This time round the update brings in following changes:
- New tegra driver for ADMA device
- Support for Xilinx AXI Direct Memory Access Engine and Xilinx AXI Central
Direct Memory Access Engine and few updates to this driver.
- New cyclic capability to sun6i and few updates.
- Slave-sg support in bcm2835.
- Updates to many drivers like designware, hsu, mv_xor, pxa, edma,
qcom_hidma & bam.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=0Sy7
-----END PGP SIGNATURE-----
Merge tag 'dmaengine-4.7-rc1' of git://git.infradead.org/users/vkoul/slave-dma
Pull dmaengine updates from Vinod Koul:
"This time round the update brings in following changes:
- new tegra driver for ADMA device
- support for Xilinx AXI Direct Memory Access Engine and Xilinx AXI
Central Direct Memory Access Engine and few updates to this driver
- new cyclic capability to sun6i and few updates
- slave-sg support in bcm2835
- updates to many drivers like designware, hsu, mv_xor, pxa, edma,
qcom_hidma & bam"
* tag 'dmaengine-4.7-rc1' of git://git.infradead.org/users/vkoul/slave-dma: (84 commits)
dmaengine: ioatdma: disable relaxed ordering for ioatdma
dmaengine: of_dma: approximate an average distribution
dmaengine: core: Use IS_ENABLED() instead of checking for built-in or module
dmaengine: edma: Re-evaluate errors when ccerr is triggered w/o error event
dmaengine: qcom_hidma: add support for object hierarchy
dmaengine: qcom_hidma: add debugfs hooks
dmaengine: qcom_hidma: implement lower level hardware interface
dmaengine: vdma: Add clock support
Documentation: DT: vdma: Add clock support for dmas
dmaengine: vdma: Add config structure to differentiate dmas
MAINTAINERS: Update Tegra DMA maintainers
dmaengine: tegra-adma: Add support for Tegra210 ADMA
Documentation: DT: Add binding documentation for NVIDIA ADMA
dmaengine: vdma: Add Support for Xilinx AXI Central Direct Memory Access Engine
Documentation: DT: vdma: update binding doc for AXI CDMA
dmaengine: vdma: Add Support for Xilinx AXI Direct Memory Access Engine
Documentation: DT: vdma: update binding doc for AXI DMA
dmaengine: vdma: Rename xilinx_vdma_ prefix to xilinx_dma
dmaengine: slave means at least one of DMA_SLAVE, DMA_CYCLIC
dmaengine: mv_xor: Allow selecting mv_xor for mvebu only compatible SoC
...
Certain Intel Sunrisepoint PCH variants report zero chip selects in SPI
capabilities register even they have one per port. Detection in
pxa2xx_spi_probe() sets master->num_chipselect to 0 leading to -EINVAL
from spi_register_master() where chip select count is validated.
Fix this by not using SPI capabilities register on Sunrisepoint. They don't
have more than one chip select so use the default value 1 instead of
detection.
Fixes: 8b136baa58 ("spi: pxa2xx: Detect number of enabled Intel LPSS SPI chip select signals")
Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Cc: stable@vger.kernel.org
The source and destination masters are reflecting buses or their layers to
where the different devices can be connected. The patch changes the master
names to reflect which one is related to which independently on the transfer
direction.
The outcome of the change is that the memory data width is now always limited
by a data width of the master which is dedicated to communicate to memory.
The patch will not break anything since all current users have the same data
width for all masters. Though it would be nice to revisit avr32 platforms to
check what is the actual hardware topology in use there. It seems that it has
one bus and two masters on it as stated by Table 8-2, that's why everything
works independently on the master in use. The purpose of the sequential patch
is to fix the driver for configuration of more than one bus.
The change is done in the assumption that src_master and dst_master are
reflecting a connection to the memory and peripheral correspondently on avr32
and otherwise on the rest.
Acked-by: Hans-Christian Egtvedt <egtvedt@samfundet.no>
Acked-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
We clamp frame_len_words to a maximum of 4096, but do not actually
limit the number of words written or read through the DATA registers
or the length added to spi_message::actual_length. This results in
silent data corruption for commands longer than this maximum.
Recalculate the length of each transfer, taking frame_len_words into
account. Use this length in qspi_{read,write}_msg(), and to increment
spi_message::actual_length.
Signed-off-by: Ben Hutchings <ben.hutchings@codethink.co.uk>
Signed-off-by: Mark Brown <broonie@kernel.org>
Cc: stable@vger.kernel.org
Each transfer can specify 8, 16 or 32 bits per word independently of
the default for the device being addressed. However, currently we
calculate the number of words in the frame assuming that the word size
is the device default.
If multiple transfers in the same message have differing
bits_per_word, we bitwise-or the different values in the WLEN register
field.
Fix both of these. Also rename 'frame_length' to 'frame_len_words' to
make clear that it's not a byte count like spi_message::frame_length.
Signed-off-by: Ben Hutchings <ben.hutchings@codethink.co.uk>
Signed-off-by: Mark Brown <broonie@kernel.org>
Cc: stable@vger.kernel.org
This reverts commit 3525e0aac9.
The DMA transfer for RX buffer was not handled correctly in this change.
The actual transfer length for DMA RX can be less than xfer->len in the
specific condition and the last words will be filled after the DMA
completion, but the commit doesn't consider it and the dmaengine is
started with rx_sg mapped by spi core.
The solution for this at least requires more lines than this commit
has inserted. So revert it for now.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
There are use cases when chip select should be triggered between transfers
in single SPI message. Current implementation does this only on last
transfer in message ignoring cs_change value provided in current transfer.
Signed-off-by: Andrey Vostrikov <andrey.vostrikov@cogentembedded.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Use dma_request_chan instead of dma_request_slave_channel,
in this case we can check EPROBE_DEFER without static
warning.
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Mark Brown <broonie@kernel.org>
Currently omap2-mcspi cannot handle dma transfer for vmalloced buffer.
I hit this problem when using mtdblock on spi-nor.
This lets the SPI core handle the page mapping for dma transfer buffer.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
External users may use spi_bus_lock to get exclusive access. This will
also grab the bus_lock_mutex and may therefore result in a deadlock if
__spi_pump_messages also tries to get the mutex.
Therefore adapt spi_pump_messages as well as spi_sync to preset the
bus_locked parameter according to the master->bus_lock_flag.
Fixes: 49023d2e4e ("spi: core: Fix deadlock when sending messages")
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Signed-off-by: Mark Brown <broonie@kernel.org>
This is smallish update with minor changes to core and new driver and usual
updates. Nothing super exciting here..
- We have made slave address as physical to enable driver to do the mapping.
- We now expose the maxburst for slave dma as new capability so clients can
know this and program accordingly
- addition of device synchronize callbacks on omap and edma.
- pl330 updates to support DMAFLUSHP for Rockchip platforms.
- Updates and improved sg handling in Xilinx VDMA driver.
- New hidma qualcomm dma driver, though some bits are still in progress
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=wCCJ
-----END PGP SIGNATURE-----
Merge tag 'dmaengine-4.6-rc1' of git://git.infradead.org/users/vkoul/slave-dma
Pull dmaengine updates from Vinod Koul:
"This is smallish update with minor changes to core and new driver and
usual updates. Nothing super exciting here..
- We have made slave address as physical to enable driver to do the
mapping.
- We now expose the maxburst for slave dma as new capability so
clients can know this and program accordingly
- addition of device synchronize callbacks on omap and edma.
- pl330 updates to support DMAFLUSHP for Rockchip platforms.
- Updates and improved sg handling in Xilinx VDMA driver.
- New hidma qualcomm dma driver, though some bits are still in
progress"
* tag 'dmaengine-4.6-rc1' of git://git.infradead.org/users/vkoul/slave-dma: (40 commits)
dmaengine: IOATDMA: revise channel reset workaround on CB3.3 platforms
dmaengine: add Qualcomm Technologies HIDMA channel driver
dmaengine: add Qualcomm Technologies HIDMA management driver
dmaengine: hidma: Add Device Tree binding
dmaengine: qcom_bam_dma: move to qcom directory
dmaengine: tegra: Move of_device_id table near to its user
dmaengine: xilinx_vdma: Remove unnecessary variable initializations
dmaengine: sirf: use __maybe_unused to hide pm functions
dmaengine: rcar-dmac: clear pertinence number of channels
dmaengine: sh: shdmac: don't open code of_device_get_match_data()
dmaengine: tegra: don't open code of_device_get_match_data()
dmaengine: qcom_bam_dma: Make driver work for BE
dmaengine: sun4i: support module autoloading
dma/mic_x100_dma: IS_ERR() vs PTR_ERR() typo
dmaengine: xilinx_vdma: Use readl_poll_timeout instead of do while loop's
dmaengine: xilinx_vdma: Simplify spin lock handling
dmaengine: xilinx_vdma: Fix issues with non-parking mode
dmaengine: xilinx_vdma: Improve SG engine handling
dmaengine: pl330: fix to support the burst mode
dmaengine: make slave address physical
...
transfer could be NULL in spi_imx_can_dma() when it's called from
spi_imx_setupxfer() with a NULL transfer. Test for a NULL pointer before
dereferencing it.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
If the SPI chip select (CS) for a dedicated channel is done manually by
the used higher device driver, the CS may be active while writing to
ECSPIx_CONFIGREG. To prevent unwanted clock edges when selecting
the clock mode, only do the necessary changes to the i.MX SPI
configuration register and leave not selected channels untouched.
To prevent unwanted clock edges on first use, an empty dummy
transmission shall be done by the initialization procedure of the device
driver of this channel. This will set the clock mode to the correct state.
Signed-off-by: Knut Wohlrab <knut.wohlrab@de.bosch.com>
Signed-off-by: Dirk Behme <dirk.behme@de.bosch.com>
Acked-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Mark Brown <broonie@kernel.org>
The spi_split_transfers_maxsize() gfp parameter is missing in the
function kernel-doc so building gives the following warning:
.//drivers/spi/spi.c:2359: warning: No description found for parameter 'gfp'
Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
rsd_nsecs is defined as u8 memeber of struct rockchip_spi,
but using of_property_read_u32. That means we take risk of
truncation by type conversion if we pass on big value from
dt.
Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Remove some of unused header files and reoder
it into alphabetical order.
Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
The device which is actually does DMA may have a limitation of the maximum
segment size. Respect this setting when preparing scatter-gather list.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Let's defer probing the driver if the return value of
dma_request_slave_channel is ERR_PTR(-EPROBE_DEFER) instead
of disabling dma capability directly.
Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
dmaengine_terminate_all is deprecated, let's use
dmaengine_terminate_async for interrupt handling.
Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
We should check return value of dmaengine_prep_slave_sg, otherwise
we take risk of null pointer.
Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
The function __spi_pump_messages() is called by spi_pump_messages() and
__spi_sync(). The function __spi_sync() has an argument 'bus_locked'
that indicates if it is called with the SPI bus mutex held or not. If
'bus_locked' is false then __spi_sync() will acquire the mutex itself.
Commit 556351f14e ("spi: introduce accelerated read support for spi
flash devices") made a change to acquire the SPI bus mutex within
__spi_pump_messages(). However, this change did not check to see if the
mutex is already held. If __spi_sync() is called with the mutex held
(ie. 'bus_locked' is true), then a deadlock occurs when
__spi_pump_messages() is called.
Fix this deadlock by passing the 'bus_locked' state from __spi_sync() to
__spi_pump_messages() and only acquire the mutex if not already held. In
the case where __spi_pump_messages() is called from spi_pump_messages()
it is assumed that the mutex is not held and so call
__spi_pump_messages() with 'bus_locked' set to false. Finally, move the
unlocking of the mutex to the end of the __spi_pump_messages() function
to simplify the code and only call cond_resched() if there are no
errors.
Fixes: 556351f14e ("spi: introduce accelerated read support for spi flash devices")
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
16-bit transfers must be in big endian mode on wire.
Signed-off-by: Alexander Kochetkov <al.kochet@gmail.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Rockchip_spi_set_cs could be called by spi_setup, but
spi_setup may be called by device driver after runtime suspend.
Then the spi clock is closed, rockchip_spi_set_cs may access the
spi registers, which causes cpu block in some socs.
Fixes: 64e36824b3 ("spi/rockchip: add driver for Rockchip RK3xxx")
Signed-off-by: Huibin Hong <huibin.hong@rock-chips.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Commit b36f09c3c4 ("dmaengine: Add transfer termination
synchronization support") marked dmaengine_terminate_all() as
deprecated and is being replaced by explicit synchronous and asynchronous
terminate functions.
Here DMA termination are done in two cases: FIFO overrun and module
removal.
FIFO overrun is handled in interrupt context and converting
dmaengine_terminate_all() to dmaengine_terminate_async() does the same than
before.
Using synchronous termination in module removal however adds a bit more
robustness as it waits all completion callbacks have finished. Although it
looks all known DMA engines used with spi-pxa2xx don't implement
device_synchronize() callback so this too appears to be a no-op in
practice.
Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Acked-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: Mark Brown <broonie@kernel.org>
Add support for third Intel Broxton variant and update comment for
A-Step variant.
Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Tested-by: Steve Sakoman <steve.sakoman@intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
The spi_lp8841_rtc_probe() function misses an initialization of the
return code when it fails to get its memory resource, as gcc notices:
drivers/spi/spi-lp8841-rtc.c: In function 'spi_lp8841_rtc_probe':
drivers/spi/spi-lp8841-rtc.c:239:9: error: 'ret' may be used uninitialized in this function [-Werror=maybe-uninitialized]
This changes the code to propagate the error from devm_ioremap_resource().
Fixes: 7ecbfff671 ("spi: master driver to enable RTC on ICPDAS LP-8841")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Tested-by: Sergei Ianovich <ynvich@gmail.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
The driver tries to be clever by only setting up DMA channels when
the corresponding sg tables are non NULL. The sg tables are embedded
structs in struct spi_transfer, so they are guaranteed to be non NULL
which makes the if(tx)/if(rx) tests completely bogus. The driver even
sets the SPI_MASTER_MUST_RX / SPI_MASTER_MUST_TX flags which makes sure
the sg tables are not only present but also non empty.
Drop the tests and make the DMA path easier to follow.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Mark Brown <broonie@kernel.org>
Now that the config function knows whether we are doing DMA or not we
can do the necessary register setup in the config function and no longer
have to do this in the trigger function. With this the trigger function
becomes a no-op for DMA, so instead of testing if we are doing DMA or
not in the trigger function we simply no longer call it in the DMA case.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Mark Brown <broonie@kernel.org>
The watermark levels in the DMA register are write only, the driver
should never have to read them back from the hardware. Replace the
current _MASK and _OFFSET defines with defines taking the watermark
level directly.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Mark Brown <broonie@kernel.org>
This reverts patch 1476253cef (spi: imx: fix ecspi mode setup)
The patch tried to fix something by clearing bits in the cfg variable,
but cfg is initialized to zero on function entry. There are no bits to
clear.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Mark Brown <broonie@kernel.org>
DMA transfer for SPI was limited to up to 8 bits word size until now.
Sync in SPI burst size and DMA bus width is necessary to correctly
support 16 and 32 BPW.
Signed-off-by: Anton Bondarenko <anton.bondarenko.sama@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Mark Brown <broonie@kernel.org>
There's no need for an extra dma_is_inited variable when we can
equally well check for the existence of a DMA channel.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Mark Brown <broonie@kernel.org>
When the MX51_ECSPI_DMA is configured we control every single bit
of the register, so there's no need to read/modify/write it. Instead
just write the value we want to have in the register. Also, drop
unnecessary check if we are actually doing DMA. The values written
to the register have no effect in PIO mode and value written there
during the last DMA transfer is still in the register, so we can
equally well always write a value.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Mark Brown <broonie@kernel.org>
The SoC specific config function does not know if DMA will be used or
not. This information will be useful to configure the SPI controller
correctly for DMA in following patches, so initialize the usedma
variable before calling into the SoC specific config function.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Mark Brown <broonie@kernel.org>
At the moment the driver decides to fallback to PIO mode the buffers
are already mapped for DMA. It's a bug to access them with the CPU
afterwards, so we cannot just fallback to PIO mode.
It should not be necessary anyway, since we only use DMA when we
verified that it's possible in the fist place, so when prep_slave_sg
fails it's a bug, either in the SDMA driver or in the can_dma
implementation.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Mark Brown <broonie@kernel.org>
ICP DAS LP-8841 contains a DS-1302 RTC. This driver provides an SPI
master which makes the RTC usable. The driver is not supposed to work
with anything else.
The driver uses the standard MicroWire half-duplex transfer timing.
Master output is set on low clock and sensed by the RTC on the rising
edge. Master input is set by the RTC on the trailing edge and is sensed
by the master on low clock.
Signed-off-by: Sergei Ianovich <ynvich@gmail.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Mark Brown <broonie@kernel.org>
Linus Torvalds