The burst length can be adjusted according to the transmission
length to improve the transmission rate
Signed-off-by: NJon Lin <jon.lin@rock-chips.com>
Tested-by: NEmil Renner Berthing <kernel@esmil.dk>
Reviewed-by: NEmil Renner Berthing <kernel@esmil.dk>
Link: https://lore.kernel.org/r/20200723004356.6390-1-jon.lin@rock-chips.comSigned-off-by: NMark Brown <broonie@kernel.org>

The original implementation set num_chipselect to ROCKCHIP_SPI_MAX_CS_NUM (2)
which seems wrong here. spi0 has 2 native cs, all others just one. With
enable and use of cs_gpiods / GPIO CS, its correct to set the num_chipselect
from the num-cs property and set max_native_cs with the define.
If num-cs is missing the default set to num_chipselect = 1.
Signed-off-by: NChris Ruehl <chris.ruehl@gtsys.com.hk>
Link: https://lore.kernel.org/r/20200511083022.23678-4-chris.ruehl@gtsys.com.hkSigned-off-by: NMark Brown <broonie@kernel.org>

Add support for spi slave mode in spi-rockchip. The register map has an entry
for it. If spi-slave is set in dts, set this corresponding bit and add to
mode_bits the SPI_NO_CS, allow slave mode without explicit CS use.
Slave abort function had been added.
Signed-off-by: NChris Ruehl <chris.ruehl@gtsys.com.hk>
Link: https://lore.kernel.org/r/20200511083022.23678-3-chris.ruehl@gtsys.com.hkSigned-off-by: NMark Brown <broonie@kernel.org>

Cleanup, move from the compatibily layer struct spi_master over
to struct spi_controller, and rename the related function calls.
Signed-off-by: NChris Ruehl <chris.ruehl@gtsys.com.hk>
Link: https://lore.kernel.org/r/20200511083022.23678-2-chris.ruehl@gtsys.com.hkSigned-off-by: NMark Brown <broonie@kernel.org>

The Rockchip spi binding is updated to yaml and new models
were added. The spi on px30,rk3308 and rk3328 are the same as
other Rockchip based SoCs, so add compatible string for it.
Signed-off-by: NJohan Jonker <jbx6244@gmail.com>
Link: https://lore.kernel.org/r/20200309151004.7780-1-jbx6244@gmail.comSigned-off-by: NMark Brown <broonie@kernel.org>

Based on 1 normalized pattern(s):

  this program is free software you can redistribute it and or modify
  it under the terms and conditions of the gnu general public license
  version 2 as published by the free software foundation this program
  is distributed in the hope it will be useful but without any
  warranty without even the implied warranty of merchantability or
  fitness for a particular purpose see the gnu general public license
  for more details

extracted by the scancode license scanner the SPDX license identifier

  GPL-2.0-only

has been chosen to replace the boilerplate/reference in 263 file(s).
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>
Reviewed-by: NAllison Randal <allison@lohutok.net>
Reviewed-by: NAlexios Zavras <alexios.zavras@intel.com>
Cc: linux-spdx@vger.kernel.org
Link: https://lkml.kernel.org/r/20190529141901.208660670@linutronix.deSigned-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

kthread.h can't be included in psi_types.h because it creates a circular
inclusion with kthread.h eventually including psi_types.h and
complaining on kthread structures not being defined because they are
defined further in the kthread.h.  Resolve this by removing psi_types.h
inclusion from the headers included from kthread.h.

Link: http://lkml.kernel.org/r/20190319235619.260832-7-surenb@google.comSigned-off-by: NSuren Baghdasaryan <surenb@google.com>
Acked-by: NJohannes Weiner <hannes@cmpxchg.org>
Cc: Dennis Zhou <dennis@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

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: NEmil Renner Berthing <kernel@esmil.dk>
Signed-off-by: NMark Brown <broonie@kernel.org>

Add missing support for lsb-first mode.
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Tested-by: NHeiko Stuebner <heiko@sntech.de>
Signed-off-by: NMark Brown <broonie@kernel.org>

The hardware supports 4, 8 and 16bit spi words,
so add the missing support for 4bit words.
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Tested-by: NHeiko Stuebner <heiko@sntech.de>
Signed-off-by: NMark Brown <broonie@kernel.org>

Register an interrupt handler to fill/empty the
tx and rx fifos rather than busy-looping.
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Tested-by: NHeiko Stuebner <heiko@sntech.de>
Signed-off-by: NMark Brown <broonie@kernel.org>

Now that we no longer potentially change spi clock
at runtime we can precompute the rx sample delay
at probe time rather than for each transfer.
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Tested-by: NHeiko Stuebner <heiko@sntech.de>
Signed-off-by: NMark Brown <broonie@kernel.org>

The driver previously checked each transfer if the
requested speed was higher than possible with the
current spi clock rate and raised the clock rate
accordingly.

However, there is no check to see if the spi clock
was actually set that high and no way to dynamically
lower the spi clock rate again.

So it seems any potiential users of this functionality
are better off just setting the spi clock rate at init
using the assigned-clock-rates devicetree property.

Removing this dynamic spi clock rate raising allows
us let the spi framework handle min/max speeds
for us.
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Tested-by: NHeiko Stuebner <heiko@sntech.de>
Signed-off-by: NMark Brown <broonie@kernel.org>

We only need to know if we're using dma when setting
up the transfer, so just use a local variable for
that.
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Tested-by: NHeiko Stuebner <heiko@sntech.de>
Signed-off-by: NMark Brown <broonie@kernel.org>

In almost all cases we already have a pointer to the
spi master structure where we have the driver data.

The only exceptions are the dma callbacks which are
easily fixed by passing them the master and using
spi_master_get_devdata to retrieve the driver data.
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Tested-by: NHeiko Stuebner <heiko@sntech.de>
Signed-off-by: NMark Brown <broonie@kernel.org>

The spi master (aka spi controller) structure already
has two fields for storing the rx and tx dma channels.
Just use them rather than duplicating them in driver data.
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Tested-by: NHeiko Stuebner <heiko@sntech.de>
Signed-off-by: NMark Brown <broonie@kernel.org>

Just read transfer info directly from the spi device
and transfer structures rather than storing it in
driver data first.
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Tested-by: NHeiko Stuebner <heiko@sntech.de>
Signed-off-by: NMark Brown <broonie@kernel.org>

Successful transfers leave the spi disabled, so if
we just make sure to disable the spi on error
there should be no need to disable the spi from
master->unprepare_message.

This also flushes the tx and rx fifos,
so no need to do that manually.
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Tested-by: NHeiko Stuebner <heiko@sntech.de>
Signed-off-by: NMark Brown <broonie@kernel.org>

The state field is currently only used to make sure
only the last of the tx and rx dma callbacks issue
an spi_finalize_current_transfer.
Rather than using a spinlock we can get away
with just turning the state field into an atomic_t.
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Tested-by: NHeiko Stuebner <heiko@sntech.de>
Signed-off-by: NMark Brown <broonie@kernel.org>

The hardware supports 3 different variants of SPI
and there were some code around it, but nothing
to actually set it to anything but "Motorola SPI".
Just drop that code and always use that mode.
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Tested-by: NHeiko Stuebner <heiko@sntech.de>
Signed-off-by: NMark Brown <broonie@kernel.org>

Use C99 designated initializers for dma slave config
structures. This also makes sure uninitialized fields
are zeroed so we don't need an explicit memset.
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Tested-by: NHeiko Stuebner <heiko@sntech.de>
Signed-off-by: NMark Brown <broonie@kernel.org>

The spi_enable_chip function takes a boolean
argument. Change the type to reflect that.
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Tested-by: NHeiko Stuebner <heiko@sntech.de>
Signed-off-by: NMark Brown <broonie@kernel.org>

Let the dma/non-dma code paths handle the spi enable
flag themselves. This removes some logic to determine
if the flag should be turned on before or after dma
and also don't leave the spi enabled if the dma path
fails.
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Signed-off-by: NMark Brown <broonie@kernel.org>

The dma direction for the tx and rx dma channels never
change, so just use the constants directly rather
than storing them in device data.
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Signed-off-by: NMark Brown <broonie@kernel.org>

The driver data has a u32 field use_dma which is
only ever used as a boolean, so change its type
to reflect that.
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Signed-off-by: NMark Brown <broonie@kernel.org>

We no longer need the dma_caps since the dma driver
already clamps the burst length to the hardware limit,
so don't request and store dma_caps in device data.
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Signed-off-by: NMark Brown <broonie@kernel.org>

Signal tx dma when spi fifo is less than half full,
and limit tx bursts to half the fifo length.

Clamp rx burst length to 1 to avoid alignment issues.
Signed-off-by: NHuibin Hong <huibin.hong@rock-chips.com>
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Signed-off-by: NMark Brown <broonie@kernel.org>

The rxconf and txconf structs are allocated on the
stack, so make sure we zero them before filling out
the relevant fields.
Signed-off-by: NHuibin Hong <huibin.hong@rock-chips.com>
Signed-off-by: NEmil Renner Berthing <kernel@esmil.dk>
Signed-off-by: NMark Brown <broonie@kernel.org>

CTRLR1 is number of data frames, when rx only.
When data frame is 8 bit, CTRLR1 is len-1.
When data frame is 16 bit, CTRLR1 is (len/2)-1.
Signed-off-by: NHuibin Hong <huibin.hong@rock-chips.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

The spi on rv1108 is the same as other rockchip based
socs, add compatible string for it.
Signed-off-by: NAndy Yan <andy.yan@rock-chips.com>
Acked-by: NRob Herring <robh@kernel.org>
Signed-off-by: NMark Brown <broonie@kernel.org>

The runtime suspend callback might be called by pm domain framework at
suspend_noirq stage. It would try to disable the clocks which already
been disabled by rockchip_spi_suspend.

Call pm_runtime_force_suspend/pm_runtime_force_resume when
suspend/resume to avoid that.
Signed-off-by: NJeffy Chen <jeffy.chen@rock-chips.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

We are assuming clocks enabled when calling rockchip_spi_remove, which
is not always true. Those clocks might already been disabled by the
runtime PM at that time.

Call pm_runtime_get_sync before trying to disable clocks to avoid that.
Signed-off-by: NJeffy Chen <jeffy.chen@rock-chips.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Slightly rework return value handling, no functional changes.
Signed-off-by: NJeffy Chen <jeffy.chen@rock-chips.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

The rockchip spi would stop driving pins when runtime suspended, which
might break slave's xfer(for example cros_ec).

Since we have pullups on those pins, we only need to care about this
when the CS asserted.

So let's keep the spi alive when chip select is asserted.

Also use pm_runtime_put instead of pm_runtime_put_sync.
Suggested-by: NDoug Anderson <dianders@chromium.org>
Signed-off-by: NJeffy Chen <jeffy.chen@rock-chips.com>
Reviewed-by: NDouglas Anderson <dianders@chromium.org>
Signed-off-by: NMark Brown <broonie@kernel.org>

The rockchip spi still requires slave selection when using GPIO CS.
Signed-off-by: NJeffy Chen <jeffy.chen@rock-chips.com>
Reviewed-by: NDouglas Anderson <dianders@chromium.org>
Signed-off-by: NMark Brown <broonie@kernel.org>

After failed to request dma tx chain, we need to disable pm_runtime.
Also cleanup error labels for better readability.
Signed-off-by: NJeffy Chen <jeffy.chen@rock-chips.com>
Reviewed-by: NBrian Norris <briannorris@chromium.org>
Signed-off-by: NMark Brown <broonie@kernel.org>

In the pattern of many other devices, support a system-sleep pin
configuration.
Signed-off-by: NBrian Norris <briannorris@chromium.org>
Reviewed-by: NDouglas Anderson <dianders@chromium.org>
Tested-by: NCaesar Wang <wxt@rock-chips.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

The Rockchip SPI controller's length register only supports 16-bits,
yielding a maximum length of 64KiB (the CTRLR1 register holds "length -
1"). Trying to transfer more than that (e.g., with a large SPI flash
read) will cause the driver to hang.

Now, it seems that while theoretically we should be able to program
CTRLR1 with 0xffff, and get a 64KiB transfer, but that also seems to
cause the core to choke, so stick with a maximum of 64K - 1 bytes --
i.e., 0xffff.
Signed-off-by: NBrian Norris <briannorris@chromium.org>
Signed-off-by: NMark Brown <broonie@kernel.org>

When using DMA, the transfer_one callback should return 1 because the
transfer hasn't finished yet.

A previous commit changed the function to return 0 when the DMA channels
were correctly prepared.

This manifested in Veyron boards with this message:

[ 1.983605] cros-ec-spi spi0.0: EC failed to respond in time

Fixes: ea984911 ("spi: rockchip: check return value of dmaengine_prep_slave_sg")
Signed-off-by: NTomeu Vizoso <tomeu.vizoso@collabora.com>
Signed-off-by: NMark Brown <broonie@kernel.org>
Cc: stable@vger.kernel.org

In gerenal, the "rockchip,rockchip-spi" string will match the dts
that's great in spi driver. After all the most of rockchip SoCs ar
same spi controller.

Then, we should keep the old style to match the dts various.
Signed-off-by: NCaesar Wang <wxt@rock-chips.com>
Signed-off-by: NMark Brown <broonie@kernel.org>