Merge branch 'topic/err_reporting' into for-linus

Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Conflicts:
drivers/dma/cppi41.c

Merge branch 'topic/dmatest' into for-linus

Merge branch 'topic/cppi' into for-linus

Merge branch 'topic/core' into for-linus

Merge branch 'topic/compile_test' into for-linus

dmaengine: virt-dma: move function declarations

We get 2 warnings when building kernel with W=1:
drivers/dma/virt-dma.c:22:14: warning: no previous prototype for 'vchan_tx_submit' [-Wmissing-prototypes]
drivers/dma/virt-dma.c:52:5: warning: no previous prototype for 'vchan_tx_desc_free' [-Wmissing-prototypes]

In fact, these two functions are incorrectly declared in a function.

So this patch moves function declarations out of this function.

Signed-off-by: Baoyou Xie <baoyou.xie@linaro.org>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

DT: dmaengine: rcar-dmac: document R8A7743/5 support

Renesas  RZ/G SoC also have the R-Car gen2/3 compatible DMA controllers.
Document RZ/G1[ME] (also known as R8A774[35]) SoC bindings.

Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Acked-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: fsldma: Unmap region obtained by of_iomap

Free memory mapping, if probe is not successful.

Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Acked-by: Li Yang <leoyang.li@nxp.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: jz4780: fix resource leaks on error exit return

In two cases when jz4780_dma_setup_hwdesc fails, there is a memory
leak on the allocated desc and associated DMA pools on the error
exit return path.  Fix this by free'ing the resources before
returning.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: coh901318: fix integer overflow when shifting more than 32 places

Currently U300_DMA_CHANNELS is set to 40, meaning that the shift of 1 can
be more than 32 places, which leads to a 32 bit integer overflow. Fix this
by using 1ULL instead of 1 before shifting it.  Also add braces on the
for-loop to keep with coding style conventions.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: edma: avoid uninitialized variable use

If edma_read_slot() gets an invalid argument, it does not set a result,
as found by "gcc -Wmaybe-uninitialized"

drivers/dma/edma.c: In function 'dma_ccerr_handler':
drivers/dma/edma.c:1499:21: error: 'p.a_b_cnt' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/dma/edma.c:1499:21: error: 'p.ccnt' may be used uninitialized in this function [-Werror=maybe-uninitialized]
  if (p.a_b_cnt == 0 && p.ccnt == 0) {

If we change the function to return an error in this case, we can handle
the failure more gracefully and treat this the same way as a null slot
that we already catch.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: ti-dma-crossbar: enable COMPILE_TEST

To get more coverage, enable COMPILE_TEST for this driver.
When compile testing eDMA or omap-dma, select also the ti-dma-crossbar so
it is also covered by the compile testing.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: omap-dma: enable COMPILE_TEST

To get more coverage, enable COMPILE_TEST for this driver.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: edma: enable COMPILE_TEST

To get more coverage, enable COMPILE_TEST for this driver.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: ti-dma-crossbar: Fix of_device_id data parameter usage

Use pointers to static constant variables for crossbar type and for DMA
offset configuration.

Fixes compiler warnings on 64bit architectures:

drivers/dma/ti-dma-crossbar.c: In function ‘ti_dra7_xbar_probe’:
drivers/dma/ti-dma-crossbar.c:398:21: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]
  xbar->dma_offset = (u32)match->data;
                     ^
drivers/dma/ti-dma-crossbar.c: In function ‘ti_dma_xbar_probe’:
drivers/dma/ti-dma-crossbar.c:431:10: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]
  switch ((u32)match->data) {
          ^

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: ti-dma-crossbar: Correct type for of_find_property() third parameter

The correct type is int and not for the third parameter of
of_find_property().
Fixes compilation for 64bit architectures (x86_64, aarch64).

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine/ARM: omap-dma: Fix the DMAengine compile test on non OMAP configs

The DMAengine driver for omap-dma use three function calls from the
plat-omap legacy driver. When the DMAengine driver is built when ARCH_OMAP
is not set, the compilation will fail due to missing symbols.
Add empty inline functions to allow the DMAengine driver to be compiled
with COMPILE_TEST.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: edma: Rename set_bits and remove unused clear_bits helper

The clear_bits() helper is not used by the driver so it can be removed.

powerpc architecture defines the set_bits() in
arch/powerpc/include/asm/bitops.h which results failed compile testing on
powerpc architecture:

>> drivers/dma/edma.c:415:20: error: conflicting types for 'set_bits'
    static inline void set_bits(int offset, int len, unsigned long *p)
                       ^~~~~~~~
   In file included from include/linux/bitops.h:36:0,
                    from include/linux/kernel.h:10,
                    from include/linux/list.h:8,
                    from include/linux/kobject.h:20,
                    from include/linux/device.h:17,
                    from include/linux/dmaengine.h:20,
                    from drivers/dma/edma.c:16:
   arch/powerpc/include/asm/bitops.h:75:14: note: previous definition of 'set_bits' was here
    DEFINE_BITOP(set_bits, or, "")
                 ^
   arch/powerpc/include/asm/bitops.h:58:24: note: in definition of macro 'DEFINE_BITOP'
    static __inline__ void fn(unsigned long mask, \
                           ^~

>> drivers/dma/edma.c:421:20: error: conflicting types for 'clear_bits'
    static inline void clear_bits(int offset, int len, unsigned long *p)
                       ^~~~~~~~~~
   In file included from include/linux/bitops.h:36:0,
                    from include/linux/kernel.h:10,
                    from include/linux/list.h:8,
                    from include/linux/kobject.h:20,
                    from include/linux/device.h:17,
                    from include/linux/dmaengine.h:20,
                    from drivers/dma/edma.c:16:
   arch/powerpc/include/asm/bitops.h:76:14: note: previous definition of 'clear_bits' was here
    DEFINE_BITOP(clear_bits, andc, "")
                 ^
   arch/powerpc/include/asm/bitops.h:58:24: note: in definition of macro 'DEFINE_BITOP'
    static __inline__ void fn(unsigned long mask, \
                           ^~

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: edma: Use correct type for of_find_property() third parameter

The correct type is int and not for the third parameter of
of_find_property().
Fixes compilation for 64bit architectures (x86_64, aarch64).

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: edma: Fix of_device_id data parameter usage (legacy vs TPCC)

Use pointers to static constant variables for eDMA binding
type (legacy vs TPCC).

Fixes the following warning when compiling the driver for 64bit
architectures (x86_64 for example):
drivers/dma/edma.c:2185:16: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]
   if (match && (u32)match->data == EDMA_BINDING_TPCC)
                ^

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: edma: Add missing MODULE_DEVICE_TABLE() for of_device_id structs

The MODULE_DEVICE_TABLE() were missing from the driver for the of_device_id
structures.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: tegra-adma: enable COMPILE_TEST

To get more coverage, enable COMPILE_TEST for this driver.

Suggested-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: sun6i: Add support for Allwinner A83T (sun8i) variant

The A83T SoC has the same dma engine as the A31 (sun6i), with a reduced
amount of endpoints and physical channels.

Signed-off-by: Jean-Francois Moine <moinejf@free.fr>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: s3c24xx: Add dma_slave_map for s3c2440 devices

This patch updates the s3c24xx dma driver to be able to pass a
dma_slave_map array via the platform data. This is needed to
be able to use the new, simpler dmaengine API [1].
I used the virtual DMA channels as a parameter for the dma_filter
function. By doing that, I could reuse the existing filter function in
drivers/dma/s3c24xx-dma.c.

I have tested this on my mini2440 board with the audio driver.
According to my observations, dma_request_slave_channel in the
function dmaengine_pcm_new in the file
sound/soc/soc-generic-dmaengine-pcm.c now returns a valid DMA channel
whereas before no DMA channel was returned at that point.

Entries for DMACH_XD0, DMACH_XD1 and DMACH_TIMER are missing because I
don't realy know which driver to use for these.

[1]
http://lists.infradead.org/pipermail/linux-arm-kernel/2015-December/393635.html

Signed-off-by: Sam Van Den Berge <sam.van.den.berge@telenet.be>
Reviewed-by: Sylwester Nawrocki <s.nawrocki@samsung.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: moxart: fix kconfig unmet direct dependencies

We have unmet direct dependencies as DMA_SUN4I selects DMA_OF so remove the
selection

warning: (DMA_SUN4I && MOXART_DMA && STM32_DMA) selects DMA_OF which has
unmet direct dependencies (DMADEVICES && OF)

Suggested-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: sun4i: fix kconfig unmet direct dependencies

We have unmet direct dependencies as DMA_SUN4I selects DMA_OF so remove the
selection

warning: (DMA_SUN4I && MOXART_DMA && STM32_DMA) selects DMA_OF which has
unmet direct dependencies (DMADEVICES && OF)

Suggested-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: zxdma: enable COMPILE_TEST

To get more coverage, enable COMPILE_TEST for this driver.

Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: timb-dma: enable COMPILE_TEST

To get more coverage, enable COMPILE_TEST for this driver.

Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: s3c24xx: enable COMPILE_TEST

To get more coverage, enable COMPILE_TEST for this driver.

Reviewed-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: s3c24xx: fix pointer cast warnings

On some systems, pointer can be large than unsigned int, triggering warning
pointer-to-int-cast on conversion.

drivers/dma/s3c24xx-dma.c: In function 's3c24xx_dma_filter':
drivers/dma/s3c24xx-dma.c:1421:24: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]

Use a long value for type conversion.

Suggested-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: s3c24xx: use correct print specifiers for size_t

This driver warns:

drivers/dma/s3c24xx-dma.c: In function 's3c24xx_dma_prep_memcpy':
drivers/dma/s3c24xx-dma.c:826:2: warning: format '%d' expects argument of type 'int', but argument 4 has type 'size_t' [-Wformat=]
drivers/dma/s3c24xx-dma.c:830:3: warning: format '%d' expects argument of type 'int', but argument 3 has type 'size_t' [-Wformat=]

We should use %zu to print 'size_t' values.

Reviewed-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: stm32-dma: make stm32_dma_set_config static

Sparse complains:
drivers/dma/stm32-dma.c:957:6: warning: symbol 'stm32_dma_set_config' was
not declared. Should it be static?

SO make stm32_dma_set_config static.

Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: stm32-dma: enable COMPILE_TEST

To get more coverage, enable COMPILE_TEST for this driver.

Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: mmp_tdma: enable COMPILE_TEST

To get more coverage, enable COMPILE_TEST for this driver.

While at it, to fix build on other archs, select MMP_SRAM only for ARCH_MMP
and also fix the platform header

Suggested-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: mmp_pdma: enable COMPILE_TEST

To get more coverage, enable COMPILE_TEST for this driver.

Acked-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: mmp_tdma: use correct print specifiers for size_t

This driver warns:

drivers/dma/mmp_tdma.c: In function 'mmp_tdma_prep_dma_cyclic':
drivers/dma/mmp_tdma.c:437:5: warning: format '%d' expects argument of type 'int', but argument 3 has type 'size_t' [-Wformat=]

We should use %zu to print 'size_t' values.

Acked-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: ep93xx: enable COMPILE_TEST

To get more coverage, enable COMPILE_TEST for this driver.

Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: ep93xx: don't use variables defined in global scope

The driver uses same variable name in local and global context in a
function, rename the local one for better readability.

Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: ep93xx: use correct print specifiers for size_t

This driver warns:

drivers/dma/ep93xx_dma.c: In function 'ep93xx_dma_prep_slave_sg':
drivers/dma/ep93xx_dma.c:1054:6: warning: format '%d' expects argument of type 'int', but argument 3 has type 'size_t' [-Wformat=]

drivers/dma/ep93xx_dma.c: In function 'ep93xx_dma_prep_dma_cyclic':
drivers/dma/ep93xx_dma.c:1129:5: warning: format '%d' expects argument of type 'int', but argument 3 has type 'size_t' [-Wformat=]

We should use %zu to print 'size_t' values.

Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: sa11x0: enable COMPILE_TEST

To get more coverage, enable COMPILE_TEST for this driver.

Acked-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: sa11x0: use correct print specifiers for size_t

This driver when compiled on 64 bits gave warnings:

drivers/dma/sa11x0-dma.c:602:2: warning: format ‘%u’ expects argument of type ‘unsigned int’, but argument 6 has type ‘size_t’ [-Wformat=]

We should use %zu to print 'size_t' values.

Acked-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: sa11x0: use correct print specifiers for u32

This driver when compiled on 64 bits gave warnings:

drivers/dma/sa11x0-dma.c:494:2: warning: format ‘%zx’ expects argument of type ‘size_t’, but argument 4 has type ‘u32’ [-Wformat=]

We should use %x to print 'u32' values.

Acked-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: sa11x0: use correct print specifiers for dma_addr_t

This driver when compiled on 64 bits gave warnings:

drivers/dma/sa11x0-dma.c:466:27: warning: format '%x' expects argument of type 'unsigned int', but argument 4 has type 'dma_addr_t' [-Wformat=]
drivers/dma/sa11x0-dma.c:554:31: warning: format '%x' expects argument of type 'unsigned int', but argument 5 has type 'dma_addr_t' [-Wformat=]
drivers/dma/sa11x0-dma.c:696:34: warning: format '%x' expects argument of type 'unsigned int', but argument 5 has type 'dma_addr_t' [-Wformat=]

We should use %pad to print 'dma_addr_t' values.

Acked-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: coh901318: enable COMPILE_TEST

To get more coverage, enable COMPILE_TEST for this driver.

Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: coh901318: use NULL for pointer initialization

Sparse complains:
drivers/dma/coh901318_lli.c:78:31: warning: Using plain integer as NULL pointer
drivers/dma/coh901318_lli.c:91:39: warning: Using plain integer as NULL pointer

Use NULL for pointer initialization rather than plain integer

Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: coh901318: fix pointer cast warnings

On some systems, pointer can be large than unsigned int, triggering warning
pointer-to-int-cast on conversion.

drivers/dma/coh901318.c: In function 'coh901318_filter_id':
drivers/dma/coh901318.c:1769:23: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]
  unsigned int ch_nr = (unsigned int) chan_id;

Also, converting an iomem pointer for print leads to warn on some system
which can be avoided with %p specfier

drivers/dma/coh901318.c: In function 'coh901318_probe':
drivers/dma/coh901318.c:2748:3: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]
   (u32) base->virtbase);

Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: coh901318: return ssize_t for coh901318_debugfs_read

The file_operations.read expects return as ssize_t, so update
coh901318_debugfs_read to return ssize_t to fix the warning:

drivers/dma/coh901318.c:1369:2: warning: initialization from incompatible pointer type [enabled by default]
  .read  = coh901318_debugfs_read,

Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: coh901318: use correct print specifiers

This driver when compiled on 64 bits gave warnings:

drivers/dma/coh901318.c: In function 'coh901318_list_print':
warning: format '%x' expects argument of type 'unsigned int', but argument 7 has type 'dma_addr_t' [-Wformat=]
warning: format '%x' expects argument of type 'unsigned int', but argument 8 has type 'dma_addr_t' [-Wformat=]
warning: format '%x' expects argument of type 'unsigned int', but argument 9 has type 'dma_addr_t' [-Wformat=]

drivers/dma/coh901318.c: In function 'coh901318_prep_memcpy':
warning: format '%x' expects argument of type 'unsigned int', but argument 6 has type 'dma_addr_t' [-Wformat=]
warning: format '%x' expects argument of type 'unsigned int', but argument 7 has type 'dma_addr_t' [-Wformat=]
warning: format '%d' expects argument of type 'int', but argument 8 has type 'size_t' [-Wformat=]

We should use %pad to print 'dma_addr_t' values and %zu to print size_t
values

Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: jz4780: make jz4780_dma_prep_dma_memcpy static

Sparse complains:
drivers/dma/dma-jz4780.c:399:32: warning: symbol
'jz4780_dma_prep_dma_memcpy' was not declared. Should it be static?

So make this static

Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: jz4780: enable COMPILE_TEST

To get more coverage, enable COMPILE_TEST for this driver.

Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: jz4740: enable COMPILE_TEST

To get more coverage, enable COMPILE_TEST for this driver.

Acked-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: jz4740: remove unused arch header

The defines in asm/mach-jz4740/dma.h are not used by driver so remove it

Acked-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: dmatest: Restore "memcpy" as default mode

Commit 0d4cb44da6ca0e8 ("dmaengine: dmatest: Add support for
scatter-gather DMA mode") changes default "dmatest" behavior by
changing default mode from "memcpy" to "scatter-gather".
Now "memcpy" gets back as default mode.

Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: cppi41: Ignore EINPROGRESS for PM runtime

We can occasionally get -EINPROGRESS for pm_runtime_get. In that case
we can just continue as we're queueing transfers anyways when
pm_runtime_active is not set.

Fixes: fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support")
Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: cleanup with list_first_entry_or_null()

The combo of list_empty() check and return list_first_entry()
can be replaced with list_first_entry_or_null().

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: NO_IRQ removal from powerpc-only drivers

We'd like to eventually remove NO_IRQ on powerpc, so remove usages of it
from powerpc-only drivers.

Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: dmatest: Apply copy_align to DMA_SG as well

The DMA_SG is still a type of memory copy operation that should
conform the hardware restriction. So this patch just applies the
copy_align to DMA_SG as well.

Signed-off-by: Nicolin Chen <nicoleotsuka@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: cppi41: mark PM functions as __maybe_unused

When CONFIG_PM_SLEEP is disabled, we get a build error in
the cppi41 dmaengine driver, since the runtime-pm functions
are hidden within the wrong #ifdef:

drivers/dma/cppi41.c:1158:21: error: 'cppi41_runtime_suspend' undeclared here (not in a function)

This removes the #ifdef and instead uses __maybe_unused
annotations that cannot have this problem.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support")
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: dmatest: exclude compare and fill time during perf report

Dmatest is currently including compare and fill time into the
calculated performance numbers. This does not reflect the HW
capability and the results vary based on the CPU speed instead of
the HW speed.

Signed-off-by: Sinan Kaya <okaya@codeaurora.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: qcom_hidma: add error reporting for tx_status

The HIDMA driver is capable of error detection. However, the error was
not being passed back to the client when tx_status API is called.

Changing the error handling behavior to follow this oder.

1. dmaengine asserts error interrupt
2. Driver receives and mark's the txn as error
3. Driver completes the txn and intimates the client. No further
   submissions. Drop the locks before calling callback, as subsequent
   processing by client maybe in callback thread.
4. Client invokes status and you can return error
5. On error, client calls terminate_all. You can reset channel, free all
   descriptors in the active, pending and completed lists
6. Client prepares new txn and so on.

As part of this work, got rid of the reset in the interrupt handler when
an error happens and the HW is put into disabled state. The only way to
recover is for the client to terminate the channel.

Signed-off-by: Sinan Kaya <okaya@codeaurora.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: qcom_hidma: report transfer errors with new interface

Pass the DMA errors to the client by passing a result argument. The HW only
supports a generic error when something goes wrong. That's why, using
DMA_TRANS_ABORTED all the time.

Signed-off-by: Sinan Kaya <okaya@codeaurora.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: qcom_hidma: release the descriptor before the callback

There is a race condition between data transfer callback and descriptor
free code. The callback routine may decide to clear the resources even
though the descriptor has not yet been freed.

Instead of calling the callback first and then releasing the memory,
this code is changing the order to return the descriptor back to the
free pool and then call the user provided callback.

Signed-off-by: Sinan Kaya <okaya@codeaurora.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: cppi41: Add basic PM runtime support

Let's keep the device enabled between cppi41_dma_issue_pending()
and dmaengine_desc_get_callback_invoke() and rely on the PM runtime
autoidle timeout elsewhere.

As the PM runtime is for whole device, not for each channel,
we need to queue pending transfers if the device is PM runtime
suspended. Then we start the pending transfers in PM runtime
resume.

Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: cppi41: Prepare to add PM runtime support

Let's just move code from cppi41_dma_issue_pending() to
push_desc_queue() as that's the only call to push_desc_queue().

We want to do this for PM runtime as we need to call push_desc_queue()
also for pending queued transfers from PM runtime resume.

No functional changes, just moves code around.

Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: device must have at least one channel

The DMA device can't be registered if it doesn't have any channels
registered at all. Moreover, it leads to memory leak and is reported by
kmemleak as (on 3.10 kernel, and same shall happen on mainline):

unreferenced object 0xffffffc09e597240 (size 64):
  comm "swapper/0", pid 1, jiffies 4294877736 (age 7060.280s)
  hex dump (first 32 bytes):
    00 00 00 00 c0 ff ff ff 30 00 00 ff 00 00 00 ff  ........0.......
    00 00 00 ff 00 00 00 ff 00 00 00 ff 00 00 00 ff  ................
  backtrace:
    [<ffffffc0003079ec>] create_object+0x148/0x2a0
    [<ffffffc000cc150c>] kmemleak_alloc+0x80/0xbc
    [<ffffffc000303a7c>] kmem_cache_alloc_trace+0x120/0x1ac
    [<ffffffc00054771c>] dma_async_device_register+0x160/0x46c
    [<ffffffc000548958>] foo_probe+0x1a0/0x264
    [<ffffffc0005d6658>] platform_drv_probe+0x14/0x20
    [<ffffffc0005d50cc>] driver_probe_device+0x160/0x374
    [<ffffffc0005d538c>] __driver_attach+0x60/0x90
    [<ffffffc0005d3e78>] bus_for_each_dev+0x7c/0xb0
    [<ffffffc0005d4a0c>] driver_attach+0x1c/0x28
    [<ffffffc0005d459c>] bus_add_driver+0x124/0x248
    [<ffffffc0005d59cc>] driver_register+0x90/0x110
    [<ffffffc0005d6bf4>] platform_driver_register+0x58/0x64
    [<ffffffc00142a70c>] foo_driver_init+0x10/0x1c
    [<ffffffc000200878>] do_one_initcall+0xac/0x148
    [<ffffffc00140096c>] kernel_init_freeable+0x1a0/0x258

Return -ENODEV from dma_async_device_register() on such a case.

Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: sh_shdma-base: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmengine: xilinx_dma: convert callback to helper function

Move the xilinx driver to new dmaengine callback

Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: documentation to the new callback mechanism

Adding documentation in dmaengine/provider.txt on the usage of callback
with result, callback_result, function pointer to retrieve transanction
result from a DMA submission.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

ntb: add DMA error handling for RX DMA

Adding support on the rx DMA path to allow recovery of errors when
DMA responds with error status and abort all the subsequent ops.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Acked-by: Allen Hubbe <Allen.Hubbe@emc.com>
Cc: Jon Mason <jdmason@kudzu.us>
Cc: linux-ntb@googlegroups.com
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

ntb: add DMA error handling for TX DMA

Adding support on the tx DMA path to allow recovery of errors when
DMA responds with error status and abort all the subsequent ops.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Acked-by: Allen Hubbe <Allen.Hubbe@emc.com>
Cc: Jon Mason <jdmason@kudzu.us>
Cc: linux-ntb@googlegroups.com
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: ioatdma: add error strings to chanerr output

Provide a mechanism to translate CHANERR bits to English strings in order
to allow user to report more concise errors.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: ioatdma: Add error handling to ioat driver

Adding error handling to the ioatdma driver so that when a
read/write error occurs the error results are reported back and
all the remaining descriptors are aborted. This utilizes the new
dmaengine callback function that allows reporting of results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: add support to provide error result from a DMA transation

Adding a new callback that will provide the error result for a transaction.
The result is allocated on the stack and the callback should create a copy
if it wishes to retain the information after exiting. The result parameter
is now defined and takes over the dummy void pointer we placed in the
helper functions previously. dmaengine drivers should start converting
to the new "callback_result" callback in order to receive transaction
results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: xgene-dma: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: virt-dma: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: txx9dmac: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: timb_dma: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: tegra20-apb-dma: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Cc: Laxman Dewangan <ldewangan@nvidia.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: ste_dma40: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Cc: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: sirf-dma: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: sh_rcar-dmac: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: qcom_hidma: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Acked-by: Sinan Kaya <okaya@codeaurora.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: ppc4xx_adma: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: pl330: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: pch_dma: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: nbpfaxi: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: mxs-dma: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: mv_xor: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: mpc512x_dma: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: mmp_tdma: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: mmp_pdma: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: mic_x100_dma: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Cc: Sudeep Dutt <sudeep.dutt@intel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: ipu: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: iop-adma: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: ioatdma: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: imx-sdma: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: imx-dma: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: fsldma: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Cc: Li Yang <leoli@freescale.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: fsl_raid: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>

dmaengine: ep93xx_dma: convert callback to helper function

This is in preperation of moving to a callback that provides results to the
callback for the transaction. The conversion will maintain current behavior
and the driver must convert to new callback mechanism at a later time in
order to receive results.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>