
and switch to using brightness_set_blocking op.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Acked-by: Pavel Machek <pavel@ucw.cz>
8 years agoDocumentation: leds: Add description of brightness setting API
Jacek Anaszewski [Wed, 7 Oct 2015 09:10:44 +0000 (11:10 +0200)]
Documentation: leds: Add description of brightness setting API

This patch adds description of the LED subsystem API for
setting an LED brightness.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
8 years agoleds: core: Drivers shouldn't enforce SYNC/ASYNC brightness setting
Jacek Anaszewski [Wed, 7 Oct 2015 09:10:43 +0000 (11:10 +0200)]
leds: core: Drivers shouldn't enforce SYNC/ASYNC brightness setting

This patch removes SET_BRIGHTNESS_ASYNC and SET_BRIGHTNESS_SYNC flags.
led_set_brightness() now calls led_set_brightness_nosleep() instead of
choosing between sync and async op basing on the flags defined by the
driver.

From now on, if a user wants to make sure that brightness will be set
synchronously, they have to use led_set_brightness_sync() API. It is now
being made publicly available since it has become apparent that it is
a caller who should decide whether brightness is to be set in
a synchronous or an asynchronous way.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
8 years agoleds: core: Use set_brightness_work for the blocking op
Jacek Anaszewski [Mon, 19 Oct 2015 07:04:01 +0000 (09:04 +0200)]
leds: core: Use set_brightness_work for the blocking op

This patch makes LED core capable of setting brightness for drivers
that implement brightness_set_blocking op. It removes from LED class
drivers responsibility for using work queues on their own.

In order to achieve this set_brightness_delayed callback is being
modified to directly call one of available ops for brightness setting.

led_set_brightness_async() function didn't set brightness in an
asynchronous way in all cases. It was mistakenly assuming that all
LED subsystem drivers used work queue in their brightness_set op,
whereas only half of them did that. Since it has no users now,
it is being removed.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
8 years agoleds: core: Add led_set_brightness_nosleep{nopm} functions
Jacek Anaszewski [Wed, 7 Oct 2015 09:10:41 +0000 (11:10 +0200)]
leds: core: Add led_set_brightness_nosleep{nopm} functions

This patch adds led_set_brightness_nosleep() and led_set_brightness_nopm()
functions, that guarantee setting LED brightness in a non-blocking way.
The latter is used from pm_ops context and doesn't modify the brightness
cached in the struct led_classdev. Its execution always ends up with
a call to brightness setting op - either directly or through
a set_brightness_work, regardless of LED_SUSPENDED flag state.

The patch also replaces led_set_brightness_async() with
led_set_brightness_nosleep() in all places where the most vital was setting
brightness in a non sleeping way but not necessarily asynchronously, which
is not needed for non-blocking drivers.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
8 years agoleds: Rename brightness_set_sync op to brightness_set_blocking
Jacek Anaszewski [Wed, 7 Oct 2015 09:10:40 +0000 (11:10 +0200)]
leds: Rename brightness_set_sync op to brightness_set_blocking

The initial purpose of brightness_set_sync op, introduced along with
the LED flash class extension, was to add a means for setting torch LED
brightness as soon as possible, which couldn't have been guaranteed by
brightness_set op. This patch renames the op to brightness_set_blocking,
which describes its purpose in a more generic way. It is beneficial
in view of the prospective changes in the LED core, aiming at removing
the need for using work queues in LED class drivers that can sleep
or use delays while setting brightness.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Acked-by: Andrew Lunn <andrew@lunn.ch>
Acked-by: Pavel Machek <pavel@ucw.cz>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
8 years agoleds: core: Add two new LED_BLINK_ flags
Jacek Anaszewski [Wed, 7 Oct 2015 09:10:39 +0000 (11:10 +0200)]
leds: core: Add two new LED_BLINK_ flags

This patch adds LED_BLINK_BRIGHTNESS_CHANGE flag to indicate that blink
brightness has changed, and LED_BLINK_DISABLE flag to indicate that
blinking deactivation has been requested. In order to use the flags
led_timer_function and set_brightness_delayed callbacks as well as
led_set_brightness() function are being modified. The main goal of these
modifications is to prepare set_brightness_work for extension of the
scope of its responsibilities.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Acked-by: Pavel Machek <pavel@ucw.cz>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
8 years agoleds: core: Use EXPORT_SYMBOL_GPL consistently
Jacek Anaszewski [Mon, 28 Sep 2015 13:07:10 +0000 (15:07 +0200)]
leds: core: Use EXPORT_SYMBOL_GPL consistently

LED core has a mixture of EXPORT_SYMBOL and EXPORT_SYMBOL_GPL macros.
This patch fixes this discrepancy and switches to using EXPORT_SYMBOL_GPL
for each exported function.

Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
Acked-by: Pavel Machek <pavel@ucw.cz>
8 years agoLinux 4.4-rc8
Linus Torvalds [Sun, 3 Jan 2016 23:15:37 +0000 (15:15 -0800)]
Linux 4.4-rc8

8 years agoMerge branch 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus
Linus Torvalds [Sun, 3 Jan 2016 19:49:31 +0000 (11:49 -0800)]
Merge branch 'upstream' of git://git.linux-mips.org/ralf/upstream-linus

Pull MIPS build fix from Ralf Baechle:
 "Fix a makefile issue resulting in build breakage with older binutils.

  This has sat in -next for a few days, testers and buildbot are happy
  with it, too though if you are going for another -rc that'd certainly
  help ironing out a few more issues"

* 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus:
  MIPS: VDSO: Fix build error with binutils 2.24 and earlier

8 years agoMerge tag 'drm-intel-fixes-2016-01-02' of git://anongit.freedesktop.org/drm-intel
Linus Torvalds [Sun, 3 Jan 2016 19:36:26 +0000 (11:36 -0800)]
Merge tag 'drm-intel-fixes-2016-01-02' of git://anongit.freedesktop.org/drm-intel

Pull i915 drm fixes from Jani Nikula:
 "Two display fixes still for v4.4.

  The new year's resolution is to start using signed tags per Linus'
  request.  This one is still unsigned; I want to fix this up in our
  maintainer scripts instead of doing it one-off"

* tag 'drm-intel-fixes-2016-01-02' of git://anongit.freedesktop.org/drm-intel:
  drm/i915: increase the tries for HDMI hotplug live status checking
  drm/i915: Unbreak check_digital_port_conflicts()

8 years agoMerge tag 'pci-v4.4-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci
Linus Torvalds [Thu, 31 Dec 2015 22:59:21 +0000 (14:59 -0800)]
Merge tag 'pci-v4.4-fixes-3' of git://git./linux/kernel/git/helgaas/pci

Pull PCI bugfix from Bjorn Helgaas:
 "Here's another fix for v4.4.

  This fixes 32-bit config reads for the HiSilicon driver.  Obviously
  the driver is completely broken without this fix (apparently it
  actually was tested internally, but got broken somehow in the process
  of upstreaming it).

  Summary:

  HiSilicon host bridge driver
    Fix 32-bit config reads (Dongdong Liu)"

* tag 'pci-v4.4-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
  PCI: hisi: Fix hisi_pcie_cfg_read() 32-bit reads

8 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc
Linus Torvalds [Thu, 31 Dec 2015 22:46:49 +0000 (14:46 -0800)]
Merge git://git./linux/kernel/git/davem/sparc

Pull sparc fixes from David Miller:
 "Just some missing syscall wire ups"

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc:
  sparc: Wire up mlock2 system call.
  sparc: Add all necessary direct socket system calls.

8 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Thu, 31 Dec 2015 22:40:43 +0000 (14:40 -0800)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Prevent XFRM per-cpu counter updates for one namespace from being
    applied to another namespace.  Fix from DanS treetman.

 2) Fix RCU de-reference in iwl_mvm_get_key_sta_id(), from Johannes
    Berg.

 3) Remove ethernet header assumption in nft_do_chain_netdev(), from
    Pablo Neira Ayuso.

 4) Fix cpsw PHY ident with multiple slaves and fixed-phy, from Pascal
    Speck.

 5) Fix use after free in sixpack_close and mkiss_close.

 6) Fix VXLAN fw assertion on bnx2x, from Yuval Mintz.

 7) natsemi doesn't check for DMA mapping errors, from Alexey
    Khoroshilov.

 8) Fix inverted test in ip6addrlbl_get(), from ANdrey Ryabinin.

 9) Missing initialization of needed_headroom in geneve tunnel driver,
    from Paolo Abeni.

10) Fix conntrack template leak in openvswitch, from Joe Stringer.

11) Mission initialization of wq->flags in sock_alloc_inode(), from
    Nicolai Stange.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (35 commits)
  sctp: sctp should release assoc when sctp_make_abort_user return NULL in sctp_close
  net, socket, socket_wq: fix missing initialization of flags
  drivers: net: cpsw: fix error return code
  openvswitch: Fix template leak in error cases.
  sctp: label accepted/peeled off sockets
  sctp: use GFP_USER for user-controlled kmalloc
  qlcnic: fix a loop exit condition better
  net: cdc_ncm: avoid changing RX/TX buffers on MTU changes
  geneve: initialize needed_headroom
  ipv6: honor ifindex in case we receive ll addresses in router advertisements
  addrconf: always initialize sysctl table data
  ipv6/addrlabel: fix ip6addrlbl_get()
  switchdev: bridge: Pass ageing time as clock_t instead of jiffies
  sh_eth: fix 16-bit descriptor field access endianness too
  veth: don’t modify ip_summed; doing so treats packets with bad checksums as good.
  net: usb: cdc_ncm: Adding Dell DW5813 LTE AT&T Mobile Broadband Card
  net: usb: cdc_ncm: Adding Dell DW5812 LTE Verizon Mobile Broadband Card
  natsemi: add checks for dma mapping errors
  rhashtable: Kill harmless RCU warning in rhashtable_walk_init
  openvswitch: correct encoding of set tunnel action attributes
  ...

8 years agosparc: Wire up mlock2 system call.
David S. Miller [Thu, 31 Dec 2015 20:38:56 +0000 (15:38 -0500)]
sparc: Wire up mlock2 system call.

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agosparc: Add all necessary direct socket system calls.
David S. Miller [Thu, 31 Dec 2015 20:18:02 +0000 (15:18 -0500)]
sparc: Add all necessary direct socket system calls.

The GLIBC folks would like to eliminate socketcall support
eventually, and this makes sense regardless so wire them
all up.

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agosctp: sctp should release assoc when sctp_make_abort_user return NULL in sctp_close
Xin Long [Tue, 29 Dec 2015 09:49:25 +0000 (17:49 +0800)]
sctp: sctp should release assoc when sctp_make_abort_user return NULL in sctp_close

In sctp_close, sctp_make_abort_user may return NULL because of memory
allocation failure. If this happens, it will bypass any state change
and never free the assoc. The assoc has no chance to be freed and it
will be kept in memory with the state it had even after the socket is
closed by sctp_close().

So if sctp_make_abort_user fails to allocate memory, we should abort
the asoc via sctp_primitive_ABORT as well. Just like the annotation in
sctp_sf_cookie_wait_prm_abort and sctp_sf_do_9_1_prm_abort said,
"Even if we can't send the ABORT due to low memory delete the TCB.
This is a departure from our typical NOMEM handling".

But then the chunk is NULL (low memory) and the SCTP_CMD_REPLY cmd would
dereference the chunk pointer, and system crash. So we should add
SCTP_CMD_REPLY cmd only when the chunk is not NULL, just like other
places where it adds SCTP_CMD_REPLY cmd.

Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge tag 'wireless-drivers-for-davem-2015-12-28' of git://git.kernel.org/pub/scm...
David S. Miller [Wed, 30 Dec 2015 21:42:31 +0000 (16:42 -0500)]
Merge tag 'wireless-drivers-for-davem-2015-12-28' of git://git./linux/kernel/git/kvalo/wireless-drivers

Kalle Valo says:

====================
iwlwifi

* don't load firmware that won't exist for 7260
* fix RCU splat
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet, socket, socket_wq: fix missing initialization of flags
Nicolai Stange [Tue, 29 Dec 2015 12:29:55 +0000 (13:29 +0100)]
net, socket, socket_wq: fix missing initialization of flags

Commit ceb5d58b2170 ("net: fix sock_wake_async() rcu protection") from
the current 4.4 release cycle introduced a new flags member in
struct socket_wq and moved SOCKWQ_ASYNC_NOSPACE and SOCKWQ_ASYNC_WAITDATA
from struct socket's flags member into that new place.

Unfortunately, the new flags field is never initialized properly, at least
not for the struct socket_wq instance created in sock_alloc_inode().

One particular issue I encountered because of this is that my GNU Emacs
failed to draw anything on my desktop -- i.e. what I got is a transparent
window, including the title bar. Bisection lead to the commit mentioned
above and further investigation by means of strace told me that Emacs
is indeed speaking to my Xorg through an O_ASYNC AF_UNIX socket. This is
reproducible 100% of times and the fact that properly initializing the
struct socket_wq ->flags fixes the issue leads me to the conclusion that
somehow SOCKWQ_ASYNC_WAITDATA got set in the uninitialized ->flags,
preventing my Emacs from receiving any SIGIO's due to data becoming
available and it got stuck.

Make sock_alloc_inode() set the newly created struct socket_wq's ->flags
member to zero.

Fixes: ceb5d58b2170 ("net: fix sock_wake_async() rcu protection")
Signed-off-by: Nicolai Stange <nicstange@gmail.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-block
Linus Torvalds [Wed, 30 Dec 2015 18:26:20 +0000 (10:26 -0800)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:
 "Make the block layer great again.

  Basically three amazing fixes in this pull request, split into 4
  patches.  Believe me, they should go into 4.4.  Two of them fix a
  regression, the third and last fixes an easy-to-trigger bug.

   - Fix a bad irq enable through null_blk, for queue_mode=1 and using
     timer completions.  Add a block helper to restart a queue
     asynchronously, and use that from null_blk.  From me.

   - Fix a performance issue in NVMe.  Some devices (Intel Pxxxx) expose
     a stripe boundary, and performance suffers if we cross it.  We took
     that into account for merging, but not for the newer splitting
     code.  Fix from Keith.

   - Fix a kernel oops in lightnvm with multiple channels.  From Matias"

* 'for-linus' of git://git.kernel.dk/linux-block:
  lightnvm: wrong offset in bad blk lun calculation
  null_blk: use async queue restart helper
  block: add blk_start_queue_async()
  block: Split bios on chunk boundaries

8 years agodrm/i915: increase the tries for HDMI hotplug live status checking
Gary Wang [Wed, 23 Dec 2015 08:11:35 +0000 (16:11 +0800)]
drm/i915: increase the tries for HDMI hotplug live status checking

The total delay of HDMI hotplug detecting with 30ms is sometimes not
enoughtfor HDMI live status up with specific HDMI monitors in BSW platform.

After doing experiments for following monitors, it needs 80ms at least
for those worst cases.

Lenovo L246 1xwA (4 failed, necessary hot-plug delay: 58/40/60/40ms)
Philips HH2AP (9 failed, necessary hot-plug delay: 80/50/50/60/46/40/58/58/39ms)
BENQ ET-0035-N (6 failed, necessary hot-plug delay: 60/50/50/80/80/40ms)
DELL U2713HM (2 failed, necessary hot-plug delay: 58/59ms)
HP HP-LP2475w (5 failed, necessary hot-plug delay: 70/50/40/60/40ms)

It looks like 70-80 ms is BSW platform needs in some bad cases of the
monitors at this end (8 times delay at most). Keep less than 100ms for
HDCP pulse HPD low (with at least 100ms) to respond a plug out.

Reviewed-by: Cooper Chiou <cooper.chiou@intel.com>
Tested-by: Gary Wang <gary.c.wang@intel.com>
Cc: Gavin Hindman <gavin.hindman@intel.com>
Cc: Sonika Jindal <sonika.jindal@intel.com>
Cc: Shashank Sharma <shashank.sharma@intel.com>
Cc: Shobhit Kumar <shobhit.kumar@intel.com>
Signed-off-by: Gary Wang <gary.c.wang@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450858295-12804-1-git-send-email-gary.c.wang@intel.com
Tested-by: Shobhit Kumar <shobhit.kumar@intel.com>
Cc: drm-intel-fixes@lists.freedesktop.org
Fixes: 237ed86c693d ("drm/i915: Check live status before reading edid")
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
(cherry picked from commit f8d03ea0053b23de42c828d559016eabe0b91523)
[Jani: undo the file mode change of the original commit]
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
8 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Wed, 30 Dec 2015 02:04:41 +0000 (18:04 -0800)]
Merge branch 'akpm' (patches from Andrew)

Merge misc fixes from Andrew Morton:
 "9 fixes"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  mm/vmstat: fix overflow in mod_zone_page_state()
  ocfs2/dlm: clear migration_pending when migration target goes down
  mm/memory_hotplug.c: check for missing sections in test_pages_in_a_zone()
  ocfs2: fix flock panic issue
  m32r: add io*_rep helpers
  m32r: fix build failure
  arch/x86/xen/suspend.c: include xen/xen.h
  mm: memcontrol: fix possible memcg leak due to interrupted reclaim
  ocfs2: fix BUG when calculate new backup super

8 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Linus Torvalds [Wed, 30 Dec 2015 01:46:29 +0000 (17:46 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/viro/vfs

Pull vfs fix from Al Viro:
 "Fix for 3.15 breakage of fcntl64() in arm OABI compat.  -stable
  fodder"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  [PATCH] arm: fix handling of F_OFD_... in oabi_fcntl64()

8 years agomm/vmstat: fix overflow in mod_zone_page_state()
Heiko Carstens [Tue, 29 Dec 2015 22:54:32 +0000 (14:54 -0800)]
mm/vmstat: fix overflow in mod_zone_page_state()

mod_zone_page_state() takes a "delta" integer argument.  delta contains
the number of pages that should be added or subtracted from a struct
zone's vm_stat field.

If a zone is larger than 8TB this will cause overflows.  E.g.  for a
zone with a size slightly larger than 8TB the line

    mod_zone_page_state(zone, NR_ALLOC_BATCH, zone->managed_pages);

in mm/page_alloc.c:free_area_init_core() will result in a negative
result for the NR_ALLOC_BATCH entry within the zone's vm_stat, since 8TB
contain 0x8xxxxxxx pages which will be sign extended to a negative
value.

Fix this by changing the delta argument to long type.

This could fix an early boot problem seen on s390, where we have a 9TB
system with only one node.  ZONE_DMA contains 2GB and ZONE_NORMAL the
rest.  The system is trying to allocate a GFP_DMA page but ZONE_DMA is
completely empty, so it tries to reclaim pages in an endless loop.

This was seen on a heavily patched 3.10 kernel.  One possible
explaination seem to be the overflows caused by mod_zone_page_state().
Unfortunately I did not have the chance to verify that this patch
actually fixes the problem, since I don't have access to the system
right now.  However the overflow problem does exist anyway.

Given the description that a system with slightly less than 8TB does
work, this seems to be a candidate for the observed problem.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agoocfs2/dlm: clear migration_pending when migration target goes down
xuejiufei [Tue, 29 Dec 2015 22:54:29 +0000 (14:54 -0800)]
ocfs2/dlm: clear migration_pending when migration target goes down

We have found a BUG on res->migration_pending when migrating lock
resources.  The situation is as follows.

dlm_mark_lockres_migration
  res->migration_pending = 1;
  __dlm_lockres_reserve_ast
  dlm_lockres_release_ast returns with res->migration_pending remains
      because other threads reserve asts
  wait dlm_migration_can_proceed returns 1
  >>>>>>> o2hb found that target goes down and remove target
          from domain_map
  dlm_migration_can_proceed returns 1
  dlm_mark_lockres_migrating returns -ESHOTDOWN with
      res->migration_pending still remains.

When reentering dlm_mark_lockres_migrating(), it will trigger the BUG_ON
with res->migration_pending.  So clear migration_pending when target is
down.

Signed-off-by: Jiufei Xue <xuejiufei@huawei.com>
Reviewed-by: Joseph Qi <joseph.qi@huawei.com>
Cc: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agomm/memory_hotplug.c: check for missing sections in test_pages_in_a_zone()
Andrew Banman [Tue, 29 Dec 2015 22:54:25 +0000 (14:54 -0800)]
mm/memory_hotplug.c: check for missing sections in test_pages_in_a_zone()

test_pages_in_a_zone() does not account for the possibility of missing
sections in the given pfn range.  pfn_valid_within always returns 1 when
CONFIG_HOLES_IN_ZONE is not set, allowing invalid pfns from missing
sections to pass the test, leading to a kernel oops.

Wrap an additional pfn loop with PAGES_PER_SECTION granularity to check
for missing sections before proceeding into the zone-check code.

This also prevents a crash from offlining memory devices with missing
sections.  Despite this, it may be a good idea to keep the related patch
'[PATCH 3/3] drivers: memory: prohibit offlining of memory blocks with
missing sections' because missing sections in a memory block may lead to
other problems not covered by the scope of this fix.

Signed-off-by: Andrew Banman <abanman@sgi.com>
Acked-by: Alex Thorlton <athorlton@sgi.com>
Cc: Russ Anderson <rja@sgi.com>
Cc: Alex Thorlton <athorlton@sgi.com>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Greg KH <greg@kroah.com>
Cc: Seth Jennings <sjennings@variantweb.net>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agoocfs2: fix flock panic issue
Junxiao Bi [Tue, 29 Dec 2015 22:54:22 +0000 (14:54 -0800)]
ocfs2: fix flock panic issue

Commit 4f6563677ae8 ("Move locks API users to locks_lock_inode_wait()")
move flock/posix lock indentify code to locks_lock_inode_wait(), but
missed to set fl_flags to FL_FLOCK which caused the following kernel
panic on 4.4.0_rc5.

  kernel BUG at fs/locks.c:1895!
  invalid opcode: 0000 [#1] SMP
  Modules linked in: ocfs2(O) ocfs2_dlmfs(O) ocfs2_stack_o2cb(O) ocfs2_dlm(O) ocfs2_nodemanager(O) ocfs2_stackglue(O) iscsi_tcp libiscsi_tcp libiscsi scsi_transport_iscsi xen_kbdfront xen_netfront xen_fbfront xen_blkfront
  CPU: 0 PID: 20268 Comm: flock_unit_test Tainted: G           O    4.4.0-rc5-next-20151217 #1
  Hardware name: Xen HVM domU, BIOS 4.3.1OVM 05/14/2014
  task: ffff88007b3672c0 ti: ffff880028b58000 task.ti: ffff880028b58000
  RIP: locks_lock_inode_wait+0x2e/0x160
  Call Trace:
    ocfs2_do_flock+0x91/0x160 [ocfs2]
    ocfs2_flock+0x76/0xd0 [ocfs2]
    SyS_flock+0x10f/0x1a0
    entry_SYSCALL_64_fastpath+0x12/0x71
  Code: e5 41 57 41 56 49 89 fe 41 55 41 54 53 48 89 f3 48 81 ec 88 00 00 00 8b 46 40 83 e0 03 83 f8 01 0f 84 ad 00 00 00 83 f8 02 74 04 <0f> 0b eb fe 4c 8d ad 60 ff ff ff 4c 8d 7b 58 e8 0e 8e 73 00 4d
  RIP  locks_lock_inode_wait+0x2e/0x160
   RSP <ffff880028b5bce8>
  ---[ end trace dfca74ec9b5b274c ]---

Fixes: 4f6563677ae8 ("Move locks API users to locks_lock_inode_wait()")
Signed-off-by: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Joseph Qi <joseph.qi@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agom32r: add io*_rep helpers
Sudip Mukherjee [Tue, 29 Dec 2015 22:54:19 +0000 (14:54 -0800)]
m32r: add io*_rep helpers

m32r allmodconfig was failing with the error:

  error: implicit declaration of function 'read'

On checking io.h it turned out that 'read' is not defined but 'readb' is
defined and 'ioread8' will then obviously mean 'readb'.

At the same time some of the helper functions ioreadN_rep() and
iowriteN_rep() were missing which also led to the build failure.

Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agom32r: fix build failure
Sudip Mukherjee [Tue, 29 Dec 2015 22:54:16 +0000 (14:54 -0800)]
m32r: fix build failure

m32r allmodconfig is failing with:

  In file included from ../include/linux/kvm_para.h:4:0,
                   from ../kernel/watchdog.c:26:
  ../include/uapi/linux/kvm_para.h:30:26: fatal error: asm/kvm_para.h: No such file or directory

kvm_para.h was not included in the build.

Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agoarch/x86/xen/suspend.c: include xen/xen.h
Andrew Morton [Tue, 29 Dec 2015 22:54:13 +0000 (14:54 -0800)]
arch/x86/xen/suspend.c: include xen/xen.h

Fix the build warning:

  arch/x86/xen/suspend.c: In function 'xen_arch_pre_suspend':
  arch/x86/xen/suspend.c:70:9: error: implicit declaration of function 'xen_pv_domain' [-Werror=implicit-function-declaration]
          if (xen_pv_domain())
              ^

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Cc: Sasha Levin <sasha.levin@oracle.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agomm: memcontrol: fix possible memcg leak due to interrupted reclaim
Vladimir Davydov [Tue, 29 Dec 2015 22:54:10 +0000 (14:54 -0800)]
mm: memcontrol: fix possible memcg leak due to interrupted reclaim

Memory cgroup reclaim can be interrupted with mem_cgroup_iter_break()
once enough pages have been reclaimed, in which case, in contrast to a
full round-trip over a cgroup sub-tree, the current position stored in
mem_cgroup_reclaim_iter of the target cgroup does not get invalidated
and so is left holding the reference to the last scanned cgroup.  If the
target cgroup does not get scanned again (we might have just reclaimed
the last page or all processes might exit and free their memory
voluntary), we will leak it, because there is nobody to put the
reference held by the iterator.

The problem is easy to reproduce by running the following command
sequence in a loop:

    mkdir /sys/fs/cgroup/memory/test
    echo 100M > /sys/fs/cgroup/memory/test/memory.limit_in_bytes
    echo $$ > /sys/fs/cgroup/memory/test/cgroup.procs
    memhog 150M
    echo $$ > /sys/fs/cgroup/memory/cgroup.procs
    rmdir test

The cgroups generated by it will never get freed.

This patch fixes this issue by making mem_cgroup_iter avoid taking
reference to the current position.  In order not to hit use-after-free
bug while running reclaim in parallel with cgroup deletion, we make use
of ->css_released cgroup callback to clear references to the dying
cgroup in all reclaim iterators that might refer to it.  This callback
is called right before scheduling rcu work which will free css, so if we
access iter->position from rcu read section, we might be sure it won't
go away under us.

[hannes@cmpxchg.org: clean up css ref handling]
Fixes: 5ac8fb31ad2e ("mm: memcontrol: convert reclaim iterator to simple css refcounting")
Signed-off-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@kernel.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: <stable@vger.kernel.org> [3.19+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agoocfs2: fix BUG when calculate new backup super
Joseph Qi [Tue, 29 Dec 2015 22:54:06 +0000 (14:54 -0800)]
ocfs2: fix BUG when calculate new backup super

When resizing, it firstly extends the last gd.  Once it should backup
super in the gd, it calculates new backup super and update the
corresponding value.

But it currently doesn't consider the situation that the backup super is
already done.  And in this case, it still sets the bit in gd bitmap and
then decrease from bg_free_bits_count, which leads to a corrupted gd and
trigger the BUG in ocfs2_block_group_set_bits:

    BUG_ON(le16_to_cpu(bg->bg_free_bits_count) < num_bits);

So check whether the backup super is done and then do the updates.

Signed-off-by: Joseph Qi <joseph.qi@huawei.com>
Reviewed-by: Jiufei Xue <xuejiufei@huawei.com>
Reviewed-by: Yiwen Jiang <jiangyiwen@huawei.com>
Cc: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agoMIPS: VDSO: Fix build error with binutils 2.24 and earlier
Guenter Roeck [Thu, 24 Dec 2015 05:04:31 +0000 (21:04 -0800)]
MIPS: VDSO: Fix build error with binutils 2.24 and earlier

Commit 2a037f310bab ("MIPS: VDSO: Fix build error") tries to fix a build
error seen with binutils 2.24 and earlier. However, the fix does not work,
and again results in the already known build errors if the kernel is built
with an earlier version of binutils.

CC      arch/mips/vdso/gettimeofday.o
/tmp/ccnOVbHT.s: Assembler messages:
/tmp/ccnOVbHT.s:50: Error: can't resolve `_start' {*UND* section} - `L0 {.text section}
/tmp/ccnOVbHT.s:374: Error: can't resolve `_start' {*UND* section} - `L0 {.text section}
scripts/Makefile.build:258: recipe for target 'arch/mips/vdso/gettimeofday.o' failed
make[2]: *** [arch/mips/vdso/gettimeofday.o] Error 1

Fixes: 2a037f310bab ("MIPS: VDSO: Fix build error")
Cc: Qais Yousef <qais.yousef@imgtec.com>
Cc: linux-mips@linux-mips.org
Cc: linux-kernel@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/11926/
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
8 years agodrivers: net: cpsw: fix error return code
Julia Lawall [Sat, 26 Dec 2015 19:12:13 +0000 (20:12 +0100)]
drivers: net: cpsw: fix error return code

Propagate the return value of platform_get_irq on failure.

A simplified version of the semantic match that finds the two cases where
no error code is returned at all is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@@
identifier ret; expression e1,e2;
@@
(
if (\(ret < 0\|ret != 0\))
 { ... return ret; }
|
ret = 0
)
... when != ret = e1
    when != &ret
*if(...)
{
  ... when != ret = e2
      when forall
 return ret;
}
// </smpl>

Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoopenvswitch: Fix template leak in error cases.
Joe Stringer [Wed, 23 Dec 2015 22:39:27 +0000 (14:39 -0800)]
openvswitch: Fix template leak in error cases.

Commit 5b48bb8506c5 ("openvswitch: Fix helper reference leak") fixed a
reference leak on helper objects, but inadvertently introduced a leak on
the ct template.

Previously, ct_info.ct->general.use was initialized to 0 by
nf_ct_tmpl_alloc() and only incremented when ovs_ct_copy_action()
returned successful. If an error occurred while adding the helper or
adding the action to the actions buffer, the __ovs_ct_free_action()
cleanup would use nf_ct_put() to free the entry; However, this relies on
atomic_dec_and_test(ct_info.ct->general.use). This reference must be
incremented first, or nf_ct_put() will never free it.

Fix the issue by acquiring a reference to the template immediately after
allocation.

Fixes: cae3a2627520 ("openvswitch: Allow attaching helpers to ct action")
Fixes: 5b48bb8506c5 ("openvswitch: Fix helper reference leak")
Signed-off-by: Joe Stringer <joe@ovn.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years ago[PATCH] arm: fix handling of F_OFD_... in oabi_fcntl64()
Al Viro [Tue, 29 Dec 2015 01:47:08 +0000 (20:47 -0500)]
[PATCH] arm: fix handling of F_OFD_... in oabi_fcntl64()

Cc: stable@vger.kernel.org # 3.15+
Reviewed-by: Jeff Layton <jeff.layton@primarydata.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
8 years agolightnvm: wrong offset in bad blk lun calculation
Matias Bjørling [Tue, 29 Dec 2015 13:37:56 +0000 (14:37 +0100)]
lightnvm: wrong offset in bad blk lun calculation

dev->nr_luns reports the total number of luns available in a device
while dev->luns_per_chnl is the number of luns per channel.

When multiple channels are available, the offset is calculated from a
channel and lun id into a linear array. As it multiplies with
the total number of luns, we go out of bound when channel id > 0 and
causes the kernel to panic when we read a protected kernel memory area.

Signed-off-by: Matias Bjørling <m@bjorling.me>
Signed-off-by: Jens Axboe <axboe@fb.com>
8 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma
Linus Torvalds [Tue, 29 Dec 2015 00:45:14 +0000 (16:45 -0800)]
Merge tag 'for-linus' of git://git./linux/kernel/git/dledford/rdma

Pull rdma fixes from Doug Ledford:
 "Three late 4.4-rc fixes.

  The first two were very small in terms of number of lines, the third
  is more lines of change than I like this late in the cycle, but there
  are positive test results from Avagotech and from my own test setup
  with the target hardware, and given the problem was a 100% failure
  case, I sent it through.

   - A previous patch updated the mlx4 driver to use vmalloc when there
     was not enough memory to get a contiguous region large enough for
     our needs, so we need kvfree() whenever we free that item.  We
     missed one place, so fix that now.

   - A previous patch added code to match incoming packets against a
     specific device, but failed to compensate for devices that have
     both InfiniBand and Ethernet ports.  Fix that.

   - Under certain vlan conditions, the ocrdma driver would fail to
     bring up any vlan interfaces and would print out a circular locking
     failure.  Fix that"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma:
  RDMA/be2net: Remove open and close entry points
  RDMA/ocrdma: Depend on async link events from CNA
  RDMA/ocrdma: Dispatch only port event when port state changes
  RDMA/ocrdma: Fix vlan-id assignment in qp parameters
  IB/mlx4: Replace kfree with kvfree in mlx4_ib_destroy_srq
  IB/cma: cma_match_net_dev needs to take into account port_num

8 years agonull_blk: use async queue restart helper
Jens Axboe [Mon, 28 Dec 2015 20:02:47 +0000 (13:02 -0700)]
null_blk: use async queue restart helper

If null_blk is run in NULL_IRQ_TIMER mode and with queue_mode NULL_Q_RQ,
we need to restart the queue from the hrtimer interrupt. We can't
directly invoke the request_fn from that context, so punt the queue run
to async kblockd context.

Tested-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Jens Axboe <axboe@fb.com>
8 years agoblock: add blk_start_queue_async()
Jens Axboe [Mon, 28 Dec 2015 20:01:22 +0000 (13:01 -0700)]
block: add blk_start_queue_async()

We currently only have an inline/sync helper to restart a stopped
queue. If drivers need an async version, they have to roll their
own. Add a generic helper instead.

Signed-off-by: Jens Axboe <axboe@fb.com>
8 years agoMerge branch 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6
Linus Torvalds [Mon, 28 Dec 2015 18:44:41 +0000 (10:44 -0800)]
Merge branch 'linus' of git://git./linux/kernel/git/herbert/crypto-2.6

Pull crypto fix from Herbert Xu:
 "This fixes a bug in the algif_skcipher interface that can trigger a
  kernel WARN_ON from user-space.  It does so by using the new skcipher
  interface which unlike the previous ablkcipher does not need to create
  extra geniv objects which is what was used to trigger the WARN_ON"

* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
  crypto: algif_skcipher - Use new skcipher interface

8 years agoMerge branch 'for-linus2' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris...
Linus Torvalds [Mon, 28 Dec 2015 18:35:19 +0000 (10:35 -0800)]
Merge branch 'for-linus2' of git://git./linux/kernel/git/jmorris/linux-security

Pull key handling bugfix from James Morris:
 "Fix a race between keyctl_read() and keyctl_revoke()"

* 'for-linus2' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security:
  KEYS: Fix race between read and revoke

8 years agoRDMA/be2net: Remove open and close entry points
Devesh Sharma [Thu, 24 Dec 2015 18:14:08 +0000 (13:14 -0500)]
RDMA/be2net: Remove open and close entry points

Recently Dough Ledford reported a deadlock happening
between ocrdma-load sequence and NetworkManager service
issueing "open" on be2net interface.

The deadlock happens when any be2net hook (e.g. open/close) is called
in parallel to insmod ocrdma.ko.

A. be2net is sending administrative open/close event to ocrdma holding
   device_list_mutex. It does this from ndo_open/ndo_stop hooks of be2net.
   So sequence of locks is rtnl_lock---> device_list lock

B.  When new ocrdma roce device gets registered, infiniband stack now
    takes rtnl_lock in ib_register_device() in GID initialization routines.
    So sequence of locks in this path is device_list lock ---> rtnl_lock.

This improper locking sequence causes deadlock.

In order to resolve the above deadlock condition, ocrdma intorduced a
patch to stop listening to administrative open/close events generated from
be2net driver. It now depends on link-state-change async-event generated from
CNA. This change leaves behind dead code which used to generate administrative
open/close events. This patch cleans-up all that dead code from be2net.

Reported-by: Doug Ledford <dledford@redhat.com>
CC: Sathya Perla <sathya.perla@avagotech.com>
Signed-off-by: Padmanabh Ratnakar <padmanabh.ratnakar@avagotech.com>
Signed-off-by: Selvin Xavier <selvin.xavier@avagotech.com>
Signed-off-by: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoRDMA/ocrdma: Depend on async link events from CNA
Devesh Sharma [Thu, 24 Dec 2015 18:14:07 +0000 (13:14 -0500)]
RDMA/ocrdma: Depend on async link events from CNA

Recently Dough Ledford reported a deadlock happening
between ocrdma-load sequence and NetworkManager service
issuing "open" on be2net interface.

The deadlock happens when any be2net hook (e.g. open/close) is called
in parallel to insmod ocrdma.ko.

A. be2net is sending administrative open/close event to ocrdma holding
   device_list_mutex. It does this from ndo_open/ndo_stop hooks of be2net.
   So sequence of locks is rtnl_lock---> device_list lock

B.  When new ocrdma roce device gets registered, infiniband stack now
    takes rtnl_lock in ib_register_device() in GID initialization routines.
    So sequence of locks in this path is device_list lock ---> rtnl_lock.

This improper locking sequence causes deadlock.

With this patch we stop using administrative open and close events
injected by be2net driver. These events were used to dispatch PORT_ACTIVE
and PORT_ERROR events to the IB-stack. This patch implements a logic
to receive async-link-events generated from CNA whenever link-state-change
is detected. Now on, these async-events will be used to dispatch
PORT_ACTIVE and PORT_ERROR events to IB-stack.

Depending on async-events from CNA removes the need to hold device-list-mutex
and thus breaks the busy-wait scenario.

Reported-by: Doug Ledford <dledford@redhat.com>
CC: Sathya Perla <sathya.perla@avagotech.com>
Signed-off-by: Padmanabh Ratnakar <padmanabh.ratnakar@avagotech.com>
Signed-off-by: Selvin Xavier <selvin.xavier@avagotech.com>
Signed-off-by: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoRDMA/ocrdma: Dispatch only port event when port state changes
Devesh Sharma [Thu, 24 Dec 2015 18:14:06 +0000 (13:14 -0500)]
RDMA/ocrdma: Dispatch only port event when port state changes

Dispatch only port event to IB stack when port state changes.
Don't explicitly modify qps to error. Let application listen to
port events on async event queue or let QP fail with retry-exceeded
completion error.

Signed-off-by: Padmanabh Ratnakar <padmanabh.ratnakar@avagotech.com>
Signed-off-by: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>