GitHub/moto-9609/android_kernel_motorola_exynos9610.git
9 years agodmaengine: edma: Add dummy driver skeleton for edma3-tptc
Peter Ujfalusi [Mon, 2 Nov 2015 13:21:40 +0000 (15:21 +0200)]
dmaengine: edma: Add dummy driver skeleton for edma3-tptc

The eDMA3 TPTC does not need any software configuration, but it is a
separate IP block in the SoC. In order the omap hwmod core to be able to
handle the TPTC resources correctly in regards of PM we need to have a
driver loaded for it.
This patch will add a dummy driver skeleton without probe or remove
callbacks provided.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Reported-by: Olof Johansson <olof@lixom.net>
Tested-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agoRevert "ARM: DTS: am33xx: Use the new DT bindings for the eDMA3"
Vinod Koul [Mon, 2 Nov 2015 15:36:46 +0000 (21:06 +0530)]
Revert "ARM: DTS: am33xx: Use the new DT bindings for the eDMA3"

This reverts commit d871cd2ec5abf8715774bcb90aa32ae5b750b587 as it causes
regression in BBB

Reported-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agoRevert "ARM: DTS: am437x: Use the new DT bindings for the eDMA3"
Vinod Koul [Mon, 2 Nov 2015 15:35:29 +0000 (21:05 +0530)]
Revert "ARM: DTS: am437x: Use the new DT bindings for the eDMA3"

This reverts commit e3faf2b8826b8ac58cdaad7f801e59e389320f0e as it causes
regression in BBB

Reported-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agoMerge branch 'topic/of' into for-linus
Vinod Koul [Sat, 31 Oct 2015 02:07:21 +0000 (07:37 +0530)]
Merge branch 'topic/of' into for-linus

9 years agoMerge branch 'topic/ioatdma' into for-linus
Vinod Koul [Sat, 31 Oct 2015 02:07:13 +0000 (07:37 +0530)]
Merge branch 'topic/ioatdma' into for-linus

9 years agoMerge branch 'topic/idma' into for-linus
Vinod Koul [Sat, 31 Oct 2015 02:07:05 +0000 (07:37 +0530)]
Merge branch 'topic/idma' into for-linus

9 years agoMerge branch 'topic/edma' into for-linus
Vinod Koul [Sat, 31 Oct 2015 02:06:55 +0000 (07:36 +0530)]
Merge branch 'topic/edma' into for-linus

Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Conflicts:
drivers/dma/edma.c

9 years agoMerge branch 'topic/dw' into for-linus
Vinod Koul [Sat, 31 Oct 2015 02:05:07 +0000 (07:35 +0530)]
Merge branch 'topic/dw' into for-linus

9 years agodmaengine: dw: some Intel devices has no memcpy support
Andy Shevchenko [Tue, 13 Oct 2015 17:09:19 +0000 (20:09 +0300)]
dmaengine: dw: some Intel devices has no memcpy support

Provide a flag to choose if the device does support memory-to-memory transfers.
At least this is not true for iDMA32 controller that might be supported in the
future. Besides that Intel BayTrail and Braswell users should not try this
feature due to HW specific behaviour.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: dw: platform: provide platform data for Intel
Andy Shevchenko [Tue, 13 Oct 2015 17:09:18 +0000 (20:09 +0300)]
dmaengine: dw: platform: provide platform data for Intel

Provide platform data explicitly for Intel SoCs where dw_dmac is enumerated by
ACPI.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: dw: don't override platform data with autocfg
Andy Shevchenko [Tue, 13 Oct 2015 17:09:17 +0000 (20:09 +0300)]
dmaengine: dw: don't override platform data with autocfg

Let probe driver decide either it wants to auto configure the driver or have
explicitly defined properties.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: hdmac: Add scatter-gathered memset support
Maxime Ripard [Thu, 22 Oct 2015 09:41:00 +0000 (11:41 +0200)]
dmaengine: hdmac: Add scatter-gathered memset support

Just like memset support, the HDMAC might be used to do a memset over a
discontiguous memory area.

In such a case, we'll just build up a chain of memset descriptors over the
contiguous chunks of memory to set, in order to allow such a support.

Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: hdmac: factorise memset descriptor allocation
Maxime Ripard [Thu, 22 Oct 2015 09:40:59 +0000 (11:40 +0200)]
dmaengine: hdmac: factorise memset descriptor allocation

The memset and scatter gathered memset are going to use some common logic
to create their descriptors.

Move that logic into a function of its own so that we can share it with the
future memset_sg callback.

Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agoMerge branch 'topic/dw' into for-linus
Vinod Koul [Thu, 29 Oct 2015 01:31:11 +0000 (10:31 +0900)]
Merge branch 'topic/dw' into for-linus

9 years agodmaengine: virt-dma: Fix kernel-doc annotations
Lars-Peter Clausen [Tue, 20 Oct 2015 11:14:45 +0000 (13:14 +0200)]
dmaengine: virt-dma: Fix kernel-doc annotations

In kernel-doc annotations parameters need to start with a @ for them to be
properly recognized. Add those where missing for virt-dma.

Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agoARM: DTS: am437x: Use the new DT bindings for the eDMA3
Peter Ujfalusi [Fri, 16 Oct 2015 07:18:12 +0000 (10:18 +0300)]
ARM: DTS: am437x: Use the new DT bindings for the eDMA3

Switch to use the ti,edma3-tpcc and ti,edma3-tptc binding for the eDMA3 and
enable the DMA even crossbar with ti,am335x-edma-crossbar.
With the new bindings boards can customize and tweak the DMA channel
priority to match their needs. With the new binding the memcpy is safe
to be used since with the old binding it was not possible for a driver
to know which channel is allowed to be used as non HW triggered channel.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agoARM: DTS: am33xx: Use the new DT bindings for the eDMA3
Peter Ujfalusi [Fri, 16 Oct 2015 07:18:11 +0000 (10:18 +0300)]
ARM: DTS: am33xx: Use the new DT bindings for the eDMA3

Switch to use the ti,edma3-tpcc and ti,edma3-tptc binding for the eDMA3 and
enable the DMA even crossbar with ti,am335x-edma-crossbar.
With the new bindings boards can customize and tweak the DMA channel
priority to match their needs. With the new binding the memcpy is safe
to be used since with the old binding it was not possible for a driver
to know which channel is allowed to be used as non HW triggered channel.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: New device tree binding
Peter Ujfalusi [Fri, 16 Oct 2015 07:18:10 +0000 (10:18 +0300)]
dmaengine: edma: New device tree binding

With the old binding and driver architecture we had many issues:
No way to assign eDMA channels to event queues, thus not able to tune the
system by moving specific DMA channels to low/high priority servicing. We
moved the cyclic channels to high priority within the code, but that was
just a workaround to this issue.
Memcopy was fundamentally broken: even if the driver scanned the DT/devices
in the booted system for direct DMA users (which is not effective when the
events are going through a crossbar) and created a map of 'used' channels,
this information was not really usable. Since via dmaengien API the eDMA
driver will be called with _some_ channel number, we would try to request
this channel when any channel is requested for memcpy. By luck we got
channel which is not used by any device most of the time so things worked,
but if a device would have been using the given channel, but not requested
it, the memcpy channel would have been waiting for HW event.
The old code had the am33xx/am43xx DMA event router handling embedded. This
should have been done in a separate driver since it is not part of the
actual eDMA IP.
There were no way to 'lock' PaRAM slots to be used by the DSP for example
when booting with DT.
In DT boot the edma node used more than one hwmod which is not a good
practice and the kernel prints warning because of this.

With the new bindings and the changes in the driver we can:
- No regression with Legacy binding and non DT boot
- DMA channels can be assigned to any TC (to set priority)
- PaRAM slots can be reserved for other cores to use
- Dynamic power management for CC and TCs, if only TC0 is used all other TC
  can be powered down for example

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: Kconfig: edma: Select TI_DMA_CROSSBAR in case of ARCH_OMAP
Peter Ujfalusi [Fri, 16 Oct 2015 07:18:09 +0000 (10:18 +0300)]
dmaengine: Kconfig: edma: Select TI_DMA_CROSSBAR in case of ARCH_OMAP

Since the crossbar is needed for eDMA when it is used on OMAP like
platforms (am335x/am437x and later DRA7xx), select the crossbar to be built
if ARCH_OMAP is set.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: ti-dma-crossbar: Add support for crossbar on AM33xx/AM43xx
Peter Ujfalusi [Fri, 16 Oct 2015 07:18:08 +0000 (10:18 +0300)]
dmaengine: ti-dma-crossbar: Add support for crossbar on AM33xx/AM43xx

The DMA event crossbar on AM33xx/AM43xx is different from the one found in
DRA7x family.
Instead of a single event crossbar it has 64 identical mux attached to each
eDMA event line. When the 0 event mux is selected, the default mapped event
is going to be routed to the corresponding eDMA event line. If different
mux is selected, then the selected event is going to be routed to the given
eDMA event.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Merge the of parsing functions
Peter Ujfalusi [Fri, 16 Oct 2015 07:18:07 +0000 (10:18 +0300)]
dmaengine: edma: Merge the of parsing functions

Instead of nesting functions just merge them since the resulting function
is still small and readable.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Do not allocate memory for edma_rsv_info in case of DT boot
Peter Ujfalusi [Fri, 16 Oct 2015 07:18:06 +0000 (10:18 +0300)]
dmaengine: edma: Do not allocate memory for edma_rsv_info in case of DT boot

The channel/slot reservation is not supported when booted with DT so there
is not need to allocate memory.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Refactor the dma device and channel struct initialization
Peter Ujfalusi [Fri, 16 Oct 2015 07:18:05 +0000 (10:18 +0300)]
dmaengine: edma: Refactor the dma device and channel struct initialization

Move all code under one function to do the dma device and eDMA channel
related setup so they are not scattered around the driver.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Get qDMA channel information from HW also
Peter Ujfalusi [Fri, 16 Oct 2015 07:18:04 +0000 (10:18 +0300)]
dmaengine: edma: Get qDMA channel information from HW also

Query the number of qDMA channels from CCCFG register.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Merge map_dmach_to_queue into assign_channel_eventq
Peter Ujfalusi [Fri, 16 Oct 2015 07:18:03 +0000 (10:18 +0300)]
dmaengine: edma: Merge map_dmach_to_queue into assign_channel_eventq

edma_assign_channel_eventq() is a wrapper around edma_map_dmach_to_queue()
We can merge the content of the later so we will have only one function
to be used for mapping channels to given eventq

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Correct PaRAM access function names (_parm_ to _param_)
Peter Ujfalusi [Fri, 16 Oct 2015 07:18:02 +0000 (10:18 +0300)]
dmaengine: edma: Correct PaRAM access function names (_parm_ to _param_)

These inline functions are designed to modify parts of the PaRAM in eDMA.
Change the names accordingly.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Simplify function parameter list for channel operations
Peter Ujfalusi [Fri, 16 Oct 2015 07:18:01 +0000 (10:18 +0300)]
dmaengine: edma: Simplify function parameter list for channel operations

Instead of passing a pointer to struct edma_cc and the channel number,
pass only the pointer to the edma_chan structure for the given channel.
This struct contains all the information needed by the functions and the
use of this makes it obvious that most of the sanity checks can be removed
from the driver.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Optimize memcpy operation
Peter Ujfalusi [Fri, 16 Oct 2015 07:18:00 +0000 (10:18 +0300)]
dmaengine: edma: Optimize memcpy operation

If the transfer is shorted then 64K we can complete it with one ACNT burst
by configuring ACNT to the length of the copy, this require one paRAM slot.
Otherwise we use two paRAM slots for the copy:
slot1: will copy (length / 32767) number of 32767 byte long blocks
slot2: will be configured to copy the remaining data.

According to tests this patch increases the throughput of memcpy from
~3MB/s to 15MB/s

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Remove alignment constraint for memcpy
Peter Ujfalusi [Fri, 16 Oct 2015 07:17:59 +0000 (10:17 +0300)]
dmaengine: edma: Remove alignment constraint for memcpy

Despite the claim by the original commit adding the memcpy
support, eDMA does not have constraint on the alignment of src, dst
or length in increment mode.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: ste_dma40: fix a trivial typo
Geliang Tang [Sun, 18 Oct 2015 15:31:10 +0000 (23:31 +0800)]
dmaengine: ste_dma40: fix a trivial typo

s/regsiter/register/

Signed-off-by: Geliang Tang <geliangtang@163.com>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: xgene-dma: Remove memcpy offload support due to performance drop
Rameshwar Prasad Sahu [Wed, 14 Oct 2015 13:55:07 +0000 (19:25 +0530)]
dmaengine: xgene-dma: Remove memcpy offload support due to performance drop

The DMA engine supports memory copy, RAID5 XOR, RAID6 PQ, and other
computations. But the bandwidth of the entire DMA engine is shared
among all channels. This patch re-configures operations availability
such that one can achieve maximum performance for XOR and PQ
computation by removing the memory offload operations.

Signed-off-by: Rameshwar Prasad Sahu <rsahu@apm.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Dynamic paRAM slot handling if HW supports it
Peter Ujfalusi [Wed, 14 Oct 2015 11:43:06 +0000 (14:43 +0300)]
dmaengine: edma: Dynamic paRAM slot handling if HW supports it

If the eDMA3 has support for channel paRAM slot mapping we can utilize it
to allocate slots on demand and save precious slots for real transfers.
On am335x the eDMA has 64 channels which means we can unlock 64 paRAM
slots out from the available 256.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Rename bitfields for slot and channel usage tracking
Peter Ujfalusi [Wed, 14 Oct 2015 11:43:05 +0000 (14:43 +0300)]
dmaengine: edma: Rename bitfields for slot and channel usage tracking

The names chosen for the bitfields were quite confusing and given no real
information on what they are used for...

edma_inuse -> slot_inuse: tracks the slot usage/availability
edma_unused -> channel_unused: tracks the channel usage/availability

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Read channel mapping support only once from HW
Peter Ujfalusi [Wed, 14 Oct 2015 11:43:04 +0000 (14:43 +0300)]
dmaengine: edma: Read channel mapping support only once from HW

Instead of directly reading it from CCCFG register take the information out
once when we set up the configuration from the HW.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Simplify and optimize ccerr interrupt handler
Peter Ujfalusi [Wed, 14 Oct 2015 11:43:03 +0000 (14:43 +0300)]
dmaengine: edma: Simplify and optimize ccerr interrupt handler

No need to run through the bits in QEMR and CCERR events since they will
not trigger any action, so just clearing the errors there is fine.
In case of the missed event the loop can be optimized so we spend less time
to handle the event.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Move the pending error check into helper function
Peter Ujfalusi [Wed, 14 Oct 2015 11:43:02 +0000 (14:43 +0300)]
dmaengine: edma: Move the pending error check into helper function

In the ccerr interrupt handler the code checks for pending errors in the
error status registers in two different places.
Move the check out to a helper function.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Simplify the interrupt handling
Peter Ujfalusi [Wed, 14 Oct 2015 11:43:01 +0000 (14:43 +0300)]
dmaengine: edma: Simplify the interrupt handling

With the merger of the arch/arm/common/edma.c code into the dmaengine
driver, there is no longer need to have per channel callback/data storage
for interrupt events.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Consolidate the comments for functions
Peter Ujfalusi [Wed, 14 Oct 2015 11:43:00 +0000 (14:43 +0300)]
dmaengine: edma: Consolidate the comments for functions

Remove or rewrite the comments for the internal functions.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Print warning when linking slots from different eDMA
Peter Ujfalusi [Wed, 14 Oct 2015 11:42:59 +0000 (14:42 +0300)]
dmaengine: edma: Print warning when linking slots from different eDMA

Warning message in case of linking between paRAM slots in different eDMA
controllers.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Use the edma_write_slot instead open coded memcpy_toio
Peter Ujfalusi [Wed, 14 Oct 2015 11:42:58 +0000 (14:42 +0300)]
dmaengine: edma: Use the edma_write_slot instead open coded memcpy_toio

edma_write_slot() is for writing an entire paRAM slot.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Use dev_dbg instead pr_debug
Peter Ujfalusi [Wed, 14 Oct 2015 11:42:57 +0000 (14:42 +0300)]
dmaengine: edma: Use dev_dbg instead pr_debug

We have access to dev, so it is better to use the dev_dbg for debug prints.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Cleanup regarding the use of dev around the code
Peter Ujfalusi [Wed, 14 Oct 2015 11:42:56 +0000 (14:42 +0300)]
dmaengine: edma: Cleanup regarding the use of dev around the code

Be consistent and do not mix the use of dev, &pdev->dev, etc in the
functions.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Use devm_kcalloc when possible
Peter Ujfalusi [Wed, 14 Oct 2015 11:42:55 +0000 (14:42 +0300)]
dmaengine: edma: Use devm_kcalloc when possible

When allocating a memory for number of items it is better (looks better)
to use devm_kcalloc.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Allocate memory dynamically for bitmaps and structures
Peter Ujfalusi [Wed, 14 Oct 2015 11:42:54 +0000 (14:42 +0300)]
dmaengine: edma: Allocate memory dynamically for bitmaps and structures

Instead of using defines to specify the size of different arrays and
bitmaps, allocate the memory for them based on the information we get from
the HW itself.
Since these defines are set based on the worst case, there are devices
where they are not valid.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agoARM/dmaengine: edma: Merge the two drivers under drivers/dma/
Peter Ujfalusi [Wed, 14 Oct 2015 11:42:53 +0000 (14:42 +0300)]
ARM/dmaengine: edma: Merge the two drivers under drivers/dma/

Move the code out from arch/arm/common and merge it inside of the dmaengine
driver.
This change is done with as minimal (if eny) functional change to the code
as possible to avoid introducing regression.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Acked-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agoARM: davinci: Add dma_mask to eDMA devices
Peter Ujfalusi [Wed, 14 Oct 2015 11:42:52 +0000 (14:42 +0300)]
ARM: davinci: Add dma_mask to eDMA devices

The upcoming change to merge the arch/arm/common/edma.c into
drivers/dma/edma.c will need this change when booting daVinci devices in
no DT mode.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Acked-by: Sekhar Nori <nsekhar@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agoARM: davinci: Use platform_device_register_full() to create pdev for eDMA
Peter Ujfalusi [Wed, 14 Oct 2015 11:42:51 +0000 (14:42 +0300)]
ARM: davinci: Use platform_device_register_full() to create pdev for eDMA

Convert the eDMA platform device creation to use
struct platform_device_info XXXXXX __initconst and
platform_device_register_full()
This will allow us to cleanly specify the dma_mask for the devices in an
upcoming patch.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Acked-by: Sekhar Nori <nsekhar@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agoARM/dmaengine: edma: Remove limitation on the number of eDMA controllers
Peter Ujfalusi [Wed, 14 Oct 2015 11:42:50 +0000 (14:42 +0300)]
ARM/dmaengine: edma: Remove limitation on the number of eDMA controllers

Since the driver stack no longer depends on lookup with id number in a
global array of pointers, the limitation for the number of eDMAs are no
longer needed. We can handle as many eDMAs in legacy and DT boot as we have
memory for them to allocate the needed structures.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agoARM/dmaengine: edma: Public API to use private struct pointer
Peter Ujfalusi [Wed, 14 Oct 2015 11:42:49 +0000 (14:42 +0300)]
ARM/dmaengine: edma: Public API to use private struct pointer

Instead of relying on indexes pointing to edma private date in the global
pointer array, pass the private data pointer via the public API.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agoARM: common: edma: Internal API to use pointer to 'struct edma'
Peter Ujfalusi [Wed, 14 Oct 2015 11:42:48 +0000 (14:42 +0300)]
ARM: common: edma: Internal API to use pointer to 'struct edma'

Merge the iomem into the 'struct edma' and change the internal (static)
functions to use pointer to the edma_cc instead of the ctlr number.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agoARM/dmaengine: edma: Move of_dma_controller_register to the dmaengine driver
Peter Ujfalusi [Wed, 14 Oct 2015 11:42:47 +0000 (14:42 +0300)]
ARM/dmaengine: edma: Move of_dma_controller_register to the dmaengine driver

If the of_dma_controller is registered in the non dmaengine driver we could
have race condition:
the of_dma_controller has been registered, but the dmaengine driver is not
yet probed. Drivers requesting DMA channels during this window will fail
since we do not yet have dmaengine drivers registered.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agoARM: davinci/common: Convert edma driver to handle one eDMA instance per driver
Peter Ujfalusi [Wed, 14 Oct 2015 11:42:46 +0000 (14:42 +0300)]
ARM: davinci/common: Convert edma driver to handle one eDMA instance per driver

Currently we have one device created to handle all (maximum 2) eDMAs in the
system.
With this change all eDMA instance will have it's own device/driver.
This change is needed for further cleanups in the eDMA driver stack since
the one device/driver to handle all eDMAs in the system was not flexible
enough and prevents the upcoming work.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Acked-by: Sekhar Nori <nsekhar@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: Simplify and optimize the edma_execute path
Peter Ujfalusi [Wed, 14 Oct 2015 11:42:45 +0000 (14:42 +0300)]
dmaengine: edma: Simplify and optimize the edma_execute path

The code path in edma_execute() and edma_callback() can be simplified
and make it more optimal.
There is not need to call in to edma_execute() when the transfer
has been finished for example.
Also the handling of missed/first or next batch of paRAMs can
be done in a more optimal way.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agoARM: common: edma: Remove unused functions
Peter Ujfalusi [Wed, 14 Oct 2015 11:42:44 +0000 (14:42 +0300)]
ARM: common: edma: Remove unused functions

We no longer have users for these functions so they can be removed.
Remove also unused enums from the header file.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agoARM: common: edma: Fix channel parameter for irq callbacks
Peter Ujfalusi [Wed, 14 Oct 2015 11:42:43 +0000 (14:42 +0300)]
ARM: common: edma: Fix channel parameter for irq callbacks

In case when the interrupt happened for the second eDMA the channel
number was incorrectly passed to the client driver.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
CC: <stable@vger.kernel.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: imx-sdma: Remove unneeded dev_info()
Fabio Estevam [Sat, 3 Oct 2015 22:37:59 +0000 (19:37 -0300)]
dmaengine: imx-sdma: Remove unneeded dev_info()

There is no need to print that the driver has been initialized
or removed, so remove such messages.

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: imx-sdma: Move message level to debug
Fabio Estevam [Sat, 3 Oct 2015 22:37:58 +0000 (19:37 -0300)]
dmaengine: imx-sdma: Move message level to debug

Since commit d078cd1b4185 ("dmaengine: imx-sdma: Add imx6sx platform
support") we get this message on every boot on mx6q:

imx-sdma 20ec000.sdma: no event needs to be remapped

, which is not very helpful.

Move the message to debug level instead.

Cc: Zidan Wang <zidan.wang@freescale.com>
Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: omap-dma: Enable packed accesses for cyclic transfers
Misael Lopez Cruz [Mon, 14 Sep 2015 12:31:05 +0000 (15:31 +0300)]
dmaengine: omap-dma: Enable packed accesses for cyclic transfers

The L3 throughput can be higher than expected when packed access
is not enabled.  The ratio depends on the number of bytes in a
transaction and the EMIF interface width.

The throughput was measured for the following settings/cases:

* Case 1: Burst size of 64 bytes, packed access disabled
* Case 2: Burst size of 64 bytes, packed access enabled
* Case 3: Burst disabled, packed access disabled

Throughput measurements were done during McASP-based audio
playback on the Jacinto6 EVM using the omapconf tool [1]:
$ omapconf trace bw -m sdma_rd

 ---------------------------------------------------------
                                  Throughput (MB/s)
  Audio parameters            Case 1    Case 2    Case 3
 ---------------------------------------------------------
  44.1kHz, 16-bits, stereo      1.41      0.18      1.41
  44.1kHz, 32-bits, stereo      1.41      0.35      1.41
  44.1kHz, 16-bits, 4-chan      2.82      0.35      2.82
  44.1kHz, 16-bits, 6-chan      4.23      0.53      4.23
  44.1kHz, 16-bits, 8-chan      5.64      0.71      5.64
 ---------------------------------------------------------

From above measurements, case 2 is the only one that delivers
the expected throughput for the given audio parameters.  For
that reason, the packed accesses are now enabled.

It's worth to mention that packed accesses cannot be enabled
for all addressing modes. In cyclic transfers, it can be
enabled in the source for MEM_TO_DEV and in dest for DEV_TO_MEM,
as they use post-increment mode which supports packed accesses.

Peter Ujfalusi:
From the TRM regarding to this:
"NOTE: Except in the constant addressing mode, the source or
destination must be specified as packed for burst transactions
to occur."

So w/o the packed setting the burst on the MEM side was not
enabled, this explains the numbers.

[1] https://github.com/omapconf/omapconf

Signed-off-by: Misael Lopez Cruz <misael.lopez@ti.com>
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: OF DMAEngine API based on CONFIG_DMA_OF instead of CONFIG_OF
Kuninori Morimoto [Fri, 21 Aug 2015 11:48:37 +0000 (11:48 +0000)]
dmaengine: OF DMAEngine API based on CONFIG_DMA_OF instead of CONFIG_OF

5fa422c ("dmaengine: move drivers/of/dma.c -> drivers/dma/of-dma.c")
moved OF base DMAEngine code to of-dma.c, then it based on CONFIG_DMA_OF.
But, OF base DMAEngine API on of_dma.h still based on CONFIG_OF now.
So, current kernel can't find OF base DMAEngine API if .config has CONFIG_OF,
but not have CONFIG_DMA_OF. This patch tidyup it.

Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: xilinx: Fix module autoload for OF platform driver
Luis de Bethencourt [Wed, 16 Sep 2015 21:00:17 +0000 (23:00 +0200)]
dmaengine: xilinx: Fix module autoload for OF platform driver

This platform driver has a OF device ID table but the OF module
alias information is not created so module autoloading won't work.

Signed-off-by: Luis de Bethencourt <luisbg@osg.samsung.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Moritz Fischer <moritz.fischer@ettus.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: sun6i: Fix module autoload for OF platform driver
Luis de Bethencourt [Wed, 16 Sep 2015 20:59:31 +0000 (22:59 +0200)]
dmaengine: sun6i: Fix module autoload for OF platform driver

This platform driver has a OF device ID table but the OF module
alias information is not created so module autoloading won't work.

Signed-off-by: Luis de Bethencourt <luisbg@osg.samsung.com>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: sirf: Fix module autoload for OF platform driver
Luis de Bethencourt [Wed, 16 Sep 2015 20:58:53 +0000 (22:58 +0200)]
dmaengine: sirf: Fix module autoload for OF platform driver

This platform driver has a OF device ID table but the OF module
alias information is not created so module autoloading won't work.

Signed-off-by: Luis de Bethencourt <luisbg@osg.samsung.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: mpc512x: Fix module autoload for OF platform driver
Luis de Bethencourt [Wed, 16 Sep 2015 20:58:22 +0000 (22:58 +0200)]
dmaengine: mpc512x: Fix module autoload for OF platform driver

This platform driver has a OF device ID table but the OF module
alias information is not created so module autoloading won't work.

Signed-off-by: Luis de Bethencourt <luisbg@osg.samsung.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: moxart-dma: Fix module autoload for OF platform driver
Luis de Bethencourt [Wed, 16 Sep 2015 20:57:51 +0000 (22:57 +0200)]
dmaengine: moxart-dma: Fix module autoload for OF platform driver

This platform driver has a OF device ID table but the OF module
alias information is not created so module autoloading won't work.

Signed-off-by: Luis de Bethencourt <luisbg@osg.samsung.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: fsldma: Fix module autoload for OF platform driver
Luis de Bethencourt [Wed, 16 Sep 2015 20:57:17 +0000 (22:57 +0200)]
dmaengine: fsldma: Fix module autoload for OF platform driver

This platform driver has a OF device ID table but the OF module
alias information is not created so module autoloading won't work.

Signed-off-by: Luis de Bethencourt <luisbg@osg.samsung.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: at_xdmac: fix memory leak in interleaved mode
Ludovic Desroches [Tue, 15 Sep 2015 13:38:24 +0000 (15:38 +0200)]
dmaengine: at_xdmac: fix memory leak in interleaved mode

In interleaved mode, when numf > 1, we have only one descriptor for the
transfer but this descriptor has to be added to the descs_list. If not,
when doing remove_xfer, the descriptor won't be put back in the
free_descs_list.

Signed-off-by: Ludovic Desroches <ludovic.desroches@atmel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: at_xdmac: handle numf > 1
Maxime Ripard [Tue, 15 Sep 2015 13:29:27 +0000 (15:29 +0200)]
dmaengine: at_xdmac: handle numf > 1

Handle 'numf > 1' case for interleaved mode.

Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Ludovic Desroches <ludovic.desroches@atmel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: zxdma: fix memset call
Vinod Koul [Thu, 1 Oct 2015 01:41:48 +0000 (07:11 +0530)]
dmaengine: zxdma: fix memset call

Fix the call to memset in this driver
[linux-4.2-next-20150911/drivers/dma/zx296702_dma.c:444]: (warning)
memset() called to fill 0 bytes of 'ds'.

Reported-by: David Binderman <dcb314@hotmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agoavr32: use dw_dmac autoconfiguration
Andy Shevchenko [Mon, 28 Sep 2015 15:57:07 +0000 (18:57 +0300)]
avr32: use dw_dmac autoconfiguration

Instead of hardcoding a platform data for dw_dmac let's use it's own
autoconfiguration feature. Thus, remove hardcoded values.

Tested on ATNGW100.

Acked-by: Hans-Christian Egtvedt <egtvedt@samfundet.no>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: dw: use dw_dmac autoconfiguration in PCI driver
Andy Shevchenko [Mon, 28 Sep 2015 15:57:05 +0000 (18:57 +0300)]
dmaengine: dw: use dw_dmac autoconfiguration in PCI driver

Instead of hardconding a platform data for dw_dmac let's use it's own
autoconfiguration feature. Thus, remove hardcoded values.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: dw: convert to __ffs()
Andy Shevchenko [Mon, 28 Sep 2015 15:57:04 +0000 (18:57 +0300)]
dmaengine: dw: convert to __ffs()

We replace __fls() by __ffs() since we have to find a *minimum* data width that
satisfies both source and destination.

While here, rename dwc_fast_fls() to dwc_fast_ffs() which it really is.

Fixes: 4c2d56c574db (dw_dmac: introduce dwc_fast_fls())
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: acpi: Use ACPI_COMPANION() instead of acpi_bus_get_device()
Jarkko Nikula [Fri, 4 Sep 2015 13:12:30 +0000 (16:12 +0300)]
dmaengine: acpi: Use ACPI_COMPANION() instead of acpi_bus_get_device()

Get pointer to the struct acpi_device by using ACPI_COMPANION() macro. This
is more efficient than using ACPI_HANDLE() and acpi_bus_get_device().

Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: edma: remove redundant conditions
Vinod Koul [Wed, 30 Sep 2015 06:36:36 +0000 (12:06 +0530)]
dmaengine: edma: remove redundant conditions

in edma_callback, driver was doing redundant check for desc, so remove that

Reported-by: David Binderman <dcb314@hotmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: idma64: use lo_hi_readq() / lo_hi_writeq()
Andy Shevchenko [Mon, 14 Sep 2015 08:55:41 +0000 (11:55 +0300)]
dmaengine: idma64: use lo_hi_readq() / lo_hi_writeq()

There are already helper functions to do 64-bit I/O on 32-bit machines, thus we
don't need to reinvent the wheel. In our case we can't use readq() / writeq()
even on 64-bit kernel since there is a hardware limitation (OCP bus is a 32-bit
bus).

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: idma64: make better performance on pause / resume
Andy Shevchenko [Mon, 14 Sep 2015 08:55:40 +0000 (11:55 +0300)]
dmaengine: idma64: make better performance on pause / resume

Accordingly to the documentation the CH_DRAIN bit enforses single bursts when
channel is going to be suspended. This, in case when channel will be resumed,
makes data to flow in non-optimal mode until DMA returns to full burst mode.
The fix differentiates pause / resume cycle from pause / terminate and sets
CH_DRAIN bit accordingly.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: idma64: this is not DesignWare
Andy Shevchenko [Mon, 14 Sep 2015 08:55:39 +0000 (11:55 +0300)]
dmaengine: idma64: this is not DesignWare

This patch fixes a comment where DesignWare is wrongly mentioned. There is no
functional change.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: idma64: useless use of min_t()
Andy Shevchenko [Mon, 14 Sep 2015 08:55:38 +0000 (11:55 +0300)]
dmaengine: idma64: useless use of min_t()

We use a pattern

x = min_t(u32, <LOG2_CONSTANT>, __ffs(expr));

There is no need to use min_t() since we can replace it by

x = __ffs(expr | <2^LOG2_CONST>);

and moreover guarantee that argument of __ffs() will be not zero.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: idma64: convert to __ffs()
Andy Shevchenko [Mon, 14 Sep 2015 08:55:37 +0000 (11:55 +0300)]
dmaengine: idma64: convert to __ffs()

We replace __fls() by __ffs() since we have to find a *minimum* data width that
satisfies both source and destination.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: idma: rename to INTEL_IDMA64
Vinod Koul [Mon, 21 Sep 2015 16:48:45 +0000 (22:18 +0530)]
dmaengine: idma: rename to INTEL_IDMA64

the symbol CONFIG_IDMA64 should rather be CONFIG_INTEL_IDMA64 to conform to
rest of the intel dmaengine drivers. This was found after sorting the
entries and trying to place this odd one

Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: drop null test before destroy functions
Julia Lawall [Sun, 13 Sep 2015 12:15:19 +0000 (14:15 +0200)]
dmaengine: 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: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: ioatdma: add PCIe AER handlers
Dave Jiang [Wed, 26 Aug 2015 20:17:30 +0000 (13:17 -0700)]
dmaengine: ioatdma: add PCIe AER handlers

Adding AER handlers in order to handle any PCIe errors.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: ioatdma: adding shutdown support
Dave Jiang [Wed, 26 Aug 2015 20:17:24 +0000 (13:17 -0700)]
dmaengine: ioatdma: adding shutdown support

The ioatdma needs to be queisced and block all additional op submission
during reboots. When NET_DMA was used, this caused issue as ops were still
being sent to ioatdma during reboots even though PCI BME has been turned
off. Even though NET_DMA has been deprecated, we need to prevent similar
situations. The shutdown handler should address that.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agodmaengine: xgene-dma: use dma_pool_zalloc
Vinod Koul [Mon, 21 Sep 2015 15:26:58 +0000 (20:56 +0530)]
dmaengine: xgene-dma: use dma_pool_zalloc

We should use shiny new dma_pool_zalloc instead of
dma_pool_alloc/memset

Reported-by: Julia Lawall <julia.lawall@lip6.fr>
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
9 years agoLinux 4.3-rc1
Linus Torvalds [Sat, 12 Sep 2015 23:35:56 +0000 (16:35 -0700)]
Linux 4.3-rc1

9 years agoMerge tag 'cris-for-4.3' of git://git.kernel.org/pub/scm/linux/kernel/git/jesper...
Linus Torvalds [Sat, 12 Sep 2015 19:24:29 +0000 (12:24 -0700)]
Merge tag 'cris-for-4.3' of git://git./linux/kernel/git/jesper/cris

Pull CRIS updates from Jesper Nilsson:
 "Mostly removal of old cruft of which we can use a generic version, or
  fixes for code not commonly run in the cris port, but also additions
  to enable some good debug"

* tag 'cris-for-4.3' of git://git.kernel.org/pub/scm/linux/kernel/git/jesper/cris: (25 commits)
  CRISv10: delete unused lib/dmacopy.c
  CRISv10: delete unused lib/old_checksum.c
  CRIS: fix switch_mm() lockdep splat
  CRISv32: enable LOCKDEP_SUPPORT
  CRIS: add STACKTRACE_SUPPORT
  CRISv32: annotate irq enable in idle loop
  CRISv32: add support for irqflags tracing
  CRIS: UAPI: use generic types.h
  CRIS: UAPI: use generic shmbuf.h
  CRIS: UAPI: use generic msgbuf.h
  CRIS: UAPI: use generic socket.h
  CRIS: UAPI: use generic sembuf.h
  CRIS: UAPI: use generic sockios.h
  CRIS: UAPI: use generic auxvec.h
  CRIS: UAPI: use generic headers via Kbuild
  CRIS: UAPI: fix elf.h export
  CRIS: don't make asm/elf.h depend on asm/user.h
  CRIS: UAPI: fix ptrace.h
  CRISv32: Squash compile warnings for axisflashmap
  CRISv32: Add GPIO driver to the default configs
  ...

9 years agoblk: rq_data_dir() should not return a boolean
Linus Torvalds [Wed, 27 May 2015 22:32:15 +0000 (15:32 -0700)]
blk: rq_data_dir() should not return a boolean

rq_data_dir() returns either READ or WRITE (0 == READ, 1 == WRITE), not
a boolean value.

Now, admittedly the "!= 0" doesn't really change the value (0 stays as
zero, 1 stays as one), but it's not only redundant, it confuses gcc, and
causes gcc to warn about the construct

    switch (rq_data_dir(req)) {
        case READ:
            ...
        case WRITE:
            ...

that we have in a few drivers.

Now, the gcc warning is silly and stupid (it seems to warn not about the
switch value having a different type from the case statements, but about
_any_ boolean switch value), but in this case the code itself is silly
and stupid too, so let's just change it, and get rid of warnings like
this:

  drivers/block/hd.c: In function ‘hd_request’:
  drivers/block/hd.c:630:11: warning: switch condition has boolean value [-Wswitch-bool]
     switch (rq_data_dir(req)) {

The odd '!= 0' came in when "cmd_flags" got turned into a "u64" in
commit 5953316dbf90 ("block: make rq->cmd_flags be 64-bit") and is
presumably because the old code (that just did a logical 'and' with 1)
would then end up making the type of rq_data_dir() be u64 too.

But if we want to retain the old regular integer type, let's just cast
the result to 'int' rather than use that rather odd '!= 0'.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoMerge branch 'writeback-plugging'
Linus Torvalds [Sat, 12 Sep 2015 18:19:01 +0000 (11:19 -0700)]
Merge branch 'writeback-plugging'

Fix up the writeback plugging introduced in commit d353d7587d02
("writeback: plug writeback at a high level") that then caused problems
due to the unplug happening with a spinlock held.

* writeback-plugging:
  writeback: plug writeback in wb_writeback() and writeback_inodes_wb()
  Revert "writeback: plug writeback at a high level"

9 years agowriteback: plug writeback in wb_writeback() and writeback_inodes_wb()
Linus Torvalds [Fri, 11 Sep 2015 20:37:19 +0000 (13:37 -0700)]
writeback: plug writeback in wb_writeback() and writeback_inodes_wb()

We had to revert the pluggin in writeback_sb_inodes() because the
wb->list_lock is held, but we could easily plug at a higher level before
taking that lock, and unplug after releasing it.  This does that.

Chris will run performance numbers, just to verify that this approach is
comparable to the alternative (we could just drop and re-take the lock
around the blk_finish_plug() rather than these two commits.

I'd have preferred waiting for actual performance numbers before picking
one approach over the other, but I don't want to release rc1 with the
known "sleeping function called from invalid context" issue, so I'll
pick this cleanup version for now.  But if the numbers show that we
really want to plug just at the writeback_sb_inodes() level, and we
should just play ugly games with the spinlock, we'll switch to that.

Cc: Chris Mason <clm@fb.com>
Cc: Josef Bacik <jbacik@fb.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Neil Brown <neilb@suse.de>
Cc: Jan Kara <jack@suse.cz>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agothermal: fix intel PCH thermal driver mismerge
Linus Torvalds [Sat, 12 Sep 2015 03:06:59 +0000 (20:06 -0700)]
thermal: fix intel PCH thermal driver mismerge

I didn't notice this when merging the thermal code from Zhang, but his
merge (commit 5a924a07f882: "Merge branches 'thermal-core' and
'thermal-intel' of .git into next") of the thermal-core and
thermal-intel branches was wrong.

In thermal-core, commit 17e8351a7739 ("thermal: consistently use int for
temperatures") converted the thermal layer to use "int" for
temperatures.

But in parallel, in the thermal-intel branch commit d0a12625d2ff
("thermal: Add Intel PCH thermal driver") added support for the intel
PCH thermal sensor using the old interfaces that used "unsigned long"
pointers.

This resulted in warnings like this:

  drivers/thermal/intel_pch_thermal.c:184:14: warning: initialization from incompatible pointer type [-Wincompatible-pointer-types]
    .get_temp = pch_thermal_get_temp,
                ^
  drivers/thermal/intel_pch_thermal.c:184:14: note: (near initialization for ‘tzd_ops.get_temp’)
  drivers/thermal/intel_pch_thermal.c:186:19: warning: initialization from incompatible pointer type [-Wincompatible-pointer-types]
    .get_trip_temp = pch_get_trip_temp,
                     ^
  drivers/thermal/intel_pch_thermal.c:186:19: note: (near initialization for ‘tzd_ops.get_trip_temp’)

This fixes it.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Sat, 12 Sep 2015 02:34:09 +0000 (19:34 -0700)]
Merge branch 'akpm' (patches from Andrew)

Merge fourth patch-bomb from Andrew Morton:

 - sys_membarier syscall

 - seq_file interface changes

 - a few misc fixups

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  revert "ocfs2/dlm: use list_for_each_entry instead of list_for_each"
  mm/early_ioremap: add explicit #include of asm/early_ioremap.h
  fs/seq_file: convert int seq_vprint/seq_printf/etc... returns to void
  selftests: enhance membarrier syscall test
  selftests: add membarrier syscall test
  sys_membarrier(): system-wide memory barrier (generic, x86)
  MODSIGN: fix a compilation warning in extract-cert

9 years agoARCv2: [axs103_smp] Reduce clk for SMP FPGA configs
Vineet Gupta [Fri, 11 Sep 2015 23:32:22 +0000 (16:32 -0700)]
ARCv2: [axs103_smp] Reduce clk for SMP FPGA configs

Newer bitfiles needs the reduced clk even for SMP builds

Cc: <stable@vger.kernel.org> #4.2
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoMerge tag 'ntb-4.3' of git://github.com/jonmason/ntb
Linus Torvalds [Sat, 12 Sep 2015 02:29:00 +0000 (19:29 -0700)]
Merge tag 'ntb-4.3' of git://github.com/jonmason/ntb

Pull NTB fixes from Jon Mason:
 "NTB bug and documentation fixes, new device IDs, performance
  improvements, and adding a mailing list to MAINTAINERS for NTB"

* tag 'ntb-4.3' of git://github.com/jonmason/ntb:
  NTB: Fix range check on memory window index
  NTB: Improve index handling in B2B MW workaround
  NTB: Fix documentation for ntb_peer_db_clear.
  NTB: Fix documentation for ntb_link_is_up
  NTB: Use unique DMA channels for TX and RX
  NTB: Remove dma_sync_wait from ntb_async_rx
  NTB: Clean up QP stats info
  NTB: Make the transport list in order of discovery
  NTB: Add PCI Device IDs for Broadwell Xeon
  NTB: Add flow control to the ntb_netdev
  NTB: Add list to MAINTAINERS

9 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Sat, 12 Sep 2015 02:17:28 +0000 (19:17 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

Pull more input updates from Dmitry Torokhov:
 "Second round of updates for the input subsystem.

  This introduces two brand new touchscreen drivers (Colibri and
  imx6ul_tsc), some small driver fixes, and we are no longer report
  errors from evdev_flush() as users do not really have a way of
  handling errors, error codes that we were returning were not on the
  list of errors supposed to be returned by close(), and errors were
  causing issues with one of older versions of systemd"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: imx_keypad - remove obsolete comment
  Input: touchscreen - add imx6ul_tsc driver support
  Input: Add touchscreen support for Colibri VF50
  Input: i8042 - lower log level for "no controller" message
  Input: evdev - do not report errors form flush()
  Input: elants_i2c - extend the calibration timeout to 12 seconds
  Input: sparcspkr - fix module autoload for OF platform drivers
  Input: regulator-haptic - fix module autoload for OF platform driver
  Input: pwm-beeper - fix module autoload for OF platform driver
  Input: ab8500-ponkey - Fix module autoload for OF platform driver
  Input: cyttsp - remove unnecessary MODULE_ALIAS()
  Input: elan_i2c - add ACPI ID "ELAN1000"

9 years agoMerge tag 'pm+acpi-4.3-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Sat, 12 Sep 2015 02:11:06 +0000 (19:11 -0700)]
Merge tag 'pm+acpi-4.3-rc1-2' of git://git./linux/kernel/git/rafael/linux-pm

Pull more power management and ACPI updates from Rafael Wysocki:
 "These are mostly fixes and cleanups on top of the previous PM+ACPI
  pull request (cpufreq core and drivers, cpuidle, generic power domains
  framework).  Some of them didn't make to that pull request and some
  fix issues introduced by it.

  The only really new thing is the support for suspend frequency in the
  cpufreq-dt driver, but it is needed to fix an issue with Exynos
  platforms.

  Specifics:

   - build fix for the new Mediatek MT8173 cpufreq driver (Guenter
     Roeck).

   - generic power domains framework fixes (power on error code path,
     subdomain removal) and cleanup of a deprecated API user (Geert
     Uytterhoeven, Jon Hunter, Ulf Hansson).

   - cpufreq-dt driver fixes including two fixes for bugs related to the
     new Operating Performance Points Device Tree bindings introduced
     recently (Viresh Kumar).

   - suspend frequency support for the cpufreq-dt driver (Bartlomiej
     Zolnierkiewicz, Viresh Kumar).

   - cpufreq core cleanups (Viresh Kumar).

   - intel_pstate driver fixes (Chen Yu, Kristen Carlson Accardi).

   - additional sanity check in the cpuidle core (Xunlei Pang).

   - fix for a comment related to CPU power management (Lina Iyer)"

* tag 'pm+acpi-4.3-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  intel_pstate: fix PCT_TO_HWP macro
  intel_pstate: Fix user input of min/max to legal policy region
  PM / OPP: Return suspend_opp only if it is enabled
  cpufreq-dt: add suspend frequency support
  cpufreq: allow cpufreq_generic_suspend() to work without suspend frequency
  PM / OPP: add dev_pm_opp_get_suspend_opp() helper
  staging: board: Migrate away from __pm_genpd_name_add_device()
  cpufreq: Use __func__ to print function's name
  cpufreq: staticize cpufreq_cpu_get_raw()
  PM / Domains: Ensure subdomain is not in use before removing
  cpufreq: Add ARM_MT8173_CPUFREQ dependency on THERMAL
  cpuidle/coupled: Add sanity check for safe_state_index
  PM / Domains: Try power off masters in error path of __pm_genpd_poweron()
  cpufreq: dt: Tolerance applies on both sides of target voltage
  cpufreq: dt: Print error on failing to mark OPPs as shared
  cpufreq: dt: Check OPP count before marking them shared
  kernel/cpu_pm: fix cpu_cluster_pm_exit comment

9 years agoMerge branch 'for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/nab/target...
Linus Torvalds [Sat, 12 Sep 2015 02:00:42 +0000 (19:00 -0700)]
Merge branch 'for-next' of git://git./linux/kernel/git/nab/target-pending

Pull SCSI target updates from Nicholas Bellinger:
 "Here are the outstanding target-pending updates for v4.3-rc1.

  Mostly bug-fixes and minor changes this round.  The fallout from the
  big v4.2-rc1 RCU conversion have (thus far) been minimal.

  The highlights this round include:

   - Move sense handling routines into scsi_common code (Sagi)

   - Return ABORTED_COMMAND sense key for PI errors (Sagi)

   - Add tpg_enabled_sendtargets attribute for disabled iscsi-target
     discovery (David)

   - Shrink target struct se_cmd by rearranging fields (Roland)

   - Drop iSCSI use of mutex around max_cmd_sn increment (Roland)

   - Replace iSCSI __kernel_sockaddr_storage with sockaddr_storage (Andy +
     Chris)

   - Honor fabric max_data_sg_nents I/O transfer limit (Arun + Himanshu +
     nab)

   - Fix EXTENDED_COPY >= v4.1 regression OOPsen (Alex + nab)"

* 'for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending: (37 commits)
  target: use stringify.h instead of own definition
  target/user: Fix UFLAG_UNKNOWN_OP handling
  target: Remove no-op conditional
  target/user: Remove unused variable
  target: Fix max_cmd_sn increment w/o cmdsn mutex regressions
  target: Attach EXTENDED_COPY local I/O descriptors to xcopy_pt_sess
  target/qla2xxx: Honor max_data_sg_nents I/O transfer limit
  target/iscsi: Replace __kernel_sockaddr_storage with sockaddr_storage
  target/iscsi: Replace conn->login_ip with login_sockaddr
  target/iscsi: Keep local_ip as the actual sockaddr
  target/iscsi: Fix np_ip bracket issue by removing np_ip
  target: Drop iSCSI use of mutex around max_cmd_sn increment
  qla2xxx: Update tcm_qla2xxx module description to 24xx+
  iscsi-target: Add tpg_enabled_sendtargets for disabled discovery
  drivers: target: Drop unlikely before IS_ERR(_OR_NULL)
  target: check DPO/FUA usage for COMPARE AND WRITE
  target: Shrink struct se_cmd by rearranging fields
  target: Remove cmd->se_ordered_id (unused except debug log lines)
  target: add support for START_STOP_UNIT SCSI opcode
  target: improve unsupported opcode message
  ...

9 years agoMerge tag 'scsi-misc' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Sat, 12 Sep 2015 01:15:18 +0000 (18:15 -0700)]
Merge tag 'scsi-misc' of git://git./linux/kernel/git/jejb/scsi

Pull second round of SCSI updates from James Bottomley:
 "There's one late arriving patch here (added today), fixing a build
  issue which the scsi_dh patch set in here uncovered.  Other than that,
  everything has been incubated in -next and the checkers for a week.

  The major pieces of this patch are a set patches facilitating better
  integration between scsi and scsi_dh (the device handling layer used
  by multi-path; all the dm parts are acked by Mike Snitzer).

  This also includes driver updates for mp3sas, scsi_debug and an
  assortment of bug fixes"

* tag 'scsi-misc' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi: (50 commits)
  scsi_dh: fix randconfig build error
  scsi: fix scsi_error_handler vs. scsi_host_dev_release race
  fcoe: Convert use of __constant_htons to htons
  mpt2sas: setpci reset kernel oops fix
  pm80xx: Don't override ts->stat on IO_OPEN_CNX_ERROR_HW_RESOURCE_BUSY
  lpfc: Fix possible use-after-free and double free in lpfc_mbx_cmpl_rdp_page_a2()
  bfa: Fix incorrect de-reference of pointer
  bfa: Fix indentation
  scsi_transport_sas: Remove check for SAS expander when querying bay/enclosure IDs.
  scsi_debug: resp_request: remove unused variable
  scsi_debug: fix REPORT LUNS Well Known LU
  scsi_debug: schedule_resp fix input variable check
  scsi_debug: make dump_sector static
  scsi_debug: vfree is null safe so drop the check
  scsi_debug: use SCSI_W_LUN_REPORT_LUNS instead of SAM2_WLUN_REPORT_LUNS;
  scsi_debug: define pr_fmt() for consistent logging
  mpt2sas: Refcount fw_events and fix unsafe list usage
  mpt2sas: Refcount sas_device objects and fix unsafe list usage
  scsi_dh: return SCSI_DH_NOTCONN in scsi_dh_activate()
  scsi_dh: don't allow to detach device handlers at runtime
  ...

9 years agoMerge tag 'media/v4.3-2' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab...
Linus Torvalds [Fri, 11 Sep 2015 23:42:39 +0000 (16:42 -0700)]
Merge tag 'media/v4.3-2' of git://git./linux/kernel/git/mchehab/linux-media

Pull media updates from Mauro Carvalho Chehab:
 "A series of patches that move part of the code used to allocate memory
  from the media subsystem to the mm subsystem"

[ The mm parts have been acked by VM people, and the series was
  apparently in -mm for a while   - Linus ]

* tag 'media/v4.3-2' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media:
  [media] drm/exynos: Convert g2d_userptr_get_dma_addr() to use get_vaddr_frames()
  [media] media: vb2: Remove unused functions
  [media] media: vb2: Convert vb2_dc_get_userptr() to use frame vector
  [media] media: vb2: Convert vb2_vmalloc_get_userptr() to use frame vector
  [media] media: vb2: Convert vb2_dma_sg_get_userptr() to use frame vector
  [media] vb2: Provide helpers for mapping virtual addresses
  [media] media: omap_vout: Convert omap_vout_uservirt_to_phys() to use get_vaddr_pfns()
  [media] mm: Provide new get_vaddr_frames() helper
  [media] vb2: Push mmap_sem down to memops

9 years agoMerge tag 'edac/v4.3-1' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab...
Linus Torvalds [Fri, 11 Sep 2015 23:21:12 +0000 (16:21 -0700)]
Merge tag 'edac/v4.3-1' of git://git./linux/kernel/git/mchehab/linux-edac

Pull edac updates from Mauro Carvalho Chehab:
 "Two EDAC fixes for Intel systems (Haswell and Ivy Bridge)"

* tag 'edac/v4.3-1' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-edac:
  sb_edac: correctly fetch DIMM width on Ivy Bridge and Haswell
  sb_edac: look harder for DDRIO on Haswell systems

9 years agoMerge branch 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/rzhang/linux
Linus Torvalds [Fri, 11 Sep 2015 23:13:47 +0000 (16:13 -0700)]
Merge branch 'next' of git://git./linux/kernel/git/rzhang/linux

Pull thermal updates from Zhang Rui:

 - use int instead of unsigned long to represent temperature to avoid
   bogus overheat detection when negative temperature reported.  From
   Sascha Hauer.

 - export available thermal governors information to user space via
   sysfs.  From Wei Ni.

 - introduce new thermal driver for Wildcat Point platform controller
   hub, which uses PCH thermal sensor and associated critical and hot
   trip points.  From Tushar Dave.

 - add suuport for Intel Skylake and Denlow platforms in powerclamp
   driver.

 - some small cleanups in thermal core.

* 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/rzhang/linux:
  thermal: Add Intel PCH thermal driver
  thermal: Add comment explaining test for critical temperature
  thermal: Use IS_ENABLED instead of #ifdef
  thermal: remove unnecessary call to thermal_zone_device_set_polling
  thermal: trivial: fix typo in comment
  thermal: consistently use int for temperatures
  thermal: add available policies sysfs attribute
  thermal/powerclamp: add cpu id for denlow platform
  thermal/powerclamp: add cpu id for Skylake u/y
  thermal/powerclamp: add cpu id for skylake h/s

9 years agorevert "ocfs2/dlm: use list_for_each_entry instead of list_for_each"
Andrew Morton [Fri, 11 Sep 2015 20:07:53 +0000 (13:07 -0700)]
revert "ocfs2/dlm: use list_for_each_entry instead of list_for_each"

Revert commit f83c7b5e9fd6 ("ocfs2/dlm: use list_for_each_entry instead
of list_for_each").

list_for_each_entry() will dereference its `pos' argument, which can be
NULL in dlm_process_recovery_data().

Reported-by: Julia Lawall <julia.lawall@lip6.fr>
Reported-by: Fengguang Wu <fengguang.wu@gmail.com>
Cc: Joseph Qi <joseph.qi@huawei.com>
Cc: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>