This fixes tx and bi-directional dma transfers on rk3399-gru-kevin.

It seems the SPI fifo must have room for 2 bursts when the dma_tx_req
signal is generated or it might skip some words. This in turn makes
the rx dma channel never complete for bi-directional transfers.

Fix it by setting tx burst length to fifo_len / 4 and the dma
watermark to fifo_len / 2.

However the rk3399 TRM says (sic):
"DMAC support incrementing-address burst and fixed-address burst. But in
the case of access SPI and UART at byte or halfword size, DMAC only
support fixed-address burst and the address must be aligned to word."

So this relies on fifo_len being a multiple of 16 such that the
burst length (= fifo_len / 4) is a multiple of 4 and the addresses
will be word-aligned.

Fixes: dcfc861d ("spi: rockchip: adjust dma watermark and burstlen")
Signed-off-by: Emil Renner Berthing <kernel@esmil.dk>
Signed-off-by: Mark Brown <broonie@kernel.org>

Support setting a delay between cs assert and deassert as
a multiple of spi clock length.
Signed-off-by: Martin Sperl <kernel@martin.sperl.org>
Signed-off-by: Mark Brown <broonie@kernel.org>

For some SPI devices that support speed_hz > 1MHz the default 10 us delay
when cs_change = 1 is typically way to long and may result in poor spi bus
utilization.

This patch makes it possible to control the delay at micro or nano second
resolution on a per spi_transfer basis. It even allows an "as fast as
possible" mode with:
    xfer.cs_change_delay_unit = SPI_DELAY_UNIT_NSECS;
    xfer.cs_change_delay = 0;

The delay code is shared between delay_usecs and cs_change_delay for
consistency and reuse, so in the future this change_delay_unit could also
apply to delay_usec as well.

Note that on slower SOCs/CPU actually reaching ns deasserts on cs is not
realistic as the gpio overhead alone (without any delays added ) may
already leave cs deasserted for more than 1us - at least on a raspberry pi.
But at the very least this way we can keep it as short as possible.
Signed-off-by: Martin Sperl <kernel@martin.sperl.org>
Signed-off-by: Mark Brown <broonie@kernel.org>

To estimate efficiency add statistics on transfer types
(polling, interrupt and dma) used to debugfs.
Signed-off-by: Martin Sperl <kernel@martin.sperl.org>

Changelog:
  V1 -> V2: applied feedback by Stefan Wahren
            reorganized patchset
	    added extra rational, descriptions
	    fixed compile issue when CONFIG_DEBUG_FS is unset
Signed-off-by: Mark Brown <broonie@kernel.org>

Under some circumstances the default 30 us polling limit is not optimal
and may lead to long delays because we are waiting on an interrupt.
with this patch we have the possibility to influence this policy.

So make this limit (in us) configurable via a module parameters
(but also modifyable via /sys/modules/...)

This replicates similar code found in spi-bcm2835aux.
Signed-off-by: Martin Sperl <kernel@martin.sperl.org>

Changelog:
  V1 -> V2: applied feedback by Stefan Wahren
            reorganized patchset
	    added extra rational, descriptions
Signed-off-by: Mark Brown <broonie@kernel.org>

The DMA mode behaves slightly different than polling or interrupt driven
mode, so just document the fact
Signed-off-by: Martin Sperl <kernel@martin.sperl.org>

Changelog:
  V1 -> V2: applied feedback by Stefan Wahren
            new in V2
Signed-off-by: Mark Brown <broonie@kernel.org>

Avoid 64 bit aritmetics when deciding if we need to use polling or not
This replicates: commit d704afff
("spi: bcm2835aux: Avoid 64-bit arithmetic in xfer len calc")
from spi-bcm2835aux
Signed-off-by: Martin Sperl <kernel@martin.sperl.org>

Changelog:
  V1 -> V2: applied feedback by Stefan Wahren
            reorganized patchset
	    added extra rational, descriptions
Signed-off-by: Mark Brown <broonie@kernel.org>

Remove the unnecessary argument of xfer_time_us when calling
bcm2835_spi_transfer_one_poll.
Signed-off-by: Martin Sperl <kernel@martin.sperl.org>

Changelog:
  V1 -> V2: applied feedback by Stefan Wahren
            reorganized patchset
	    added extra rational, descriptions
Signed-off-by: Mark Brown <broonie@kernel.org>

This patch adds support for DMA. Transfers are done with dma only if
they are longer than 16 bytes in order to achieve a better performance.
DMA setup introduces a little overhead and for transfers shorter than 16
bytes there is no performance improvement.
Signed-off-by: Radu Pirea <radu_nicolae.pirea@upb.ro>
Signed-off-by: Mark Brown <broonie@kernel.org>

The bindings for DMA are now common for both drivers of the USART
IP.

The node given as an example for USART in SPI mode has been updated in
order to include DMA bindings.
Signed-off-by: Radu Pirea <radu_nicolae.pirea@upb.ro>
Signed-off-by: Mark Brown <broonie@kernel.org>

My previous patch leaves a dangling variable in the driver.
get rid of it.

Fixes: 06a391b1621e ("spi: ep93xx: Convert to use CS GPIO descriptors")
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Mark Brown <broonie@kernel.org>

Process handling QSPI when transmit/receive at qspi_trigger_transfer_out_in() as follows:
Setting the trigger, is the number of bytes in the FIFO buffer to determine
when there is an interrupt. Then check if the value of triggering number is
32-bytes or 1-byte, there will be corresponding processing
Handling (if (n == QSPI_BUFFER_SIZE) esle) this is unnecessary, leads to the
same processing of data transmission or reception, The difference here are with
ret = rspi_wait_for_tx_empty(rspi);
ret = rspi_wait_for_rx_full(rspi);

When the nummber trigger is 32 bytes, we only write into FIFO when the FIFO is completely empty
(interrupt transmission), and only receive if FIFO is full of 32 bytes of data.

In the case of a nummber trigger that is 1 byte, in principle we still need to process
rspi_wait_for_tx_empty/full so that FIFO is empty only with the amount of data we need to write to
or equal to the number of bytes we need to receive, There is currently no processing of this.
And in the current case with this patch, at this time it only needs at least 1 byte received in
FIFO that has interrupt received, or FIFO at least 1bytes free can be written into FIFO,
This patch therefore does not affect this processing.
So we need to eliminate unnecessary waste processing (if (n == QSPI_BUFFER_SIZE) esle),
more precisely in waiting for FIFO status.
The same with handling in qspi_transfer_out()/qspi_transfer_in().
Signed-off-by: Hoan Nguyen An <na-hoan@jinso.co.jp>
Signed-off-by: Mark Brown <broonie@kernel.org>

atmel_qspi objects are kept in spi_controller objects, so, first get
pointer to spi_controller object and then get atmel_qspi object from
spi_controller object.

Fixes: 2d30ac5e ("mtd: spi-nor: atmel-quadspi: Use spi-mem interface for atmel-quadspi driver")
Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Reviewed-by: Tudor Ambarus <tudor.ambarus@microchip.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

During probe, return the "get_irq" error value instead of -ENOENT. This
allows the driver to be defer probed if needed.
Signed-off-by: Fabien Dessenne <fabien.dessenne@st.com>
Acked-by: Amelie Delaunay <amelie.delaunay@st.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

This patch fixes PIO mode transfer to use PIO bit in SPI_COMMAND1 register.
Current driver uses DMA_EN instead of PIO bit.
Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Calculate the divisor for the SCR (Serial Clock Rate), avoiding
that the SSP transmission rate can be greater than the device rate.

When the division between the SSP clock and the device rate generates
a reminder, we have to increment by one the divisor.
In this way the resulting SSP clock will never be greater than the
device SPI max frequency.

For example, with:

 - ssp_clk  = 50 MHz
 - dev freq = 15 MHz

without this patch the SSP clock will be greater than 15 MHz:

 - 25 MHz for PXA25x_SSP and CE4100_SSP
 - 16,56 MHz for the others

Instead, with this patch, we have in both case an SSP clock of 12.5MHz,
so the max rate of the SPI device clock is respected.
Signed-off-by: Flavio Suligoi <f.suligoi@asem.it>
Reviewed-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Reviewed-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

When GPIO chip-select is used nothing prevents any available SPI
controllers to work with both CS-high and traditional CS-low modes.
In fact the SPI bus core code already does it, so we don't need to
introduce any modification there. But spi_setup() still fails to
switch the interface settings if CS-high flag is set for the case
of GPIO-driven slave chip-select when the SPI controller doesn't
support the hardwired CS-inversion. Lets fix it by clearing the
SPI_CS_HIGH flag out from bad_bits (unsupported by controller) when
client chip is selected by GPIO.

This feature is useful for slave devices, which in accordance with
communication protocol can work with both active-high and active-low
chip-selects. I am aware of one such device. It is MMC-SPI interface,
when at init sequence the driver needs to perform a read operation with
low and high chip-select sequentially (requirement of 74 clock cycles
with both chipselect, see the mmc_spi driver for details).
Signed-off-by: Serge Semin <fancer.lancer@gmail.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

This converts the EP93xx SPI master driver to use GPIO
descriptors for chip select handling.

EP93xx was using platform data to pass in GPIO lines,
by converting all board files to use GPIO descriptor
tables the core will look up the GPIO lines from the
SPI device in the same manner as for device tree.
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Mark Brown <broonie@kernel.org>

Add missing <of_device_id> table for SPI driver relying on SPI
device match since compatible is in a DT binding or in a DTS.

Before this patch:
modinfo sound/soc/codecs/snd-soc-adau1977-spi.ko | grep alias
alias:          spi:adau1979
alias:          spi:adau1978
alias:          spi:adau1977

After this patch:
modinfo sound/soc/codecs/snd-soc-adau1977-spi.ko | grep alias
alias:          of:N*T*Cadi,adau1979C*
alias:          of:N*T*Cadi,adau1979
alias:          of:N*T*Cadi,adau1978C*
alias:          of:N*T*Cadi,adau1978
alias:          of:N*T*Cadi,adau1977C*
alias:          of:N*T*Cadi,adau1977
alias:          spi:adau1979
alias:          spi:adau1978
alias:          spi:adau1977
Reported-by: Javier Martinez Canillas <javier@dowhile0.org>
Signed-off-by: Daniel Gomez <dagmcr@gmail.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Alpha and some of the architectures are missing readsl/writesl functions.
so the zynq-qspi driver won't be able to build on these arches. hence use
ioread32_rep()/iowrite32_rep() instead of readsl()/writesl().
Signed-off-by: Naga Sureshkumar Relli <naga.sureshkumar.relli@xilinx.com>
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

During probe, check the "get_irq" error value.
Signed-off-by: Fabien Dessenne <fabien.dessenne@st.com>
Acked-by: Ludovic Barre <ludovic.barre@st.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Split spi transfers into chunks of <=65532 to enable the driver to
perform DMA transfer on big buffers. The DLEN register specifies the
number of bytes to transfer in DMA mode. It is 16-bit wide and thus the
maximum DMA transfer is 65535 bytes. Set the maximum to 65532 (4 byte
aligned) since the FIFO in DMA mode is accessed in 4 byte chunks.

->max_dma_len is the value the spi core uses when splitting the buffer
into scatter gather entries.
The BCM2835 DMA block has 2 types of channels/engines:
- Normal: Max length: 1G
- Lite: Max length: 65535

Even when using a Lite channel we will not exceed the max length, so
let's drop setting ->max_dma_len.
Signed-off-by: Meghana Madhyastha <meghana.madhyastha@gmail.com>
Signed-off-by: Noralf Trønnes <noralf@tronnes.org>
Reviewed-by: Lukas Wunner <lukas@wunner.de>
Signed-off-by: Mark Brown <broonie@kernel.org>

This patch changes mode and mode_bits from u16 to u32 to allow more
mode configurations.
Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Add binding documentation of spi-mt65xx for MT8516 SOC.
Signed-off-by: Leilk Liu <leilk.liu@mediatek.com>
Reviewed-by: Matthias Brugger <matthias.bgg@gmail.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Add PCI IDs for SPI on Comet Lake.
Signed-off-by: Evan Green <evgreen@chromium.org>
Acked-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Reviewed-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Large transfers (64kB) doesn't show up in the trace. Not sure why, but
since printk can only display buffers up to 64 bytes in length, we only
need to store the first 64 bytes.

Cc: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Noralf Trønnes <noralf@tronnes.org>
Signed-off-by: Mark Brown <broonie@kernel.org>

spi_split_transfers_maxsize() can be used to split a transfer. This
function uses spi_res to lifetime manage the added transfer structures.
So in order to finalize the current message while it contains the split
transfers, spi_res_release() must be called after finalizing.
Signed-off-by: Noralf Trønnes <noralf@tronnes.org>
Signed-off-by: Mark Brown <broonie@kernel.org>

Don't warn about splitting transfers, the info is available in the
statistics if needed.
Signed-off-by: Noralf Trønnes <noralf@tronnes.org>
Signed-off-by: Mark Brown <broonie@kernel.org>

Falling back to maximum speed of the controller in case of SPI slave
maximum speed is not set is needless. It already defaults to maximum
speed of the controller since commit 052eb2d4 ("spi: core: Set
max_speed_hz of spi_device default to max_speed_hz of controller").
Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Document SoC specific bindings for R-Car RZ/G1C(r8a77470) SoC.
Signed-off-by: Cao Van Dong <cv-dong@jinso.co.jp>
Signed-off-by: Mark Brown <broonie@kernel.org>

With dw_dmac, sometimes the request of a DMA channel fails because
the DMA driver is not ready, so an explicit dependency request
is necessary.
Signed-off-by: Flavio Suligoi <f.suligoi@asem.it>
Signed-off-by: Mark Brown <broonie@kernel.org>

Stub helper spi_mem_default_supports_op() should
be set to static inline
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

When building with CONFIG_SPI_MEM is not set
gc warns this:

drivers/spi/spi-zynq-qspi.o: In function `zynq_qspi_supports_op':
spi-zynq-qspi.c:(.text+0x1da): undefined reference to `spi_mem_default_supports_op'

Fixes: 67dca5e5 ("spi: spi-mem: Add support for Zynq QSPI controller")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Pull ARM fixes from Russell King:
 "A small number of ARM fixes

   - Fix function tracer and unwinder dependencies so that we don't end
     up building kernels that will crash

   - Fix ARMv7M nommu initialisation (missing register initialisation)

   - Fix EFI decompressor entry (ensuring barrier instructions are
     enabled prior to use)"

* tag 'for-linus' of git://git.armlinux.org.uk/~rmk/linux-arm:
  ARM: 8857/1: efi: enable CP15 DMB instructions before cleaning the cache
  ARM: 8856/1: NOMMU: Fix CCR register faulty initialization when MPU is disabled
  ARM: fix function graph tracer and unwinder dependencies

Pull powerpc fixes from Michael Ellerman:
 "A one-liner to make our Radix MMU support depend on HUGETLB_PAGE. We
  use some of the hugetlb inlines (eg. pud_huge()) when operating on the
  linear mapping and if they're compiled into empty wrappers we can
  corrupt memory.

  Then two fixes to our VFIO IOMMU code. The first is not a regression
  but fixes the locking to avoid a user-triggerable deadlock.

  The second does fix a regression since rc1, and depends on the first
  fix. It makes it possible to run guests with large amounts of memory
  again (~256GB).

  Thanks to Alexey Kardashevskiy"

* tag 'powerpc-5.1-6' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/mm_iommu: Allow pinning large regions
  powerpc/mm_iommu: Fix potential deadlock
  powerpc/mm/radix: Make Radix require HUGETLB_PAGE

Pull block fixes from Jens Axboe:
 "A set of io_uring fixes that should go into this release. In
  particular, this contains:

   - The mutex lock vs ctx ref count fix (me)

   - Removal of a dead variable (me)

   - Two race fixes (Stefan)

   - Ring head/tail condition fix for poll full SQ detection (Stefan)"

* tag 'for-linus-20190428' of git://git.kernel.dk/linux-block:
  io_uring: remove 'state' argument from io_{read,write} path
  io_uring: fix poll full SQ detection
  io_uring: fix race condition when sq threads goes sleeping
  io_uring: fix race condition reading SQ entries
  io_uring: fail io_uring_register(2) on a dying io_uring instance

Pull rdma fixes from Jason Gunthorpe:
 "One core bug fix and a few driver ones

   - FRWR memory registration for hfi1/qib didn't work with with some
     iovas causing a NFSoRDMA failure regression due to a fix in the NFS
     side

   - A command flow error in mlx5 allowed user space to send a corrupt
     command (and also smash the kernel stack we've since learned)

   - Fix a regression and some bugs with device hot unplug that was
     discovered while reviewing Andrea's patches

   - hns has a failure if the user asks for certain QP configurations"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
  RDMA/hns: Bugfix for mapping user db
  RDMA/ucontext: Fix regression with disassociate
  RDMA/mlx5: Use rdma_user_map_io for mapping BAR pages
  RDMA/mlx5: Do not allow the user to write to the clock page
  IB/mlx5: Fix scatter to CQE in DCT QP creation
  IB/rdmavt: Fix frwr memory registration