GitHub/moto-9609/android_kernel_motorola_exynos9610.git
9 years agobus: sunxi-rsb: Add Allwinner Reduced Serial Bus (RSB) controller bindings
Chen-Yu Tsai [Fri, 23 Oct 2015 18:41:30 +0000 (20:41 +0200)]
bus: sunxi-rsb: Add Allwinner Reduced Serial Bus (RSB) controller bindings

Reduced Serial Bus is a proprietary 2-line push-pull serial bus supporting
multiple slave devices. It was developed by Allwinner, Inc. and used by
Allwinner and X-Powers, Inc. for their line of PMICs and other peripheral
ICs.

Recent Allwinner SoCs, starting with the A23, have an RSB controller. This
is used to talk to the PMIC, and later with the A80 and A83 platform, the
audio codec IC.

Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Olof Johansson <olof@lixom.net>
9 years agoMerge tag 'v4.4-rockchip-drivers2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Olof Johansson [Fri, 23 Oct 2015 17:18:34 +0000 (10:18 -0700)]
Merge tag 'v4.4-rockchip-drivers2' of git://git./linux/kernel/git/mmind/linux-rockchip into next/drivers

Some fixes for the new power-domain driver, including restricting
Rockchip Kconfig options, fixing a possible null-pointer and a
typo in the dt-bindings.

* tag 'v4.4-rockchip-drivers2' of git://git.kernel.org/pub/scm/linux/kernel/git/mmind/linux-rockchip:
  dt-bindings: Correct paths in Rockchip power domains binding document
  soc: rockchip: power-domain: don't try to print the clock name in error case
  soc: rockchip: Restrict to ARCH_ROCKCHIP

Signed-off-by: Olof Johansson <olof@lixom.net>
9 years agoMerge branch 'drivers/psci2' into next/drivers
Olof Johansson [Fri, 23 Oct 2015 16:57:16 +0000 (09:57 -0700)]
Merge branch 'drivers/psci2' into next/drivers

* drivers/psci2:
  drivers: psci: make PSCI 1.0 functions initialization version dependent

Signed-off-by: Olof Johansson <olof@lixom.net>
9 years agodrivers: psci: make PSCI 1.0 functions initialization version dependent
Lorenzo Pieralisi [Fri, 23 Oct 2015 14:46:50 +0000 (15:46 +0100)]
drivers: psci: make PSCI 1.0 functions initialization version dependent

The PSCI specifications [1] and the SMC calling convention mandate
that unimplemented functions ids must return NOT_SUPPORTED (0xffffffff)
if a function id is called but it is not implemented.

Consequently, PSCI 1.0 function ids that require the 1.0 PSCI_FEATURES
call to be initialized:

CPU_SUSPEND (psci_init_cpu_suspend())
SYSTEM_SUSPEND (psci_init_system_suspend())

call the PSCI_FEATURES function id independently of the detected
PSCI firmware version, since, if the PSCI_FEATURES function id is not
implemented, it must return NOT_SUPPORTED according to the PSCI
specifications, causing the initialization functions to fail as expected.

Some existing PSCI implementations (ie Qemu PSCI emulation), do not
comply with the SMC calling convention and fail if function ids that are
not implemented are called from the OS, causing boot failures.

To solve this issue, this patch adds code that checks the PSCI firmware
version before calling PSCI 1.0 initialization functions so that the
OS makes sure that it is calling 1.0 functions only if the firmware
version detected is 1.0 or greater, therefore avoiding PSCI calls
that are bound to fail and might cause system boot failures owing
to non-compliant PSCI firmware implementations.

[1] http://infocenter.arm.com/help/topic/com.arm.doc.den0022c/DEN0022C_Power_State_Coordination_Interface.pdf

Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Mark Rutland <mark.rutland@arm.com>
Tested-by: Kevin Hilman <khilman@kernel.org>
Acked-by: Sudeep Holla <sudeep.holla@arm.com>
Signed-off-by: Olof Johansson <olof@lixom.net>
9 years agoMerge tag 'firmware/psci-1.0' of git://git.kernel.org/pub/scm/linux/kernel/git/lpiera...
Olof Johansson [Thu, 22 Oct 2015 17:02:10 +0000 (10:02 -0700)]
Merge tag 'firmware/psci-1.0' of git://git./linux/kernel/git/lpieralisi/linux into next/drivers

This pull request contains patches that enable PSCI 1.0 firmware
features for arm/arm64 platforms:

- Lorenzo Pieralisi adds support for the PSCI_FEATURES call, manages
  various 1.0 specifications updates (power state id and functions return
  values) and provides PSCI v1.0 DT bindings
- Sudeep Holla implements PSCI v1.0 system suspend support to enable PSCI
  based suspend-to-RAM

* tag 'firmware/psci-1.0' of git://git.kernel.org/pub/scm/linux/kernel/git/lpieralisi/linux:
  drivers: firmware: psci: add system suspend support
  drivers: firmware: psci: define more generic PSCI_FN_NATIVE macro
  drivers: firmware: psci: add PSCI v1.0 DT bindings
  drivers: firmware: psci: add extended stateid power_state support
  drivers: firmware: psci: add PSCI_FEATURES call
  drivers: firmware: psci: move power_state handling to generic code
  drivers: firmware: psci: add INVALID_ADDRESS return value

Signed-off-by: Olof Johansson <olof@lixom.net>
9 years agodt-bindings: Correct paths in Rockchip power domains binding document
Geert Uytterhoeven [Fri, 16 Oct 2015 14:11:41 +0000 (16:11 +0200)]
dt-bindings: Correct paths in Rockchip power domains binding document

Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
9 years agosoc: rockchip: power-domain: don't try to print the clock name in error case
Heiko Stuebner [Thu, 15 Oct 2015 12:35:01 +0000 (14:35 +0200)]
soc: rockchip: power-domain: don't try to print the clock name in error case

When we never got the the clock-reference, i.e. when IS_ERR(clk) is true,
don't try to print the clock name via %pC as this of course produces a
null-pointer-dereference in __clk_get_name().

Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Reviewed-by: Caesar Wang <wxt@rock-chips.com>
9 years agosoc: qcom/smem: add HWSPINLOCK dependency
Arnd Bergmann [Tue, 13 Oct 2015 15:05:39 +0000 (17:05 +0200)]
soc: qcom/smem: add HWSPINLOCK dependency

This fixes a build error when smem is enabled without hwspinlock:

drivers/built-in.o: In function `qcom_smem_alloc':
rockchip-efuse.c:(.text+0x7a3e4): undefined reference to `__hwspin_lock_timeout'
rockchip-efuse.c:(.text+0x7a568): undefined reference to `__hwspin_unlock'
drivers/built-in.o: In function `qcom_smem_remove':
rockchip-efuse.c:(.text+0x7a5cc): undefined reference to `hwspin_lock_free'
drivers/built-in.o: In function `qcom_smem_probe':
rockchip-efuse.c:(.text+0x7a960): undefined reference to `hwspin_lock_request_specific'
rockchip-efuse.c:(.text+0x7a988): undefined reference to `of_hwspin_lock_get_id'
drivers/built-in.o: In function `qcom_smem_get':
rockchip-efuse.c:(.text+0x7aa24): undefined reference to `__hwspin_lock_timeout'
rockchip-efuse.c:(.text+0x7aafc): undefined reference to `__hwspin_unlock'

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
9 years agoMerge tag 'qcom-soc-for-4.4' of git://codeaurora.org/quic/kernel/agross-msm into...
Arnd Bergmann [Thu, 15 Oct 2015 21:03:24 +0000 (23:03 +0200)]
Merge tag 'qcom-soc-for-4.4' of git://codeaurora.org/quic/kernel/agross-msm into next/drivers

Pull "Qualcomm ARM Based SoC Updates for 4.4" from Andy Gross:

* Implement id_table driver matching in SMD
* Avoid NULL pointer exception on remove of SMEM
* Reorder SMEM/SMD configs
* Make qcom_smem_get() return a pointer
* Handle big endian CPUs correctly in SMEM
* Represent SMD channel layout in structures
* Use __iowrite32_copy() in SMD
* Remove use of VLAIs in SMD
* Handle big endian CPUs correctly in SMD/RPM
* Handle big endian CPUs corretly in SMD
* Reject sending SMD packets that are too large
* Fix endianness issue in SCM __qcom_scm_is_call_available
* Add missing prototype for qcom_scm_is_available()
* Correct SMEM items for upper channels
* Use architecture level to build SCM correctly
* Delete unneeded of_node_put in SMD
* Correct active/slep state flagging in SMD/RPM
* Move RPM message ram out of SMEM DT node

* tag 'qcom-soc-for-4.4' of git://codeaurora.org/quic/kernel/agross-msm:
  soc: qcom: smem: Move RPM message ram out of smem DT node
  soc: qcom: smd-rpm: Correct the active vs sleep state flagging
  soc: qcom: smd: delete unneeded of_node_put
  firmware: qcom-scm: build for correct architecture level
  soc: qcom: smd: Correct SMEM items for upper channels
  qcom-scm: add missing prototype for qcom_scm_is_available()
  qcom-scm: fix endianess issue in __qcom_scm_is_call_available
  soc: qcom: smd: Reject send of too big packets
  soc: qcom: smd: Handle big endian CPUs
  soc: qcom: smd_rpm: Handle big endian CPUs
  soc: qcom: smd: Remove use of VLAIS
  soc: qcom: smd: Use __iowrite32_copy() instead of open-coding it
  soc: qcom: smd: Represent channel layout in structures
  soc: qcom: smem: Handle big endian CPUs
  soc: qcom: Make qcom_smem_get() return a pointer
  soc: qcom: Reorder SMEM/SMD configs
  soc: qcom: smem: Avoid NULL pointer exception on remove
  soc: qcom: smd: Implement id_table driver matching

9 years agoMerge tag 'berlin-new-cpuclk-for-4.4-1' of git://git.infradead.org/users/hesselba...
Arnd Bergmann [Thu, 15 Oct 2015 20:41:04 +0000 (22:41 +0200)]
Merge tag 'berlin-new-cpuclk-for-4.4-1' of git://git.infradead.org/users/hesselba/linux-berlin into next/drivers

Merge "Marvell Berlin BG2Q CPU clock driver" from Sebastian Hesselbarth:

- add BG2Q CPU clock to clk driver

* tag 'berlin-new-cpuclk-for-4.4-1' of git://git.infradead.org/users/hesselba/linux-berlin:
  clk: berlin: add cpuclk
  ARM: berlin: dts: add CLKID_CPU for BG2Q

9 years agoMerge tag 'drivers_pl172_for_4.4' of https://github.com/manabian/linux-lpc into next...
Arnd Bergmann [Thu, 15 Oct 2015 20:26:03 +0000 (22:26 +0200)]
Merge tag 'drivers_pl172_for_4.4' of https://github.com/manabian/linux-lpc into next/drivers

Merge "PL172 driver updates for v4.4" from Joachim Eastwood:

Support for additional ARM MPMCs to the PL172 driver and an update to
the bindings documentation to reflect this from Vladimir Zapolskiy.

"The change adds support of ARM PrimeCell PL175 MPMC and PL176 MPMC,
 the static memory controllers on devices are similar to one found on
 ARM PrimeCell PL172, add support to the existing driver."

* tag 'drivers_pl172_for_4.4' of https://github.com/manabian/linux-lpc:
  doc: dt: arm,pl172: add description of PL175 and PL176 controllers
  memory: pl172: add ARM PrimeCell PL176 MPMC support
  memory: pl172: add ARM PrimeCell PL175 MPMC support
  memory: pl172: correct MPMC peripheral ID register bits

9 years agoclk: berlin: add cpuclk
Antoine Tenart [Thu, 15 Oct 2015 18:55:55 +0000 (20:55 +0200)]
clk: berlin: add cpuclk

Add cpuclk in the Berlin BG2Q clock driver. This clk has a divider
fixed to 1.

Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Acked-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
9 years agoMerge tag 'berlin-dt-cpuclk-for-4.4-1' into berlin/cpuclk
Sebastian Hesselbarth [Thu, 15 Oct 2015 18:49:33 +0000 (20:49 +0200)]
Merge tag 'berlin-dt-cpuclk-for-4.4-1' into berlin/cpuclk

Marvell Berlin DT CPU clock for 4.4
- add missing CLKID_CPU for Berlin BG2Q

9 years agoARM: berlin: dts: add CLKID_CPU for BG2Q
Sebastian Hesselbarth [Thu, 15 Oct 2015 18:28:46 +0000 (20:28 +0200)]
ARM: berlin: dts: add CLKID_CPU for BG2Q

Marvell Berlin BG2Q SoC also has a clock for the CPU, add a
corresponding CLKID to the dt-binding include.

Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Signed-off-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
9 years agosoc: qcom: smem: Move RPM message ram out of smem DT node
Stephen Boyd [Thu, 8 Oct 2015 18:34:09 +0000 (13:34 -0500)]
soc: qcom: smem: Move RPM message ram out of smem DT node

SMEM is a software construct built on top of a DDR reserved region
and sometimes a device memory region called RPM message ram. Having
the RPM message ram in the smem DT node's reg property leads to the
smem node being located in different places depending on if the
message ram is being used or not. Let's add a qcom specific
property, qcom,rpm-msg-ram, and point to the device memory from
the SMEM node via a phandle.  As SMEM is a software construct, it
really needs to reside at the root of the DT regardless of whether
it's using the message ram or not.

Cc: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Andy Gross <agross@codeaurora.org>
9 years agosoc: qcom: smd-rpm: Correct the active vs sleep state flagging
Bjorn Andersson [Tue, 13 Oct 2015 20:57:43 +0000 (13:57 -0700)]
soc: qcom: smd-rpm: Correct the active vs sleep state flagging

The BIT() was incorrectly inherited from family A and should not be used
on family B where the state is denoted by an enum.

Reported-by: Georgi Djakov <georgi.djakov@linaro.org>
Signed-off-by: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Tested-by: Georgi Djakov <georgi.djakov@linaro.org>
Signed-off-by: Andy Gross <agross@codeaurora.org>
9 years agosoc: qcom: smd: delete unneeded of_node_put
Julia Lawall [Mon, 12 Oct 2015 20:43:15 +0000 (22:43 +0200)]
soc: qcom: smd: delete unneeded of_node_put

Device node iterators perform an of_node_put on each iteration, so putting
an of_node_put before a continue results in a double put.

A simplified version of the semantic match that finds this problem is as
follows (http://coccinelle.lip6.fr):

// <smpl>
@@
expression root,e;
local idexpression child;
iterator i;
@@

 i(..., child, ...) {
   ... when != of_node_get(child)
*  of_node_put(child);
   ...
*  continue;
}
// </smpl>

Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: Andy Gross <agross@codeaurora.org>
9 years agofirmware: qcom-scm: build for correct architecture level
Arnd Bergmann [Mon, 12 Oct 2015 14:56:19 +0000 (16:56 +0200)]
firmware: qcom-scm: build for correct architecture level

The ".arch_extension sec" directive is only available on ARMv6 or higher,
so if we enable the SCM driver while building a kernel for an older CPU,
we get a build error:

/tmp/ccUyhMOY.s:130: Error: selected processor does not support ARM mode `smc #0'
/tmp/ccUyhMOY.s:216: Error: selected processor does not support ARM mode `smc #0'
/tmp/ccUyhMOY.s:373: Error: selected processor does not support ARM mode `smc #0'
make[4]: *** [drivers/firmware/qcom_scm-32.o] Error 1

This changes the Makefile so we pass the ARMv7 architecture level both
for the check and for the actual compilation of the scm driver.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andy Gross <agross@codeaurora.org>
9 years agosoc: qcom: smd: Correct SMEM items for upper channels
Bjorn Andersson [Sat, 10 Oct 2015 05:06:44 +0000 (22:06 -0700)]
soc: qcom: smd: Correct SMEM items for upper channels

Update the SMEM items for the second set of SMD channels, as these where
incorrect.

Signed-off-by: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Andy Gross <agross@codeaurora.org>
9 years agoqcom-scm: add missing prototype for qcom_scm_is_available()
Rob Clark [Tue, 29 Sep 2015 19:48:55 +0000 (15:48 -0400)]
qcom-scm: add missing prototype for qcom_scm_is_available()

Signed-off-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Andy Gross <agross@codeaurora.org>
9 years agoqcom-scm: fix endianess issue in __qcom_scm_is_call_available
Rob Clark [Tue, 29 Sep 2015 19:48:53 +0000 (15:48 -0400)]
qcom-scm: fix endianess issue in __qcom_scm_is_call_available

Signed-off-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Andy Gross <agross@codeaurora.org>
9 years agosoc: qcom: smd: Reject send of too big packets
Bjorn Andersson [Fri, 25 Sep 2015 01:37:18 +0000 (18:37 -0700)]
soc: qcom: smd: Reject send of too big packets

Attempting to find room for a packet that's bigger than the fifo will
never succeed and the calling process will be sleeping forever in the
loop, waiting for enough room. So fail early instead.

Reported-by: Courtney Cavin <courtney.cavin@sonymobile.com>
Signed-off-by: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Reviewed-by: Andy Gross <agross@codeaurora.org>
Signed-off-by: Andy Gross <agross@codeaurora.org>
9 years agosoc: qcom: smd: Handle big endian CPUs
Stephen Boyd [Thu, 17 Sep 2015 21:50:53 +0000 (14:50 -0700)]
soc: qcom: smd: Handle big endian CPUs

The smd structures are always in little endian, but the smd
driver is not capable of being used on big endian CPUs. Annotate
the little endian data members and update the code to do the
proper byte swapping.

Cc: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Andy Gross <agross@codeaurora.org>
9 years agosoc: qcom: smd_rpm: Handle big endian CPUs
Stephen Boyd [Wed, 2 Sep 2015 22:46:50 +0000 (15:46 -0700)]
soc: qcom: smd_rpm: Handle big endian CPUs

The smd rpm structures are always in little endian, but this
driver is not capable of being used on big endian CPUs. Annotate
the little endian data members and update the code to do the
proper byte swapping.

Cc: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Reviewed-by: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Andy Gross <agross@codeaurora.org>
9 years agosoc: qcom: smd: Remove use of VLAIS
Stephen Boyd [Wed, 2 Sep 2015 22:46:48 +0000 (15:46 -0700)]
soc: qcom: smd: Remove use of VLAIS

Usage of VLAIS prevents clang from compiling this file, and it
also opens us to the possibility of allocating a large structure
on the stack to the point that we blow past the limit of the
kernel stack. Remove the VLAIS and allocate a structure on the
heap with kmalloc so that we're safer and more clang friendly.

Cc: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Reviewed-by: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Andy Gross <agross@codeaurora.org>
9 years agosoc: qcom: smd: Use __iowrite32_copy() instead of open-coding it
Stephen Boyd [Wed, 2 Sep 2015 22:46:47 +0000 (15:46 -0700)]
soc: qcom: smd: Use __iowrite32_copy() instead of open-coding it

We already have a function to do this and it silences some sparse
warnings along the way.

Reviewed-by: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Andy Gross <agross@codeaurora.org>
9 years agosoc: qcom: smd: Represent channel layout in structures
Stephen Boyd [Wed, 2 Sep 2015 22:46:46 +0000 (15:46 -0700)]
soc: qcom: smd: Represent channel layout in structures

The rx and tx channel info are laid out in memory next to each
other, and there are two types of channel info structures, byte
based and word based. We have 4 pointers to these info
structures, when we really only need two to point to the
different types of structures. Encapsulate the byte based and
word based tx/rx structures in a "channel pair" structure that
describes the layout of memory and reduces the number of pointers
in the smd channel structure by two.

Reviewed-by: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Andy Gross <agross@codeaurora.org>
9 years agosoc: qcom: smem: Handle big endian CPUs
Stephen Boyd [Wed, 2 Sep 2015 22:46:45 +0000 (15:46 -0700)]
soc: qcom: smem: Handle big endian CPUs

The contents of smem are always in little endian, but the smem
driver is not capable of being used on big endian CPUs. Annotate
the little endian data members and update the code to do the
proper byte swapping.

Cc: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Reviewed-by: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Andy Gross <agross@codeaurora.org>
9 years agosoc: qcom: Make qcom_smem_get() return a pointer
Stephen Boyd [Wed, 2 Sep 2015 22:46:44 +0000 (15:46 -0700)]
soc: qcom: Make qcom_smem_get() return a pointer

Passing a void ** almost always requires a cast at the call site.
Instead of littering the code with casts every time this function
is called, have qcom_smem_get() return a void pointer to the
location of the smem item. This frees the caller from having to
cast the pointer.

Cc: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Reviewed-by: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Andy Gross <agross@codeaurora.org>
9 years agosoc: qcom: Reorder SMEM/SMD configs
Stephen Boyd [Fri, 28 Aug 2015 18:23:33 +0000 (11:23 -0700)]
soc: qcom: Reorder SMEM/SMD configs

When I make nconfig, having the SMEM option after the SMD option
causes the configurator to get confused when I'm enabling and
disabling these options. Let's move SMEM before SMD so there's a
clear indented dependency chain.

Cc: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Reviewed-by: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Andy Gross <agross@codeaurora.org>
9 years agosoc: qcom: smem: Avoid NULL pointer exception on remove
Stephen Boyd [Fri, 28 Aug 2015 18:18:49 +0000 (11:18 -0700)]
soc: qcom: smem: Avoid NULL pointer exception on remove

Don't set a pointer to NULL and then dereference it in the next
line.

Cc: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Reviewed-by: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Andy Gross <agross@codeaurora.org>
9 years agosoc: qcom: smd: Implement id_table driver matching
Bjorn Andersson [Fri, 28 Aug 2015 17:39:20 +0000 (10:39 -0700)]
soc: qcom: smd: Implement id_table driver matching

Implement a id_table based driver maching mechanism for drivers that
binds to fixed channels and doesn't need any additional configuration,
e.g. IPCRTR and DIAG.

Signed-off-by: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Andy Gross <agross@codeaurora.org>
9 years agoMerge tag 'arm-scpi-for-v4.4' of git://git.kernel.org/pub/scm/linux/kernel/git/sudeep...
Arnd Bergmann [Wed, 14 Oct 2015 15:07:32 +0000 (17:07 +0200)]
Merge tag 'arm-scpi-for-v4.4' of git://git./linux/kernel/git/sudeep.holla/linux into next/drivers

Merge "ARM System Control and Power Interface(SCPI) support" from Sudeep Holla

It adds support for the following features provided by SCP firmware
using different subsystems in Linux:
  1. SCPI mailbox protocol driver which using mailbox framework
  2. Clocks provided by SCP using clock framework
  3. CPU DVFS(cpufreq) using existing arm-big-little driver
  4. SCPI based sensors including temperature sensors

* tag 'arm-scpi-for-v4.4' of git://git.kernel.org/pub/scm/linux/kernel/git/sudeep.holla/linux:
  hwmon: Support thermal zones registration for SCP temperature sensors
  hwmon: Support sensors exported via ARM SCP interface
  firmware: arm_scpi: Extend to support sensors
  Documentation: add DT bindings for ARM SCPI sensors
  cpufreq: arm_big_little: add SCPI interface driver
  clk: scpi: add support for cpufreq virtual device
  clk: add support for clocks provided by SCP(System Control Processor)
  firmware: add support for ARM System Control and Power Interface(SCPI) protocol
  Documentation: add DT binding for ARM System Control and Power Interface(SCPI) protocol

9 years agosoc: rockchip: Restrict to ARCH_ROCKCHIP
Thierry Reding [Fri, 9 Oct 2015 13:46:18 +0000 (15:46 +0200)]
soc: rockchip: Restrict to ARCH_ROCKCHIP

By definition this directory contains drivers that are specific to the
Rockchip architecture. All Kconfig options should therefore depend on
ARCH_ROCKCHIP to avoid exposing these symbols on other architectures.
For example, this options currently shows up as new when doing an
incremental build on PowerPC.

Signed-off-by: Thierry Reding <treding@nvidia.com>
[add COMPILE_TEST alternative condition]
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
9 years agoMerge tag 'v4.4-rockchip-drivers1' of git://git.kernel.org/pub/scm/linux/kernel/git...
Arnd Bergmann [Fri, 9 Oct 2015 15:08:28 +0000 (17:08 +0200)]
Merge tag 'v4.4-rockchip-drivers1' of git://git./linux/kernel/git/mmind/linux-rockchip into next/drivers

Merge "Rockchip power-domain drivers for 4.4" from Heiko Stuebner:

Add the power-domain base-driver which currently contains
support for the rk3288 powerdomain layout but can be easily
extended for the socs (including arm64) later on.
A big thanks to Ceasar Wang for pulling through on this
during 18 revisions.
Also included is a fix to the pm-clock handling in the generic
powerdomains to adapt it to the per-user clock handling we now
do, Acked by Rafael Wysocki.

* tag 'v4.4-rockchip-drivers1' of git://git.kernel.org/pub/scm/linux/kernel/git/mmind/linux-rockchip:
  soc: rockchip: power-domain: Add power domain driver
  dt-bindings: add document of Rockchip power domains
  PM / clk: Do not __clk_get passed in clock-references
  dt-bindings: add power-domain header for RK3288 SoCs

9 years agohwmon: Support thermal zones registration for SCP temperature sensors
Punit Agrawal [Wed, 24 Jun 2015 18:03:16 +0000 (19:03 +0100)]
hwmon: Support thermal zones registration for SCP temperature sensors

Add support to create thermal zones based on the temperature sensors
provided by the SCP. The thermal zones can be defined using the
thermal DT bindings and should refer to the SCP sensor id to select
the sensor.

Signed-off-by: Punit Agrawal <punit.agrawal@arm.com>
Acked-by: Guenter Roeck <linux@roeck-us.net>
Cc: Eduardo Valentin <edubezval@gmail.com>
9 years agohwmon: Support sensors exported via ARM SCP interface
Punit Agrawal [Thu, 21 May 2015 14:08:45 +0000 (15:08 +0100)]
hwmon: Support sensors exported via ARM SCP interface

Create a driver to add support for SoC sensors exported by the System
Control Processor (SCP) via the System Control and Power Interface
(SCPI). The supported sensor types is one of voltage, temperature,
current, and power.

The sensor labels and values provided by the SCP are exported via the
hwmon sysfs interface.

Signed-off-by: Punit Agrawal <punit.agrawal@arm.com>
Acked-by: Guenter Roeck <linux@roeck-us.net>
Cc: Sudeep Holla <sudeep.holla@arm.com>
9 years agofirmware: arm_scpi: Extend to support sensors
Punit Agrawal [Fri, 19 Jun 2015 14:31:46 +0000 (15:31 +0100)]
firmware: arm_scpi: Extend to support sensors

ARM System Control Processor (SCP) provides an API to query and use
the sensors available in the system. Extend the SCPI driver to support
 sensor messages.

Signed-off-by: Punit Agrawal <punit.agrawal@arm.com>
Acked-by: Sudeep Holla <sudeep.holla@arm.com>
9 years agoDocumentation: add DT bindings for ARM SCPI sensors
Punit Agrawal [Thu, 2 Jul 2015 18:30:01 +0000 (19:30 +0100)]
Documentation: add DT bindings for ARM SCPI sensors

The System Control Processor (SCP) provides access to SoC sensors via
the System Control and Power Interface (SCPI) Message Protocol. Add
bindings to allow probing of these sensors. Also support referencing
of the sensors for setting up thermal zones via the thermal DT
bindings.

Signed-off-by: Punit Agrawal <punit.agrawal@arm.com>
Acked-by: Rob Herring <robh+dt@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Sudeep Holla <sudeep.holla@arm.com>
9 years agoMerge tag 'at91-cleanup-4.4' of git://git.kernel.org/pub/scm/linux/kernel/git/abellon...
Arnd Bergmann [Thu, 8 Oct 2015 15:26:27 +0000 (17:26 +0200)]
Merge tag 'at91-cleanup-4.4' of git://git./linux/kernel/git/abelloni/linux into next/drivers

Merge "First batch of cleanups for 4.4:" from Alexandre Belloni:
 - properly get the slow clock from timer-atmel-st, tcb_clksrc and pwm-atmel-tcb
 - small fix in an error path for tcb_clksrc

* tag 'at91-cleanup-4.4' of git://git.kernel.org/pub/scm/linux/kernel/git/abelloni/linux:
  misc: atmel_tclib: get and use slow clock
  clocksource: tcb_clksrc: fix setup_clkevents error path
  clocksource: atmel-st: get and use slow clock

9 years agoMerge tag 'qcom-fixes-for-4.3-rc1' into base-for-4.4
Andy Gross [Thu, 8 Oct 2015 04:13:49 +0000 (23:13 -0500)]
Merge tag 'qcom-fixes-for-4.3-rc1' into base-for-4.4

Qualcomm fixes for v4.3-rc1

* Add SCM function call stubs on ARM64

9 years agoMerge tag 'renesas-clk-for-v4.4' of git://git.kernel.org/pub/scm/linux/kernel/git...
Arnd Bergmann [Tue, 6 Oct 2015 14:46:41 +0000 (16:46 +0200)]
Merge tag 'renesas-clk-for-v4.4' of git://git./linux/kernel/git/horms/renesas into next/drivers

Merge "Renesas ARM Based SoC Clk Updates for v4.4" from Simon Horman:

* Consider "zb_clk" suitable for power management

  This part of a multi-stage effort by Geert Uytterhoeven to add:

  "Clock Domain support to the Clock Pulse Generator (CPG) Module Stop
   (MSTP) Clocks driver using the generic PM Domain, to be used on shmobile
   SoCs without device power domains (R-Car Gen1 and Gen2, RZ).  This allows
   to power-manage the module clocks of SoC devices that are part of the
   CPG/MSTP Clock Domain using Runtime PM, or for system suspend/resume,
   similar to SoCs with device power domains (SH-Mobile and R-Mobile)."

* tag 'renesas-clk-for-v4.4' of git://git.kernel.org/pub/scm/linux/kernel/git/horms/renesas:
  clk: shmobile: mstp: Consider "zb_clk" suitable for power management

9 years agomisc: atmel_tclib: get and use slow clock
Boris Brezillon [Sun, 16 Aug 2015 09:23:46 +0000 (11:23 +0200)]
misc: atmel_tclib: get and use slow clock

Commit dca1a4b5ff6e ("clk: at91: keep slow clk enabled to prevent system
hang") added a workaround for the slow clock as it is not properly handled
by its users.

Get and use the slow clock as it is necessary for the timer counters.

Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Thierry Reding <thierry.reding@gmail.com>
9 years agoclocksource: tcb_clksrc: fix setup_clkevents error path
Boris Brezillon [Sun, 16 Aug 2015 09:23:45 +0000 (11:23 +0200)]
clocksource: tcb_clksrc: fix setup_clkevents error path

t2_clk is already disabled before request_irq(), it must not be disabled
again.

Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: Daniel Lezcano <daniel.lezcano@linaro.org>
9 years agoclocksource: atmel-st: get and use slow clock
Alexandre Belloni [Sun, 16 Aug 2015 09:23:44 +0000 (11:23 +0200)]
clocksource: atmel-st: get and use slow clock

The current slow clock rate is hardcoded. Properly get the slow clock
and use its rate.

Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: Daniel Lezcano <daniel.lezcano@linaro.org>
9 years agosoc: rockchip: power-domain: Add power domain driver
Caesar Wang [Tue, 8 Sep 2015 06:18:22 +0000 (14:18 +0800)]
soc: rockchip: power-domain: Add power domain driver

This driver is found on RK3288 SoCs.

In order to meet high performance and low power requirements, a power
management unit is designed or saving power when RK3288 in low power
mode.
The RK3288 PMU is dedicated for managing the power of the whole chip.

PMU can work in the Low Power Mode by setting bit[0] of PMU_PWRMODE_CON
register. After setting the register, PMU would enter the Low Power mode.
In the low power mode, pmu will auto power on/off the specified power
domain, send idle req to specified power domain, shut down/up pll and
so on. All of above are configurable by setting corresponding registers.

Signed-off-by: Caesar Wang <wxt@rock-chips.com>
Reviewed-by: Kevin Hilman <khilman@linaro.org>
[replace dsb() with dsb(sy) for arm64 buildability; sy is the default,
so no functional change; adapt to per-user clocks in genpd]
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
9 years agodt-bindings: add document of Rockchip power domains
Caesar Wang [Tue, 8 Sep 2015 06:18:20 +0000 (14:18 +0800)]
dt-bindings: add document of Rockchip power domains

This add the necessary binding documentation for the power domains
found on Rockchip SoCs.

Signed-off-by: Caesar Wang <wxt@rock-chips.com>
Reviewed-by: Kevin Hilman <khilman@linaro.org>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
9 years agoMerge branch 'v4.4-armsoc/pd-headers' into v4.4-armsoc/drivers
Heiko Stuebner [Tue, 6 Oct 2015 07:47:26 +0000 (09:47 +0200)]
Merge branch 'v4.4-armsoc/pd-headers' into v4.4-armsoc/drivers

9 years agoPM / clk: Do not __clk_get passed in clock-references
Stephen Boyd [Thu, 1 Oct 2015 19:05:09 +0000 (21:05 +0200)]
PM / clk: Do not __clk_get passed in clock-references

Clock references are on a per-user basis now, so they are not supposed
to be refcounted by itself anymore. Therefore multiple cascaded get and
put calls will fail.

When a clock reference gets passed into pm_clk_add_clk we can assume
that the pm clock handling will take control of the clock reference,
so after this functions returns the caller should've given up control
of that handle.

So remove the additional call to __clk_get() in __pm_clk_add().

The only current user of pm_clk_add_clk is drivers/clk/shmobile/clk-mstp.c
which already follows this paradigm by only getting the clock but not
puting it after passing the reference into pm_clk_add_clk.

In the error case the caller is expected to clean up the clock, as it
may very well try to do something different if pm_clk_add_clk() fails.

Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
[add commit-message]
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
9 years agoLinux 4.3-rc4
Linus Torvalds [Sun, 4 Oct 2015 15:57:17 +0000 (16:57 +0100)]
Linux 4.3-rc4

9 years agoMerge branch 'strscpy' of git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf...
Linus Torvalds [Sun, 4 Oct 2015 15:31:13 +0000 (16:31 +0100)]
Merge branch 'strscpy' of git://git./linux/kernel/git/cmetcalf/linux-tile

Pull strscpy string copy function implementation from Chris Metcalf.

Chris sent this during the merge window, but I waffled back and forth on
the pull request, which is why it's going in only now.

The new "strscpy()" function is definitely easier to use and more secure
than either strncpy() or strlcpy(), both of which are horrible nasty
interfaces that have serious and irredeemable problems.

strncpy() has a useless return value, and doesn't NUL-terminate an
overlong result.  To make matters worse, it pads a short result with
zeroes, which is a performance disaster if you have big buffers.

strlcpy(), by contrast, is a mis-designed "fix" for strlcpy(), lacking
the insane NUL padding, but having a differently broken return value
which returns the original length of the source string.  Which means
that it will read characters past the count from the source buffer, and
you have to trust the source to be properly terminated.  It also makes
error handling fragile, since the test for overflow is unnecessarily
subtle.

strscpy() avoids both these problems, guaranteeing the NUL termination
(but not excessive padding) if the destination size wasn't zero, and
making the overflow condition very obvious by returning -E2BIG.  It also
doesn't read past the size of the source, and can thus be used for
untrusted source data too.

So why did I waffle about this for so long?

Every time we introduce a new-and-improved interface, people start doing
these interminable series of trivial conversion patches.

And every time that happens, somebody does some silly mistake, and the
conversion patch to the improved interface actually makes things worse.
Because the patch is mindnumbing and trivial, nobody has the attention
span to look at it carefully, and it's usually done over large swatches
of source code which means that not every conversion gets tested.

So I'm pulling the strscpy() support because it *is* a better interface.
But I will refuse to pull mindless conversion patches.  Use this in
places where it makes sense, but don't do trivial patches to fix things
that aren't actually known to be broken.

* 'strscpy' of git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile:
  tile: use global strscpy() rather than private copy
  string: provide strscpy()
  Make asm/word-at-a-time.h available on all architectures

9 years agoMerge tag 'md/4.3-fixes' of git://neil.brown.name/md
Linus Torvalds [Sun, 4 Oct 2015 10:47:28 +0000 (11:47 +0100)]
Merge tag 'md/4.3-fixes' of git://neil.brown.name/md

Pull md fixes from Neil Brown:
 "Assorted fixes for md in 4.3-rc.

  Two tagged for -stable, and one is really a cleanup to match and
  improve kmemcache interface.

* tag 'md/4.3-fixes' of git://neil.brown.name/md:
  md/bitmap: don't pass -1 to bitmap_storage_alloc.
  md/raid1: Avoid raid1 resync getting stuck
  md: drop null test before destroy functions
  md: clear CHANGE_PENDING in readonly array
  md/raid0: apply base queue limits *before* disk_stack_limits
  md/raid5: don't index beyond end of array in need_this_block().
  raid5: update analysis state for failed stripe
  md: wait for pending superblock updates before switching to read-only

9 years agoMerge branch 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus
Linus Torvalds [Sun, 4 Oct 2015 10:41:58 +0000 (11:41 +0100)]
Merge branch 'upstream' of git://git.linux-mips.org/ralf/upstream-linus

Pull MIPS updates from Ralf Baechle:
 "This week's round of MIPS fixes:
   - Fix JZ4740 build
   - Fix fallback to GFP_DMA
   - FP seccomp in case of ENOSYS
   - Fix bootmem panic
   - A number of FP and CPS fixes
   - Wire up new syscalls
   - Make sure BPF assembler objects can properly be disassembled
   - Fix BPF assembler code for MIPS I"

* 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus:
  MIPS: scall: Always run the seccomp syscall filters
  MIPS: Octeon: Fix kernel panic on startup from memory corruption
  MIPS: Fix R2300 FP context switch handling
  MIPS: Fix octeon FP context switch handling
  MIPS: BPF: Fix load delay slots.
  MIPS: BPF: Do all exports of symbols with FEXPORT().
  MIPS: Fix the build on jz4740 after removing the custom gpio.h
  MIPS: CPS: #ifdef on CONFIG_MIPS_MT_SMP rather than CONFIG_MIPS_MT
  MIPS: CPS: Don't include MT code in non-MT kernels.
  MIPS: CPS: Stop dangling delay slot from has_mt.
  MIPS: dma-default: Fix 32-bit fall back to GFP_DMA
  MIPS: Wire up userfaultfd and membarrier syscalls.

9 years agoMerge branch 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 4 Oct 2015 10:40:09 +0000 (11:40 +0100)]
Merge branch 'irq-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull irq fixes from Thomas Gleixner:
 "This update contains:

   - Fix for a long standing race affecting /proc/irq/NNN

   - One line fix for ARM GICV3-ITS counting the wrong data

   - Warning silencing in ARM GICV3-ITS.  Another GCC trying to be
     overly clever issue"

* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  irqchip/gic-v3-its: Count additional LPIs for the aliased devices
  irqchip/gic-v3-its: Silence warning when its_lpi_alloc_chunks gets inlined
  genirq: Fix race in register_irq_proc()

9 years agoMIPS: scall: Always run the seccomp syscall filters
Markos Chandras [Fri, 25 Sep 2015 07:17:42 +0000 (08:17 +0100)]
MIPS: scall: Always run the seccomp syscall filters

The MIPS syscall handler code used to return -ENOSYS on invalid
syscalls. Whilst this is expected, it caused problems for seccomp
filters because the said filters never had the change to run since
the code returned -ENOSYS before triggering them. This caused
problems on the chromium testsuite for filters looking for invalid
syscalls. This has now changed and the seccomp filters are always
run even if the syscall is invalid. We return -ENOSYS once we
return from the seccomp filters. Moreover, similar codepaths have
been merged in the process which simplifies somewhat the overall
syscall code.

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/11236/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
9 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 3 Oct 2015 14:53:05 +0000 (10:53 -0400)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Ingo Molnar:
 "Fixes all around the map: W+X kernel mapping fix, WCHAN fixes, two
  build failure fixes for corner case configs, x32 header fix and a
  speling fix"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/headers/uapi: Fix __BITS_PER_LONG value for x32 builds
  x86/mm: Set NX on gap between __ex_table and rodata
  x86/kexec: Fix kexec crash in syscall kexec_file_load()
  x86/process: Unify 32bit and 64bit implementations of get_wchan()
  x86/process: Add proper bound checks in 64bit get_wchan()
  x86, efi, kasan: Fix build failure on !KASAN && KMEMCHECK=y kernels
  x86/hyperv: Fix the build in the !CONFIG_KEXEC_CORE case
  x86/cpufeatures: Correct spelling of the HWP_NOTIFY flag

9 years agoMerge branch 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 3 Oct 2015 14:51:41 +0000 (10:51 -0400)]
Merge branch 'timers-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull timer fixes from Ingo Molnar:
 "An abs64() fix in the watchdog driver, and two clocksource driver
  NO_IRQ assumption fixes"

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  clocksource: Fix abs() usage w/ 64bit values
  clocksource/drivers/keystone: Fix bad NO_IRQ usage
  clocksource/drivers/rockchip: Fix bad NO_IRQ usage

9 years agoMerge branch 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 3 Oct 2015 14:46:41 +0000 (10:46 -0400)]
Merge branch 'core-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull EFI fixes from Ingo Molnar:
 "Two EFI fixes: one for x86, one for ARM, fixing a boot crash bug that
  can trigger under newer EFI firmware"

* 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  arm64/efi: Fix boot crash by not padding between EFI_MEMORY_RUNTIME regions
  x86/efi: Fix boot crash by mapping EFI memmap entries bottom-up at runtime, instead of top-down

9 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Sat, 3 Oct 2015 14:39:31 +0000 (10:39 -0400)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "Bunch of fixes all over the place, all pretty small: amdgpu, i915,
  exynos, one qxl and one vmwgfx.

  There is also a bunch of mst fixes, I left some cleanups in the series
  as I didn't think it was worth splitting up the tested series"

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux: (37 commits)
  drm/dp/mst: add some defines for logical/physical ports
  drm/dp/mst: drop cancel work sync in the mstb destroy path (v2)
  drm/dp/mst: split connector registration into two parts (v2)
  drm/dp/mst: update the link_address_sent before sending the link address (v3)
  drm/dp/mst: fixup handling hotplug on port removal.
  drm/dp/mst: don't pass port into the path builder function
  drm/radeon: drop radeon_fb_helper_set_par
  drm: handle cursor_set2 in restore_fbdev_mode
  drm/exynos: Staticize local function in exynos_drm_gem.c
  drm/exynos: fimd: actually disable dp clock
  drm/exynos: dp: remove suspend/resume functions
  drm/qxl: recreate the primary surface when the bo is not primary
  drm/amdgpu: only print meaningful VM faults
  drm/amdgpu/cgs: remove import_gpu_mem
  drm/i915: Call non-locking version of drm_kms_helper_poll_enable(), v2
  drm: Add a non-locking version of drm_kms_helper_poll_enable(), v2
  drm/vmwgfx: Fix a command submission hang regression
  drm/exynos: remove unused mode_fixup() code
  drm/exynos: remove decon_mode_fixup()
  drm/exynos: remove fimd_mode_fixup()
  ...

9 years agodt-bindings: add power-domain header for RK3288 SoCs
Caesar Wang [Tue, 8 Sep 2015 06:18:21 +0000 (14:18 +0800)]
dt-bindings: add power-domain header for RK3288 SoCs

According to a description from TRM, add all the power domains.

At the moment, we can support some domains on RK3288.
We can add more types on RK3288 in the future, that's need to do.

Signed-off-by: Caesar Wang <wxt@rock-chips.com>
Reviewed-by: Kevin Hilman <khilman@linaro.org>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
9 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Fri, 2 Oct 2015 21:53:25 +0000 (17:53 -0400)]
Merge branch 'for-linus' of git://git./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

9 years agoclocksource: Fix abs() usage w/ 64bit values
John Stultz [Tue, 15 Sep 2015 01:05:20 +0000 (18:05 -0700)]
clocksource: Fix abs() usage w/ 64bit values

This patch fixes one cases where abs() was being used with 64-bit
nanosecond values, where the result may be capped at 32-bits.

This potentially could cause watchdog false negatives on 32-bit
systems, so this patch addresses the issue by using abs64().

Signed-off-by: John Stultz <john.stultz@linaro.org>
Cc: Prarit Bhargava <prarit@redhat.com>
Cc: Richard Cochran <richardcochran@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: stable@vger.kernel.org
Link: http://lkml.kernel.org/r/1442279124-7309-2-git-send-email-john.stultz@linaro.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
9 years agoMerge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Linus Torvalds [Fri, 2 Oct 2015 18:54:16 +0000 (14:54 -0400)]
Merge tag 'arm64-fixes' of git://git./linux/kernel/git/arm64/linux

Pull arm64 fixes from Catalin Marinas:

 - Fix for transparent huge page change_protection() logic which was
   inadvertently changing a huge pmd page into a pmd table entry.

 - Function graph tracer panic fix caused by the return_to_handler code
   corrupting the multi-regs function return value (composite types).

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64: ftrace: fix function_graph tracer panic
  arm64: Fix THP protection change logic

9 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/geert/linux...
Linus Torvalds [Fri, 2 Oct 2015 18:51:46 +0000 (14:51 -0400)]
Merge branch 'for-linus' of git://git./linux/kernel/git/geert/linux-m68k

Pull m68k updates from Geert Uytterhoeven:
 "Summary:
   - Fix for accidental modification of arguments of syscall functions
   - Wire up new syscalls
   - Update defconfigs"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/geert/linux-m68k:
  m68k/defconfig: Update defconfigs for v4.3-rc1
  m68k: Define asmlinkage_protect
  m68k: Wire up membarrier
  m68k: Wire up userfaultfd
  m68k: Wire up direct socket calls

9 years agoirqchip/gic-v3-its: Count additional LPIs for the aliased devices
Marc Zyngier [Fri, 2 Oct 2015 15:44:06 +0000 (16:44 +0100)]
irqchip/gic-v3-its: Count additional LPIs for the aliased devices

When configuring the interrupt mapping for a new device, we
iterate over all the possible aliases to account for their
maximum MSI allocation. This was introduced by e8137f4f5088
("irqchip: gicv3-its: Iterate over PCI aliases to generate ITS configuration").

Turns out that the code doing that is a bit braindead, and repeatedly
accounts for the same device over and over.

Fix this by counting the actual alias that is passed to us by the
core code.

Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Cc: Alex Shi <alex.shi@linaro.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: David Daney <ddaney.cavm@gmail.com>
Cc: Jason Cooper <jason@lakedaemon.net>
Link: http://lkml.kernel.org/r/1443800646-8074-3-git-send-email-marc.zyngier@arm.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
9 years agoirqchip/gic-v3-its: Silence warning when its_lpi_alloc_chunks gets inlined
Marc Zyngier [Fri, 2 Oct 2015 15:44:05 +0000 (16:44 +0100)]
irqchip/gic-v3-its: Silence warning when its_lpi_alloc_chunks gets inlined

More agressive inlining in recent versions of GCC have uncovered
a new set of warnings:

 drivers/irqchip/irq-gic-v3-its.c: In function its_msi_prepare:
  drivers/irqchip/irq-gic-v3-its.c:1148:26: warning: lpi_base may be used
    uninitialized in this function [-Wmaybe-uninitialized]
     dev->event_map.lpi_base = lpi_base;
                          ^
 drivers/irqchip/irq-gic-v3-its.c:1116:6: note: lpi_base was declared here
  int lpi_base;
      ^
 drivers/irqchip/irq-gic-v3-its.c:1149:25: warning: nr_lpis may be used
  uninitialized in this function [-Wmaybe-uninitialized]
   dev->event_map.nr_lpis = nr_lpis;
                         ^
 drivers/irqchip/irq-gic-v3-its.c:1117:6: note: nr_lpis was declared here
  int nr_lpis;
      ^
The warning is fairly benign (there is no code path that could
actually use uninitialized variables), but let's silence it anyway
by zeroing the variables on the error path.

Reported-by: Alex Shi <alex.shi@linaro.org>
Tested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Cc: David Daney <ddaney.cavm@gmail.com>
Cc: Jason Cooper <jason@lakedaemon.net>
Link: http://lkml.kernel.org/r/1443800646-8074-2-git-send-email-marc.zyngier@arm.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
9 years agoMerge tag 'dmaengine-fix-4.3-rc4' of git://git.infradead.org/users/vkoul/slave-dma
Linus Torvalds [Fri, 2 Oct 2015 18:46:15 +0000 (14:46 -0400)]
Merge tag 'dmaengine-fix-4.3-rc4' of git://git.infradead.org/users/vkoul/slave-dma

Pull dmaengine fixes from Vinod Koul:
 "This contains fixes spread throughout the drivers, and also fixes one
  more instance of privatecnt in dmaengine.

  Driver fixes summary:
   - bunch of pxa_dma fixes for reuse of descriptor issue, residue and
     no-requestor
   - odd fixes in xgene, idma, sun4i and zxdma
   - at_xdmac fixes for cleaning descriptor and block addr mode"

* tag 'dmaengine-fix-4.3-rc4' of git://git.infradead.org/users/vkoul/slave-dma:
  dmaengine: pxa_dma: fix residue corner case
  dmaengine: pxa_dma: fix the no-requestor case
  dmaengine: zxdma: Fix off-by-one for testing valid pchan request
  dmaengine: at_xdmac: clean used descriptor
  dmaengine: at_xdmac: change block increment addressing mode
  dmaengine: dw: properly read DWC_PARAMS register
  dmaengine: xgene-dma: Fix overwritting DMA tx ring
  dmaengine: fix balance of privatecnt
  dmaengine: sun4i: fix unsafe list iteration
  dmaengine: idma64: improve residue estimation
  dmaengine: xgene-dma: fix handling xgene_dma_get_ring_size result
  dmaengine: pxa_dma: fix initial list move

9 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-block
Linus Torvalds [Fri, 2 Oct 2015 18:40:57 +0000 (14:40 -0400)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:
 "Another week, another round of fixes.

  These have been brewing for a bit and in various iterations, but I
  feel pretty comfortable about the quality of them.  They fix real
  issues.  The pull request is mostly blk-mq related, and the only one
  not fixing a real bug, is the tag iterator abstraction from Christoph.
  But it's pretty trivial, and we'll need it for another fix soon.

  Apart from the blk-mq fixes, there's an NVMe affinity fix from Keith,
  and a single fix for xen-blkback from Roger fixing failure to free
  requests on disconnect"

* 'for-linus' of git://git.kernel.dk/linux-block:
  blk-mq: factor out a helper to iterate all tags for a request_queue
  blk-mq: fix racy updates of rq->errors
  blk-mq: fix deadlock when reading cpu_list
  blk-mq: avoid inserting requests before establishing new mapping
  blk-mq: fix q->mq_usage_counter access race
  blk-mq: Fix use after of free q->mq_map
  blk-mq: fix sysfs registration/unregistration race
  blk-mq: avoid setting hctx->tags->cpumask before allocation
  NVMe: Set affinity after allocating request queues
  xen/blkback: free requests on disconnection

9 years agoRevert "Input: synaptics - fix handling of disabling gesture mode"
Dmitry Torokhov [Fri, 2 Oct 2015 17:31:32 +0000 (10:31 -0700)]
Revert "Input: synaptics - fix handling of disabling gesture mode"

This reverts commit e51e38494a8ecc18650efb0c840600637891de2c: we
actually do want the device to work in extended W mode, as this is the
mode that allows us receiving multiple contact information.

Cc: stable@vger.kernel.org
9 years agoMIPS: Octeon: Fix kernel panic on startup from memory corruption
Matt Bennett [Wed, 30 Sep 2015 04:40:42 +0000 (17:40 +1300)]
MIPS: Octeon: Fix kernel panic on startup from memory corruption

During development it was found that a number of builds would panic
during the kernel init process, more specifically in 'delayed_fput()'.
The panic showed the kernel trying to access a memory address of
'0xb7fdc00' while traversing the 'delayed_fput_list' structure.
Comparing this memory address to the value of the pointer used on
builds that did not panic confirmed that the pointer on crashing
builds must have been corrupted at some stage earlier in the init
process.

By traversing the list earlier and earlier in the code it was found
that 'plat_mem_setup()' was responsible for corrupting the list.
Specifically the line:

    memory = cvmx_bootmem_phy_alloc(mem_alloc_size,
__pa_symbol(&__init_end), -1,
0x100000,
CVMX_BOOTMEM_FLAG_NO_LOCKING);

Which would eventually call:

    cvmx_bootmem_phy_set_size(new_ent_addr,
cvmx_bootmem_phy_get_size
(ent_addr) -
(desired_min_addr -
ent_addr));

Where 'new_ent_addr'=0x4800000 (the address of 'delayed_fput_list')
and the second argument (size)=0xb7fdc00 (the address causing the
kernel panic). The job of this part of 'plat_mem_setup()' is to
allocate chunks of memory for the kernel to use. At the start of
each chunk of memory the size of the chunk is written, hence the
value 0xb7fdc00 is written onto memory at 0x4800000, therefore the
kernel panics when it goes back to access 'delayed_fput_list' later
on in the initialisation process.

On builds that were not crashing it was found that the compiler had
placed 'delayed_fput_list' at 0x4800008, meaning it wasn't corrupted
(but something else in memory was overwritten).

As can be seen in the first function call above the code begins to
allocate chunks of memory beginning from the symbol '__init_end'.
The MIPS linker script (vmlinux.lds.S) however defines the .bss
section to begin after '__init_end'. Therefore memory within the
.bss section is allocated to the kernel to use (System.map shows
'delayed_fput_list' and other kernel structures to be in .bss).

To stop the kernel panic (and the .bss section being corrupted)
memory should begin being allocated from the symbol '_end'.

Signed-off-by: Matt Bennett <matt.bennett@alliedtelesis.co.nz>
Acked-by: David Daney <david.daney@cavium.com>
Cc: linux-mips@linux-mips.org
Cc: aleksey.makarov@auriga.com
Patchwork: https://patchwork.linux-mips.org/patch/11251/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
9 years agoMIPS: Fix R2300 FP context switch handling
Paul Burton [Mon, 21 Sep 2015 17:07:42 +0000 (10:07 -0700)]
MIPS: Fix R2300 FP context switch handling

Commit 1a3d59579b9f ("MIPS: Tidy up FPU context switching") removed FP
context saving from the asm-written resume function in favour of reusing
existing code to perform the same task. However it only removed the FP
context saving code from the r4k_switch.S implementation of resume.
Remove it from the r2300_switch.S implementation too in order to prevent
attempting to save the FP context twice, which would likely lead to an
exception from the second save because the FPU had already been disabled
by the first save.

This patch has only been build tested, using rbtx49xx_defconfig.

Fixes: 1a3d59579b9f ("MIPS: Tidy up FPU context switching")
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Cc: linux-mips@linux-mips.org
Cc: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-kernel@vger.kernel.org
Cc: Manuel Lauss <manuel.lauss@gmail.com>
Patchwork: https://patchwork.linux-mips.org/patch/11167/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
9 years agoMIPS: Fix octeon FP context switch handling
Paul Burton [Mon, 21 Sep 2015 17:07:41 +0000 (10:07 -0700)]
MIPS: Fix octeon FP context switch handling

Commit 1a3d59579b9f ("MIPS: Tidy up FPU context switching") removed FP
context saving from the asm-written resume function in favour of reusing
existing code to perform the same task. However it only removed the FP
context saving code from the r4k_switch.S implementation of resume.
Octeon uses its own implementation in octeon_switch.S, so remove FP
context saving there too in order to prevent attempting to save context
twice. That formerly led to an exception from the second save as follows
because the FPU had already been disabled by the first save:

    do_cpu invoked from kernel context![#1]:
    CPU: 0 PID: 2 Comm: kthreadd Not tainted 4.3.0-rc2-dirty #2
    task: 800000041f84a008 ti: 800000041f864000 task.ti: 800000041f864000
    $ 0   : 0000000000000000 0000000010008ce1 0000000000100000 ffffffffbfffffff
    $ 4   : 800000041f84a008 800000041f84ac08 800000041f84c000 0000000000000004
    $ 8   : 0000000000000001 0000000000000000 0000000000000000 0000000000000001
    $12   : 0000000010008ce3 0000000000119c60 0000000000000036 800000041f864000
    $16   : 800000041f84ac08 800000000792ce80 800000041f84a008 ffffffff81758b00
    $20   : 0000000000000000 ffffffff8175ae50 0000000000000000 ffffffff8176c740
    $24   : 0000000000000006 ffffffff81170300
    $28   : 800000041f864000 800000041f867d90 0000000000000000 ffffffff815f3fa0
    Hi    : 0000000000fa8257
    Lo    : ffffffffe15cfc00
    epc   : ffffffff8112821c resume+0x9c/0x200
    ra    : ffffffff815f3fa0 __schedule+0x3f0/0x7d8
    Status: 10008ce2        KX SX UX KERNEL EXL
    Cause : 1080002c (ExcCode 0b)
    PrId  : 000d0601 (Cavium Octeon+)
    Modules linked in:
    Process kthreadd (pid: 2, threadinfo=800000041f864000, task=800000041f84a008, tls=0000000000000000)
    Stack : ffffffff81604218 ffffffff815f7e08 800000041f84a008 ffffffff811681b0
              800000041f84a008 ffffffff817e9878 0000000000000000 ffffffff81770000
              ffffffff81768340 ffffffff81161398 0000000000000001 0000000000000000
              0000000000000000 ffffffff815f4424 0000000000000000 ffffffff81161d68
              ffffffff81161be8 0000000000000000 0000000000000000 0000000000000000
              0000000000000000 0000000000000000 0000000000000000 ffffffff8111e16c
              0000000000000000 0000000000000000 0000000000000000 0000000000000000
              0000000000000000 0000000000000000 0000000000000000 0000000000000000
              0000000000000000 0000000000000000 0000000000000000 0000000000000000
              0000000000000000 0000000000000000 0000000000000000 0000000000000000
              ...
    Call Trace:
    [<ffffffff8112821c>] resume+0x9c/0x200
    [<ffffffff815f3fa0>] __schedule+0x3f0/0x7d8
    [<ffffffff815f4424>] schedule+0x34/0x98
    [<ffffffff81161d68>] kthreadd+0x180/0x198
    [<ffffffff8111e16c>] ret_from_kernel_thread+0x14/0x1c

Tested using cavium_octeon_defconfig on an EdgeRouter Lite.

Fixes: 1a3d59579b9f ("MIPS: Tidy up FPU context switching")
Reported-by: Aaro Koskinen <aaro.koskinen@nokia.com>
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Cc: linux-mips@linux-mips.org
Cc: Aleksey Makarov <aleksey.makarov@auriga.com>
Cc: linux-kernel@vger.kernel.org
Cc: Chandrakala Chavva <cchavva@caviumnetworks.com>
Cc: David Daney <david.daney@cavium.com>
Cc: Leonid Rosenboim <lrosenboim@caviumnetworks.com>
Patchwork: https://patchwork.linux-mips.org/patch/11166/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
9 years agodrivers: firmware: psci: add system suspend support
Sudeep Holla [Thu, 18 Jun 2015 14:41:34 +0000 (15:41 +0100)]
drivers: firmware: psci: add system suspend support

PSCI v1.0 introduces a new API called PSCI_SYSTEM_SUSPEND. This API
provides the mechanism by which the calling OS can request entry into
the deepest possible system sleep state.

It meets all the necessary preconditions for entering suspend to RAM
state in Linux. This patch adds support for PSCI_SYSTEM_SUSPEND in psci
firmware and registers a psci system suspend operation to implement the
suspend-to-RAM(s2r) in a generic way on all the platforms implementing
PSCI.

Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Acked-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
9 years agodrivers: firmware: psci: define more generic PSCI_FN_NATIVE macro
Sudeep Holla [Thu, 18 Jun 2015 14:41:33 +0000 (15:41 +0100)]
drivers: firmware: psci: define more generic PSCI_FN_NATIVE macro

This patch replaces the definition and usage of PSCI_0_2_FN_NATIVE with
the new and more generic macro PSCI_FN_NATIVE that can be used with any
version. This will be useful for the new features introduced in PSCIv1.0
and for any future revisions.

Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Acked-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
9 years agodrivers: firmware: psci: add PSCI v1.0 DT bindings
Lorenzo Pieralisi [Tue, 26 May 2015 12:06:57 +0000 (13:06 +0100)]
drivers: firmware: psci: add PSCI v1.0 DT bindings

PSCI 1.0 is designed to be fully compliant to the PSCI 0.2
specification, with minor differences that are described in the
PSCI specification.

In particular, PSCI v1.0 augments the specification with a new
power_state format (extended stateid - probeable through the
PSCI_FEATURES call), changes some function return codes and
functions usage requirements wrt PSCI 0.2. These changes mean
that 1.0 vs 0.2 compliancy should be enforced through a DT
compatible string that allows firmware to specify 1.0 only
compliancy so that older kernels are prevented from using
PSCI 1.0 FW implementations in a non-compatible way (eg by
calling a 1.0 FW implementation and expecting 0.2 behaviour).

This patch adds PSCI 1.0 DT bindings and related compatible
string.

Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Acked-by: Sudeep Holla <sudeep.holla@arm.com>
Tested-by: Jisheng Zhang <jszhang@marvell.com>
Cc: Mark Rutland <mark.rutland@arm.com>
9 years agodrivers: firmware: psci: add extended stateid power_state support
Lorenzo Pieralisi [Tue, 26 May 2015 16:10:32 +0000 (17:10 +0100)]
drivers: firmware: psci: add extended stateid power_state support

PSCI v1.0 augmented the power_state parameter format specification
(extended stateid) and introduced a way to probe it through the
PSCI_FEATURES interface.

This patch implements code that detects the power_state format at
run-time through the PSCI_FEATURES interface, so that the power_state
argument can be properly detected and validated in the kernel according
to the information provided through firmware.

Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Tested-by: Jisheng Zhang <jszhang@marvell.com>
Cc: Mark Rutland <mark.rutland@arm.com>
9 years agodrivers: firmware: psci: add PSCI_FEATURES call
Lorenzo Pieralisi [Tue, 26 May 2015 16:06:21 +0000 (17:06 +0100)]
drivers: firmware: psci: add PSCI_FEATURES call

PSCI v1.0 introduces a PSCI_FEATURES call that allows to probe for
features related to a specific function identifier.

This patch adds PSCI_FEATURES support to the PSCI firmware layer.

Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Tested-by: Jisheng Zhang <jszhang@marvell.com>
Cc: Mark Rutland <mark.rutland@arm.com>
9 years agodrivers: firmware: psci: move power_state handling to generic code
Lorenzo Pieralisi [Tue, 26 May 2015 15:49:01 +0000 (16:49 +0100)]
drivers: firmware: psci: move power_state handling to generic code

Functions implemented on arm64 to check if a power_state parameter
is valid and if the power_state implies context loss are not
arm64 specific and should be moved to generic code so that they
can be reused on arm systems too.

This patch moves the functions handling the power_state parameter
to generic PSCI firmware layer code.

Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Acked-by: Sudeep Holla <sudeep.holla@arm.com>
Tested-by: Jisheng Zhang <jszhang@marvell.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
9 years agodrivers: firmware: psci: add INVALID_ADDRESS return value
Lorenzo Pieralisi [Fri, 22 May 2015 13:31:37 +0000 (14:31 +0100)]
drivers: firmware: psci: add INVALID_ADDRESS return value

PSCI 1.0 introduces the INVALID_ADDRESS return value for functions
that take an address as input parameter (eg CPU_SUSPEND).

This patch adds INVALID_ADDRESS return value to kernel code and
updates the PSCI to linux error conversion to take it into account.

The kernel error value associated to INVALID_ADDRESS is set to
the error returned when the PSCI error code is INVALID_PARAMETERS
to comply with current call sites expected return value, given
that the kernel at present has no use for the additional error
information reported.

Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Acked-by: Sudeep Holla <sudeep.holla@arm.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Jisheng Zhang <jszhang@marvell.com>
9 years agoMerge tag 'mmc-v4.3-rc3' of git://git.linaro.org/people/ulf.hansson/mmc
Linus Torvalds [Fri, 2 Oct 2015 12:03:04 +0000 (08:03 -0400)]
Merge tag 'mmc-v4.3-rc3' of git://git.linaro.org/people/ulf.hansson/mmc

Pull MMC fixes from Ulf Hansson:
 "Here are some mmc fixes intended for v4.3 rc4:

  MMC core:
   - Allow users of mmc_of_parse() to succeed when CONFIG_GPIOLIB is
     unset
   - Prevent infinite loop of re-tuning for CRC-errors for CMD19 and
     CMD21

   MMC host:
   - pxamci: Fix issues with card detect
   - sunxi: Fix clk-delay settings"

* tag 'mmc-v4.3-rc3' of git://git.linaro.org/people/ulf.hansson/mmc:
  mmc: core: fix dead loop of mmc_retune
  mmc: pxamci: fix card detect with slot-gpio API
  mmc: sunxi: Fix clk-delay settings
  mmc: core: Don't return an error for CD/WP GPIOs when GPIOLIB is unset

9 years agoMerge git://git.infradead.org/intel-iommu
Linus Torvalds [Fri, 2 Oct 2015 11:59:29 +0000 (07:59 -0400)]
Merge git://git.infradead.org/intel-iommu

Pull IOVA fixes from David Woodhouse:
 "The main fix here is the first one, fixing the over-allocation of
   size-aligned requests.  The other patches simply make the existing
  IOVA code available to users other than the Intel VT-d driver, with no
  functional change.

  I concede the latter really *should* have been submitted during the
  merge window, but since it's basically risk-free and people are
  waiting to build on top of it and it's my fault I didn't get it in, I
  (and they) would be grateful if you'd take it"

* git://git.infradead.org/intel-iommu:
  iommu: Make the iova library a module
  iommu: iova: Export symbols
  iommu: iova: Move iova cache management to the iova library
  iommu/iova: Avoid over-allocating when size-aligned

9 years agoarm64: ftrace: fix function_graph tracer panic
Li Bin [Wed, 30 Sep 2015 02:49:55 +0000 (10:49 +0800)]
arm64: ftrace: fix function_graph tracer panic

When function graph tracer is enabled, the following operation
will trigger panic:

mount -t debugfs nodev /sys/kernel
echo next_tgid > /sys/kernel/tracing/set_ftrace_filter
echo function_graph > /sys/kernel/tracing/current_tracer
ls /proc/

------------[ cut here ]------------
[  198.501417] Unable to handle kernel paging request at virtual address cb88537fdc8ba316
[  198.506126] pgd = ffffffc008f79000
[  198.509363] [cb88537fdc8ba316] *pgd=00000000488c6003, *pud=00000000488c6003, *pmd=0000000000000000
[  198.517726] Internal error: Oops: 94000005 [#1] SMP
[  198.518798] Modules linked in:
[  198.520582] CPU: 1 PID: 1388 Comm: ls Tainted: G
[  198.521800] Hardware name: linux,dummy-virt (DT)
[  198.522852] task: ffffffc0fa9e8000 ti: ffffffc0f9ab0000 task.ti: ffffffc0f9ab0000
[  198.524306] PC is at next_tgid+0x30/0x100
[  198.525205] LR is at return_to_handler+0x0/0x20
[  198.526090] pc : [<ffffffc0002a1070>] lr : [<ffffffc0000907c0>] pstate: 60000145
[  198.527392] sp : ffffffc0f9ab3d40
[  198.528084] x29: ffffffc0f9ab3d40 x28: ffffffc0f9ab0000
[  198.529406] x27: ffffffc000d6a000 x26: ffffffc000b786e8
[  198.530659] x25: ffffffc0002a1900 x24: ffffffc0faf16c00
[  198.531942] x23: ffffffc0f9ab3ea0 x22: 0000000000000002
[  198.533202] x21: ffffffc000d85050 x20: 0000000000000002
[  198.534446] x19: 0000000000000002 x18: 0000000000000000
[  198.535719] x17: 000000000049fa08 x16: ffffffc000242efc
[  198.537030] x15: 0000007fa472b54c x14: ffffffffff000000
[  198.538347] x13: ffffffc0fada84a0 x12: 0000000000000001
[  198.539634] x11: ffffffc0f9ab3d70 x10: ffffffc0f9ab3d70
[  198.540915] x9 : ffffffc0000907c0 x8 : ffffffc0f9ab3d40
[  198.542215] x7 : 0000002e330f08f0 x6 : 0000000000000015
[  198.543508] x5 : 0000000000000f08 x4 : ffffffc0f9835ec0
[  198.544792] x3 : cb88537fdc8ba316 x2 : cb88537fdc8ba306
[  198.546108] x1 : 0000000000000002 x0 : ffffffc000d85050
[  198.547432]
[  198.547920] Process ls (pid: 1388, stack limit = 0xffffffc0f9ab0020)
[  198.549170] Stack: (0xffffffc0f9ab3d40 to 0xffffffc0f9ab4000)
[  198.582568] Call trace:
[  198.583313] [<ffffffc0002a1070>] next_tgid+0x30/0x100
[  198.584359] [<ffffffc0000907bc>] ftrace_graph_caller+0x6c/0x70
[  198.585503] [<ffffffc0000907bc>] ftrace_graph_caller+0x6c/0x70
[  198.586574] [<ffffffc0000907bc>] ftrace_graph_caller+0x6c/0x70
[  198.587660] [<ffffffc0000907bc>] ftrace_graph_caller+0x6c/0x70
[  198.588896] Code: aa0003f5 2a0103f4 b4000102 91004043 (885f7c60)
[  198.591092] ---[ end trace 6a346f8f20949ac8 ]---

This is because when using function graph tracer, if the traced
function return value is in multi regs ([x0-x7]), return_to_handler
may corrupt them. So in return_to_handler, the parameter regs should
be protected properly.

Cc: <stable@vger.kernel.org> # 3.18+
Signed-off-by: Li Bin <huawei.libin@huawei.com>
Acked-by: AKASHI Takahiro <takahiro.akashi@linaro.org>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
9 years agoMIPS: BPF: Fix load delay slots.
Ralf Baechle [Fri, 2 Oct 2015 07:48:57 +0000 (09:48 +0200)]
MIPS: BPF: Fix load delay slots.

The entire bpf_jit_asm.S is written in noreorder mode because "we know
better" according to a comment.  This also prevented the assembler from
throwing in the required NOPs for MIPS I processors which have no
load-use interlock, thus the load's consumer might end up using the
old value of the register from prior to the load.

Fixed by putting the assembler in reorder mode for just the affected
load instructions.  This is not enough for gas to actually try to be
clever by looking at the next instruction and inserting a nop only
when needed but as the comment said "we know better", so getting gas
to unconditionally emit a NOP is just right in this case and prevents
adding further ifdefery.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
9 years agox86/headers/uapi: Fix __BITS_PER_LONG value for x32 builds
Ben Hutchings [Thu, 1 Oct 2015 00:40:43 +0000 (01:40 +0100)]
x86/headers/uapi: Fix __BITS_PER_LONG value for x32 builds

On x32, gcc predefines __x86_64__ but long is only 32-bit.  Use
__ILP32__ to distinguish x32.

Fixes this compiler error in perf:

tools/include/asm-generic/bitops/__ffs.h: In function '__ffs':
tools/include/asm-generic/bitops/__ffs.h:19:8: error: right shift count >= width of type [-Werror=shift-count-overflow]
  word >>= 32;
       ^

This isn't sufficient to build perf for x32, though.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Link: http://lkml.kernel.org/r/1443660043.2730.15.camel@decadent.org.uk
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agomd/bitmap: don't pass -1 to bitmap_storage_alloc.
NeilBrown [Thu, 1 Oct 2015 06:03:38 +0000 (16:03 +1000)]
md/bitmap: don't pass -1 to bitmap_storage_alloc.

Passing -1 to bitmap_storage_alloc() causes page->index to be set to
-1, which is quite problematic.

So only pass ->cluster_slot if mddev_is_clustered().

Fixes: b97e92574c0b ("Use separate bitmaps for each nodes in the cluster")
Cc: stable@vger.kernel.org (v4.1+)
Signed-off-by: NeilBrown <neilb@suse.com>
9 years agomd/raid1: Avoid raid1 resync getting stuck
Jes Sorensen [Wed, 16 Sep 2015 14:20:05 +0000 (10:20 -0400)]
md/raid1: Avoid raid1 resync getting stuck

close_sync() needs to set conf->next_resync to a large, but safe value
below MaxSector and use it to determine whether or not to set
start_next_window in wait_barrier()

Solution suggested by Neil Brown.

Reported-by: Nate Dailey <nate.dailey@stratus.com>
Tested-by: Xiao Ni <xni@redhat.com>
Signed-off-by: Jes Sorensen <Jes.Sorensen@redhat.com>
Signed-off-by: NeilBrown <neilb@suse.com>
9 years agomd: drop null test before destroy functions
Julia Lawall [Sun, 13 Sep 2015 12:15:10 +0000 (14:15 +0200)]
md: drop null test before destroy functions

Remove unneeded NULL test.

The semantic patch that makes this change is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@@ expression x; @@
-if (x != NULL)
  \(kmem_cache_destroy\|mempool_destroy\|dma_pool_destroy\)(x);
// </smpl>

Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: NeilBrown <neilb@suse.com>
9 years agomd: clear CHANGE_PENDING in readonly array
Shaohua Li [Fri, 18 Sep 2015 17:20:12 +0000 (10:20 -0700)]
md: clear CHANGE_PENDING in readonly array

If faulty disks of an array are more than allowed degraded number, the
array enters error handling. It will be marked as read-only with
MD_CHANGE_PENDING/RECOVERY_NEEDED set. But currently recovery doesn't
clear CHANGE_PENDING bit for read-only array.  If MD_CHANGE_PENDING is
set for a raid5 array, all returned IO will be hold on a list till the
bit is clear. But recovery nevery clears this bit, the IO is always in
pending state and nevery finish. This has bad effects like upper layer
can't get an IO error and the array can't be stopped.

Fixes: c3cce6cda162 ("md/raid5: ensure device failure recorded before write request returns.")
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
9 years agomd/raid0: apply base queue limits *before* disk_stack_limits
NeilBrown [Thu, 24 Sep 2015 05:47:47 +0000 (15:47 +1000)]
md/raid0: apply base queue limits *before* disk_stack_limits

Calling e.g. blk_queue_max_hw_sectors() after calls to
disk_stack_limits() discards the settings determined by
disk_stack_limits().
So we need to make those calls first.

Fixes: 199dc6ed5179 ("md/raid0: update queue parameter in a safer location.")
Cc: stable@vger.kernel.org (v2.6.35+ - please apply with 199dc6ed5179).
Reported-by: Jes Sorensen <Jes.Sorensen@redhat.com>
Signed-off-by: NeilBrown <neilb@suse.com>
9 years agomd/raid5: don't index beyond end of array in need_this_block().
NeilBrown [Thu, 24 Sep 2015 05:25:36 +0000 (15:25 +1000)]
md/raid5: don't index beyond end of array in need_this_block().

When need_this_block probably shouldn't be called when there
are more than 2 failed devices, we really don't want it to try
indexing beyond the end of the failed_num[] of fdev[] arrays.

So limit the loops to at most 2 iterations.

Reported-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.de>
9 years agoraid5: update analysis state for failed stripe
Shaohua Li [Fri, 18 Sep 2015 17:20:13 +0000 (10:20 -0700)]
raid5: update analysis state for failed stripe

handle_failed_stripe() makes the stripe fail, eg, all IO will return
with a failure, but it doesn't update stripe_head_state. Later
handle_stripe() has special handling for raid6 for handle_stripe_fill().
That check before handle_stripe_fill() doesn't skip the failed stripe
and we get a kernel crash in need_this_block.  This patch clear the
analysis state to make sure no functions wrongly called after
handle_failed_stripe()

Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
9 years agomd: wait for pending superblock updates before switching to read-only
NeilBrown [Thu, 24 Sep 2015 04:00:51 +0000 (14:00 +1000)]
md: wait for pending superblock updates before switching to read-only

If a superblock update is pending, wait for it to complete before
letting md_set_readonly() switch to readonly.
Otherwise we might lose important information about a device having
failed.

For external arrays, waiting for superblock updates can wait on
user-space, so in that case, just return an error.

Reported-and-tested-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
9 years agox86/mm: Set NX on gap between __ex_table and rodata
Stephen Smalley [Thu, 1 Oct 2015 13:04:22 +0000 (09:04 -0400)]
x86/mm: Set NX on gap between __ex_table and rodata

Unused space between the end of __ex_table and the start of
rodata can be left W+x in the kernel page tables.  Extend the
setting of the NX bit to cover this gap by starting from
text_end rather than rodata_start.

  Before:
  ---[ High Kernel Mapping ]---
  0xffffffff80000000-0xffffffff81000000          16M                               pmd
  0xffffffff81000000-0xffffffff81600000           6M     ro         PSE     GLB x  pmd
  0xffffffff81600000-0xffffffff81754000        1360K     ro                 GLB x  pte
  0xffffffff81754000-0xffffffff81800000         688K     RW                 GLB x  pte
  0xffffffff81800000-0xffffffff81a00000           2M     ro         PSE     GLB NX pmd
  0xffffffff81a00000-0xffffffff81b3b000        1260K     ro                 GLB NX pte
  0xffffffff81b3b000-0xffffffff82000000        4884K     RW                 GLB NX pte
  0xffffffff82000000-0xffffffff82200000           2M     RW         PSE     GLB NX pmd
  0xffffffff82200000-0xffffffffa0000000         478M                               pmd

  After:
  ---[ High Kernel Mapping ]---
  0xffffffff80000000-0xffffffff81000000          16M                               pmd
  0xffffffff81000000-0xffffffff81600000           6M     ro         PSE     GLB x  pmd
  0xffffffff81600000-0xffffffff81754000        1360K     ro                 GLB x  pte
  0xffffffff81754000-0xffffffff81800000         688K     RW                 GLB NX pte
  0xffffffff81800000-0xffffffff81a00000           2M     ro         PSE     GLB NX pmd
  0xffffffff81a00000-0xffffffff81b3b000        1260K     ro                 GLB NX pte
  0xffffffff81b3b000-0xffffffff82000000        4884K     RW                 GLB NX pte
  0xffffffff82000000-0xffffffff82200000           2M     RW         PSE     GLB NX pmd
  0xffffffff82200000-0xffffffffa0000000         478M                               pmd

Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Acked-by: Kees Cook <keescook@chromium.org>
Cc: <stable@vger.kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Link: http://lkml.kernel.org/r/1443704662-3138-1-git-send-email-sds@tycho.nsa.gov
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agox86/kexec: Fix kexec crash in syscall kexec_file_load()
Lee, Chun-Yi [Tue, 29 Sep 2015 12:58:57 +0000 (20:58 +0800)]
x86/kexec: Fix kexec crash in syscall kexec_file_load()

The original bug is a page fault crash that sometimes happens
on big machines when preparing ELF headers:

    BUG: unable to handle kernel paging request at ffffc90613fc9000
    IP: [<ffffffff8103d645>] prepare_elf64_ram_headers_callback+0x165/0x260

The bug is caused by us under-counting the number of memory ranges
and subsequently not allocating enough ELF header space for them.
The bug is typically masked on smaller systems, because the ELF header
allocation is rounded up to the next page.

This patch modifies the code in fill_up_crash_elf_data() by using
walk_system_ram_res() instead of walk_system_ram_range() to correctly
count the max number of crash memory ranges. That's because the
walk_system_ram_range() filters out small memory regions that
reside in the same page, but walk_system_ram_res() does not.

Here's how I found the bug:

After tracing prepare_elf64_headers() and prepare_elf64_ram_headers_callback(),
the code uses walk_system_ram_res() to fill-in crash memory regions information
to the program header, so it counts those small memory regions that
reside in a page area.

But, when the kernel was using walk_system_ram_range() in
fill_up_crash_elf_data() to count the number of crash memory regions,
it filters out small regions.

I printed those small memory regions, for example:

  kexec: Get nr_ram ranges. vaddr=0xffff880077592258 paddr=0x77592258, sz=0xdc0

Based on the code in walk_system_ram_range(), this memory region
will be filtered out:

  pfn = (0x77592258 + 0x1000 - 1) >> 12 = 0x77593
  end_pfn = (0x77592258 + 0xfc0 -1 + 1) >> 12 = 0x77593
  end_pfn - pfn = 0x77593 - 0x77593 = 0  <=== if (end_pfn > pfn) is FALSE

So, the max_nr_ranges that's counted by the kernel doesn't include
small memory regions - causing us to under-allocate the required space.
That causes the page fault crash that happens in a later code path
when preparing ELF headers.

This bug is not easy to reproduce on small machines that have few
CPUs, because the allocated page aligned ELF buffer has more free
space to cover those small memory regions' PT_LOAD headers.

Signed-off-by: Lee, Chun-Yi <jlee@suse.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Baoquan He <bhe@redhat.com>
Cc: Jiang Liu <jiang.liu@linux.intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Takashi Iwai <tiwai@suse.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Cc: Vivek Goyal <vgoyal@redhat.com>
Cc: kexec@lists.infradead.org
Cc: linux-kernel@vger.kernel.org
Cc: <stable@vger.kernel.org>
Link: http://lkml.kernel.org/r/1443531537-29436-1-git-send-email-jlee@suse.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agodrm/dp/mst: add some defines for logical/physical ports
Dave Airlie [Thu, 1 Oct 2015 06:28:25 +0000 (16:28 +1000)]
drm/dp/mst: add some defines for logical/physical ports

This just removes the magic number.

Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agodrm/dp/mst: drop cancel work sync in the mstb destroy path (v2)
Dave Airlie [Wed, 30 Sep 2015 00:39:42 +0000 (10:39 +1000)]
drm/dp/mst: drop cancel work sync in the mstb destroy path (v2)

Since 9eb1e57f564d4e6e10991402726cc83fe0b9172f
drm/dp/mst: make sure mst_primary mstb is valid in work function

we validate the mstb structs in the work function, and doing
that takes a reference. So we should never get here with the
work function running using the mstb device, only if the work
function hasn't run yet or is running for another mstb.

So we don't need to sync the work here, this was causing
lockdep spew as below.

[  +0.000160] =============================================
[  +0.000001] [ INFO: possible recursive locking detected ]
[  +0.000002] 3.10.0-320.el7.rhel72.stable.backport.3.x86_64.debug #1 Tainted: G        W      ------------
[  +0.000001] ---------------------------------------------
[  +0.000001] kworker/4:2/1262 is trying to acquire lock:
[  +0.000001]  ((&mgr->work)){+.+.+.}, at: [<ffffffff810b29a5>] flush_work+0x5/0x2e0
[  +0.000007]
but task is already holding lock:
[  +0.000001]  ((&mgr->work)){+.+.+.}, at: [<ffffffff810b57e4>] process_one_work+0x1b4/0x710
[  +0.000004]
other info that might help us debug this:
[  +0.000001]  Possible unsafe locking scenario:

[  +0.000002]        CPU0
[  +0.000000]        ----
[  +0.000001]   lock((&mgr->work));
[  +0.000002]   lock((&mgr->work));
[  +0.000001]
 *** DEADLOCK ***

[  +0.000001]  May be due to missing lock nesting notation

[  +0.000002] 2 locks held by kworker/4:2/1262:
[  +0.000001]  #0:  (events_long){.+.+.+}, at: [<ffffffff810b57e4>] process_one_work+0x1b4/0x710
[  +0.000004]  #1:  ((&mgr->work)){+.+.+.}, at: [<ffffffff810b57e4>] process_one_work+0x1b4/0x710
[  +0.000003]
stack backtrace:
[  +0.000003] CPU: 4 PID: 1262 Comm: kworker/4:2 Tainted: G        W      ------------   3.10.0-320.el7.rhel72.stable.backport.3.x86_64.debug #1
[  +0.000001] Hardware name: LENOVO 20EGS0R600/20EGS0R600, BIOS GNET71WW (2.19 ) 02/05/2015
[  +0.000008] Workqueue: events_long drm_dp_mst_link_probe_work [drm_kms_helper]
[  +0.000001]  ffffffff82c26c90 00000000a527b914 ffff88046399bae8 ffffffff816fe04d
[  +0.000004]  ffff88046399bb58 ffffffff8110f47f ffff880461438000 0001009b840fc003
[  +0.000002]  ffff880461438a98 0000000000000000 0000000804dc26e1 ffffffff824a2c00
[  +0.000003] Call Trace:
[  +0.000004]  [<ffffffff816fe04d>] dump_stack+0x19/0x1b
[  +0.000004]  [<ffffffff8110f47f>] __lock_acquire+0x115f/0x1250
[  +0.000002]  [<ffffffff8110fd49>] lock_acquire+0x99/0x1e0
[  +0.000002]  [<ffffffff810b29a5>] ? flush_work+0x5/0x2e0
[  +0.000002]  [<ffffffff810b29ee>] flush_work+0x4e/0x2e0
[  +0.000002]  [<ffffffff810b29a5>] ? flush_work+0x5/0x2e0
[  +0.000004]  [<ffffffff81025905>] ? native_sched_clock+0x35/0x80
[  +0.000002]  [<ffffffff81025959>] ? sched_clock+0x9/0x10
[  +0.000002]  [<ffffffff810da1f5>] ? local_clock+0x25/0x30
[  +0.000002]  [<ffffffff8110dca9>] ? mark_held_locks+0xb9/0x140
[  +0.000003]  [<ffffffff810b4ed5>] ? __cancel_work_timer+0x95/0x160
[  +0.000002]  [<ffffffff810b4ee8>] __cancel_work_timer+0xa8/0x160
[  +0.000002]  [<ffffffff810b4fb0>] cancel_work_sync+0x10/0x20
[  +0.000007]  [<ffffffffa0160d17>] drm_dp_destroy_mst_branch_device+0x27/0x120 [drm_kms_helper]
[  +0.000006]  [<ffffffffa0163968>] drm_dp_mst_link_probe_work+0x78/0xa0 [drm_kms_helper]
[  +0.000002]  [<ffffffff810b5850>] process_one_work+0x220/0x710
[  +0.000002]  [<ffffffff810b57e4>] ? process_one_work+0x1b4/0x710
[  +0.000005]  [<ffffffff810b5e5b>] worker_thread+0x11b/0x3a0
[  +0.000003]  [<ffffffff810b5d40>] ? process_one_work+0x710/0x710
[  +0.000002]  [<ffffffff810beced>] kthread+0xed/0x100
[  +0.000003]  [<ffffffff810bec00>] ? insert_kthread_work+0x80/0x80
[  +0.000003]  [<ffffffff817121d8>] ret_from_fork+0x58/0x90

v2: add flush_work.

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: stable@vger.kernel.org
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agodrm/dp/mst: split connector registration into two parts (v2)
Dave Airlie [Wed, 16 Sep 2015 07:55:23 +0000 (17:55 +1000)]
drm/dp/mst: split connector registration into two parts (v2)

In order to cache the EDID properly for tiled displays, we
need to retrieve it before we register the connector with
userspace, otherwise userspace can call get resources
and try and get the edid before we've even cached it.

This fixes some problems when hotplugging mst monitors,
with X/mutter running. As mutter seems to get 0 modes
for one of the monitors in the tile.

v2: fix warning in radeon
handle tile setting in cached path rather than
get edid path.

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: stable@vger.kernel.org
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agodrm/dp/mst: update the link_address_sent before sending the link address (v3)
Dave Airlie [Sun, 6 Sep 2015 08:53:00 +0000 (18:53 +1000)]
drm/dp/mst: update the link_address_sent before sending the link address (v3)

Update the state before sending the msg to close it.

v2: reset value if return indicates we haven't send the msg.
v3: just clean the code up.
Pointed out by Adam J Richter on

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91481

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agodrm/dp/mst: fixup handling hotplug on port removal.
Dave Airlie [Wed, 16 Sep 2015 00:37:28 +0000 (10:37 +1000)]
drm/dp/mst: fixup handling hotplug on port removal.

output ports should always have a connector, unless
in the rare case connector allocation fails in the
driver.

In this case we only need to teardown the pdt,
and free the struct, and there is no need to
send a hotplug msg.

In the case were we add the port to the destroy
list we need to send a hotplug if we destroy
any connectors, so userspace knows to reprobe
stuff.

this patch also handles port->connector allocation
failing which should be a rare event, but makes
the code consistent.

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: stable@vger.kernel.org
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agodrm/dp/mst: don't pass port into the path builder function
Dave Airlie [Wed, 16 Sep 2015 01:04:49 +0000 (11:04 +1000)]
drm/dp/mst: don't pass port into the path builder function

This is unnecessary and it makes it easier to see what is needed
from port.

also add blank line to make things nicer.

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>