| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi into spi-5.11
|
|
SPI driver probing currently comprises two steps, whereas removal
comprises only one step:
spi_alloc_master()
spi_register_controller()
spi_unregister_controller()
That's because spi_unregister_controller() calls device_unregister()
instead of device_del(), thereby releasing the reference on the
spi_controller which was obtained by spi_alloc_master().
An SPI driver's private data is contained in the same memory allocation
as the spi_controller struct. Thus, once spi_unregister_controller()
has been called, the private data is inaccessible. But some drivers
need to access it after spi_unregister_controller() to perform further
teardown steps.
Introduce devm_spi_alloc_master() and devm_spi_alloc_slave(), which
release a reference on the spi_controller struct only after the driver
has unbound, thereby keeping the memory allocation accessible. Change
spi_unregister_controller() to not release a reference if the
spi_controller was allocated by one of these new devm functions.
The present commit is small enough to be backportable to stable.
It allows fixing drivers which use the private data in their ->remove()
hook after it's been freed. It also allows fixing drivers which neglect
to release a reference on the spi_controller in the probe error path.
Long-term, most SPI drivers shall be moved over to the devm functions
introduced herein. The few that can't shall be changed in a treewide
commit to explicitly release the last reference on the controller.
That commit shall amend spi_unregister_controller() to no longer release
a reference, thereby completing the migration.
As a result, the behaviour will be less surprising and more consistent
with subsystems such as IIO, which also includes the private data in the
allocation of the generic iio_dev struct, but calls device_del() in
iio_device_unregister().
Signed-off-by: Lukas Wunner <[email protected]>
Link: https://lore.kernel.org/r/272bae2ef08abd21388c98e23729886663d19192.1605121038.git.lukas@wunner.de
Signed-off-by: Mark Brown <[email protected]>
|
|
Provide a macro to filter all SPI_MODE_0,1,2,3 mode in one run.
The latest SPI framework will parse the devicetree in following call
sequence: of_register_spi_device() -> of_spi_parse_dt()
So, driver do not need to pars the devicetree and will get prepared
flags in the probe.
On one hand it is good far most drivers. On other hand some drivers need to
filter flags provide by SPI framework and apply know to work flags. This drivers
may use SPI_MODE_X_MASK to filter MODE flags and set own, known flags:
spi->flags &= ~SPI_MODE_X_MASK;
spi->flags |= SPI_MODE_0;
Signed-off-by: Oleksij Rempel <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
Silence documentation build warnings by correcting kernel-doc comment
for spi_transfer struct.
Signed-off-by: Colton Lewis <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
Change doubled word "as" to "as a".
Change "Return: Return:" in kernel-doc notation to have only one
"Return:".
Signed-off-by: Randy Dunlap <[email protected]>
Cc: Mark Brown <[email protected]>
Cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
Use kthread_create_worker() helper to simplify the code. It uses
the kthread worker API the right way. It will eventually allow
to remove the FIXME in kthread_worker_fn() and add more consistency
checks in the future.
Signed-off-by: Marek Szyprowski <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
On some SPI controllers (like spi-geni-qcom) setting the chip select
is a heavy operation. For instance on spi-geni-qcom, with the current
code, is was measured as taking upwards of 20 us. Even on SPI
controllers that aren't as heavy, setting the chip select is at least
something like a MMIO operation over some peripheral bus which isn't
as fast as a RAM access.
While it would be good to find ways to mitigate problems like this in
the drivers for those SPI controllers, it can also be noted that the
SPI framework could also help out. Specifically, in some situations,
we can see the SPI framework calling the driver's set_cs() with the
same parameter several times in a row. This is specifically observed
when looking at the way the Chrome OS EC SPI driver (cros_ec_spi)
works but other drivers likely trip it to some extent.
Let's solve this by caching the chip select state in the core and only
calling into the controller if there was a change. We check not only
the "enable" state but also the chip select mode (active high or
active low) since controllers may care about both the mode and the
enable flag in their callback.
Signed-off-by: Douglas Anderson <[email protected]>
Link: https://lore.kernel.org/r/20200629164103.1.Ied8e8ad8bbb2df7f947e3bc5ea1c315e041785a2@changeid
Signed-off-by: Mark Brown <[email protected]>
|
|
Add fallback to pio mode in case dma transfer failed with error status
SPI_TRANS_FAIL_NO_START.
If spi client driver want to enable this feature please set xfer->error in
the proper place such as dmaengine_prep_slave_sg() failure detect(but no
any data put into spi bus yet). Besides, add master->fallback checking in
its can_dma() so that spi core could switch to pio next time. Please refer
to spi-imx.c.
Signed-off-by: Robin Gong <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
We need to add a blank line to avoid this warning:
./include/linux/spi/spi.h:401: WARNING: Unexpected indentation.
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
Link: https://lore.kernel.org/r/1c701b3ac903dc0bc304dca958fbdee53bd38dc3.1586881715.git.mchehab+huawei@kernel.org
Signed-off-by: Jonathan Corbet <[email protected]>
|
|
some members were not described in documentation.
Signed-off-by: Qiujun Huang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
When dealing with a SPI controller driver that is sending more than 1
byte at once (or the entire buffer at once), and the SPI peripheral
driver has requested timestamping for a byte in the middle of the
buffer, we find that spi_take_timestamp_pre never records a "pre"
timestamp.
This happens because the function currently expects to be called with
the "progress" argument >= to what the peripheral has requested to be
timestamped. But clearly there are cases when that isn't going to fly.
And since we can't change the past when we realize that the opportunity
to take a "pre" timestamp has just passed and there isn't going to be
another one, the approach taken is to keep recording the "pre" timestamp
on each call, overwriting the previously recorded one until the "post"
timestamp is also taken.
Signed-off-by: Vladimir Oltean <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
Currently ACPI firmware description for a SPI device does not have any
method to describe the data buswidth on the board.
So even through the controller and device may support higher modes than
standard SPI, it cannot be assumed that the board does - as such, that
device is limited to standard SPI in such a circumstance.
As a workaround, allow the controller driver supply buswidth override bits,
which are used inform the core code that the controller driver knows the
buswidth supported on that board for that device.
A host controller driver might know this info from DMI tables, for example.
Signed-off-by: John Garry <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
|
|
Some SPI master controllers always drive a native chip select when
performing a transfer. Hence when using both native and GPIO chip
selects, at least one native chip select must be left unused, to be
driven when performing transfers with slave devices using GPIO chip
selects.
Currently, to find an unused native chip select, SPI controller drivers
need to parse and process cs-gpios theirselves. This is not only
duplicated in each driver that needs it, but also duplicates part of the
work done later at SPI controller registration time. Note that this
cannot be done after spi_register_controller() returns, as at that time,
slave devices may have been probed already.
Hence add generic support to the SPI subsystem for finding an unused
native chip select. Optionally, this unused native chip select, and all
other in-use native chip selects, can be validated against the maximum
number of native chip selects available on the controller hardware.
Signed-off-by: Geert Uytterhoeven <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
The API for PTP system timestamping (associating a SPI transaction with
the system time at which it was transferred) is flawed: it assumes that
the xfer->tx_buf pointer will always be present.
This is, of course, not always the case.
So introduce a "progress" variable that denotes how many word have been
transferred.
Fix the Freescale DSPI driver, the only user of the API so far, in the
same patch.
Fixes: b42faeee718c ("spi: Add a PTP system timestamp to the transfer structure")
Fixes: d6b71dfaeeba ("spi: spi-fsl-dspi: Implement the PTP system timestamping for TCFQ mode")
Signed-off-by: Vladimir Oltean <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
This change documents the CS setup, host & inactive times. They were
omitted when the fields were added, and were caught by one of the build
bots.
Fixes: 25093bdeb6bc ("spi: implement SW control for CS times")
Reported-by: kbuild test robot <[email protected]>
Signed-off-by: Alexandru Ardelean <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
The way the max delay is computed for this controller, it looks like it is
searching for the max delay from an SPI message a using that.
No idea if this is valid. But this change should support both `delay_usecs`
and the new `delay` data which is of `spi_delay` type.
Signed-off-by: Alexandru Ardelean <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
This change implements CS control for setup, hold & inactive delays.
The `cs_setup` delay is completely new, and can help with cases where
asserting the CS, also brings the device out of power-sleep, where there
needs to be a longer (than usual), before transferring data.
The `cs_hold` time can overlap with the `delay` (or `delay_usecs`) from an
SPI transfer. The main difference is that `cs_hold` implies that CS will be
de-asserted.
The `cs_inactive` delay does not have a clear use-case yet. It has been
implemented mostly because the `spi_set_cs_timing()` function implements
it. To some degree, this could overlap or replace `cs_change_delay`, but
this will require more consideration/investigation in the future.
All these delays have been added to the `spi_controller` struct, as they
would typically be configured by calling `spi_set_cs_timing()` after an
`spi_setup()` call.
Software-mode for CS control, implies that the `set_cs_timing()` hook has
not been provided for the `spi_controller` object.
Signed-off-by: Alexandru Ardelean <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
The initial version of `spi_set_cs_timing()` was implemented with
consideration only for clock-cycles as delay.
For cases like `CS setup` time, it's sometimes needed that micro-seconds
(or nano-seconds) are required, or sometimes even longer delays, for cases
where the device needs a little longer to start transferring that after CS
is asserted.
Signed-off-by: Alexandru Ardelean <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
The change introduces the `delay` field to the `spi_transfer` struct as an
`struct spi_delay` type.
This intends to eventually replace `delay_usecs`.
But, since there are many users of `delay_usecs`, this needs some
intermediate work.
A helper called `spi_transfer_delay_exec()` is also added, which maintains
backwards compatibility with `delay_usecs`, by assigning the value to
`delay` if non-zero.
This should maintain backwards compatibility with current users of
`udelay_usecs`.
Signed-off-by: Alexandru Ardelean <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
This change does a conversion from the `word_delay_usecs` -> `word_delay`
for the `spi_device` struct.
This allows users to specify inter-word delays in other unit types
(nano-seconds or clock cycles), depending on how users want.
The Atmel SPI driver is the only current user of the `word_delay_usecs`
field (from the `spi_device` struct).
So, it needed a slight conversion to use the `word_delay` as an `spi_delay`
struct.
In SPI core, the only required mechanism is to update the `word_delay`
information per `spi_transfer`. This requires a bit more logic than before,
because it needs that both delays be converted to a common unit
(nano-seconds) for comparison.
Signed-off-by: Alexandru Ardelean <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
The Spreadtrum SPI driver is the only user of the `word_delay` field in
the `spi_transfer` struct.
This change converts the field to use the `spi_delay` struct. This also
enforces the users to specify the delay unit to be `SPI_DELAY_UNIT_SCK`.
Signed-off-by: Alexandru Ardelean <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
Since the logic for `spi_delay` struct + `spi_delay_exec()` has been copied
from the `cs_change_delay` logic, it's natural to make this delay, the
first user.
The `cs_change_delay` logic requires that the default remain 10 uS, in case
it is unspecified/unconfigured. So, there is some special handling needed
to do that.
The ADIS library is one of the few users of the new `cs_change_delay`
parameter for an spi_transfer.
The introduction of the `spi_delay` struct, requires that the users of of
`cs_change_delay` get an update. This change also updates the ADIS library.
Signed-off-by: Alexandru Ardelean <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
There are plenty of delays that have been introduced in SPI core. Most of
them are in micro-seconds, some need to be in nano-seconds, and some in
clock-cycles.
For some of these delays (related to transfers & CS timing) it may make
sense to have a `spi_delay` struct that abstracts these a bit.
The important element of these delays [for unification] seems to be the
`unit` of the delay.
It looks like micro-seconds is good enough for most people, but every-once
in a while, some delays seem to require other units of measurement.
This change adds the `spi_delay` struct & a `spi_delay_exec()` function
that processes a `spi_delay` object/struct to execute the delay.
It's a copy of the `cs_change_delay` mechanism, but without the default
for 10 uS.
The clock-cycle delay unit is a bit special, as it needs to be bound to an
`spi_transfer` object to execute.
Signed-off-by: Alexandru Ardelean <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
SPI is one of the interfaces used to access devices which have a POSIX
clock driver (real time clocks, 1588 timers etc). The fact that the SPI
bus is slow is not what the main problem is, but rather the fact that
drivers don't take a constant amount of time in transferring data over
SPI. When there is a high delay in the readout of time, there will be
uncertainty in the value that has been read out of the peripheral.
When that delay is constant, the uncertainty can at least be
approximated with a certain accuracy which is fine more often than not.
Timing jitter occurs all over in the kernel code, and is mainly caused
by having to let go of the CPU for various reasons such as preemption,
servicing interrupts, going to sleep, etc. Another major reason is CPU
dynamic frequency scaling.
It turns out that the problem of retrieving time from a SPI peripheral
with high accuracy can be solved by the use of "PTP system
timestamping" - a mechanism to correlate the time when the device has
snapshotted its internal time counter with the Linux system time at that
same moment. This is sufficient for having a precise time measurement -
it is not necessary for the whole SPI transfer to be transmitted "as
fast as possible", or "as low-jitter as possible". The system has to be
low-jitter for a very short amount of time to be effective.
This patch introduces a PTP system timestamping mechanism in struct
spi_transfer. This is to be used by SPI device drivers when they need to
know the exact time at which the underlying device's time was
snapshotted. More often than not, SPI peripherals have a very exact
timing for when their SPI-to-interconnect bridge issues a transaction
for snapshotting and reading the time register, and that will be
dependent on when the SPI-to-interconnect bridge figures out that this
is what it should do, aka as soon as it sees byte N of the SPI transfer.
Since spi_device drivers are the ones who'd know best how the peripheral
behaves in this regard, expose a mechanism in spi_transfer which allows
them to specify which word (or word range) from the transfer should be
timestamped.
Add a default implementation of the PTP system timestamping in the SPI
core. This is not going to be satisfactory performance-wise, but should
at least increase the likelihood that SPI device drivers will use PTP
system timestamping in the future.
There are 3 entry points from the core towards the SPI controller
drivers:
- transfer_one: The driver is passed individual spi_transfers to
execute. This is the easiest to timestamp.
- transfer_one_message: The core passes the driver an entire spi_message
(a potential batch of spi_transfers). The core puts the same pre and
post timestamp to all transfers within a message. This is not ideal,
but nothing better can be done by default anyway, since the core has
no insight into how the driver batches the transfers.
- transfer: Like transfer_one_message, but for unqueued drivers (i.e.
the driver implements its own queue scheduling).
Signed-off-by: Vladimir Oltean <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
|
|
Compiler is not happy about spi_set_cs_timing() prototype.
drivers/spi/spi.c:3016:6: warning: no previous prototype for ‘spi_set_cs_timing’ [-Wmissing-prototypes]
void spi_set_cs_timing(struct spi_device *spi, u8 setup, u8 hold,
^~~~~~~~~~~~~~~~~
Let's add it to the header.
Signed-off-by: Andy Shevchenko <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
|
|
Right now the only way to get the SPI pumping thread bumped up to
realtime priority is for the controller to request it. However it may
be that the controller works fine with the normal priority but
communication to a particular SPI device on the bus needs realtime
priority.
Let's add a way for devices to request realtime priority when they set
themselves up.
NOTE: this will just affect the priority of transfers that end up on
the SPI core's pumping thread. In many cases transfers happen in the
context of the caller so if you need realtime priority for all
transfers you should ensure the calling context is also realtime
priority.
Signed-off-by: Douglas Anderson <[email protected]>
Reviewed-by: Guenter Roeck <[email protected]>
Tested-by: Enric Balletbo i Serra <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
|
|
Provide a means for the spi bus driver to report the effectively used
spi clock frequency used for each spi_transfer.
Signed-off-by: Martin Sperl <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
|
|
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi into spi-5.3
spi: Add spi_is_bpw_supported()
Lets client drivers check and potentially handle issues.
|
|
Support setting a delay between cs assert and deassert as
a multiple of spi clock length.
Signed-off-by: Martin Sperl <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
|
|
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 <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
|
|
This patch changes mode and mode_bits from u16 to u32 to allow more
mode configurations.
Signed-off-by: Sowjanya Komatineni <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
|
|
This let SPI clients check if the controller supports a particular word
width. drivers/gpu/drm/tinydrm/mipi-dbi.c will use this to determine if
the controller supports 16-bit for RGB565 pixels. If it doesn't it will
swap the bytes before transfer on little endian machines.
Signed-off-by: Noralf Trønnes <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
|
|
This patch creates set_cs_timing SPI master optional method for
SPI masters to implement configuring CS timing if applicable.
This patch also creates spi_cs_timing accessory for SPI clients to
use for requesting SPI master controllers to configure device requested
CS setup time, hold time and inactive delay.
Signed-off-by: Sowjanya Komatineni <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
|
|
Geert points out that I confused the min/max arguments that are
reversed between SPI_BPW_RANGE_MASK() and GENMASK(). This time
I have verified the result of the macro after fixing the arguments.
Cc: Geert Uytterhoeven <[email protected]>
Fixes: eefffb42f665 ("spi: work around clang bug in SPI_BPW_RANGE_MASK()")
Signed-off-by: Arnd Bergmann <[email protected]>
Reviewed-by: Geert Uytterhoeven <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
|
|
Clang-8 evaluates both sides of a ?: expression to check for
valid arithmetic even in the side that is never taken. This
results in a build warning:
drivers/spi/spi-sh-msiof.c:1052:24: error: shift count >= width of type [-Werror,-Wshift-count-overflow]
.bits_per_word_mask = SPI_BPW_RANGE_MASK(8, 32),
^~~~~~~~~~~~~~~~~~~~~~~~~
Change the implementation to use the GENMASK() macro that does
what we want here but does not have a problem with the shift
count overflow.
Link: https://bugs.llvm.org/show_bug.cgi?id=38789
Signed-off-by: Arnd Bergmann <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
|
|
Some devices are slow and cannot keep up with the SPI bus and therefore
require a short delay between words of the SPI transfer.
The example of this that I'm looking at is a SAMA5D2 with a minimum SPI
clock of 400kHz talking to an AVR-based SPI slave. The AVR cannot put
bytes on the bus fast enough to keep up with the SoC's SPI controller
even at the lowest bus speed.
This patch introduces the ability to specify a required inter-word
delay for SPI devices. It is up to the controller driver to configure
itself accordingly in order to introduce the requested delay.
Note that, for spi_transfer, there is already a field word_delay that
provides similar functionality. This field, however, is specified in
clock cycles (and worse, SPI controller cycles, not SCK cycles); that
makes this value dependent on the master clock instead of the device
clock for which the delay is intended to provide some relief. This
patch leaves this old word_delay in place and provides a time-based
word_delay_us alongside it; the new field fits in the struct padding
so struct size is constant. There is only one in-kernel user of the
word_delay field and presumably that driver could be reworked to use
the time-based value instead.
The time-based delay is limited to 8 bits as these delays are intended
to be short. The SAMA5D2 that I've tested this on limits delays to a
maximum of ~100us, which is already many word-transfer periods even at
the minimum transfer speed supported by the controller.
Signed-off-by: Jonas Bonn <[email protected]>
CC: Mark Brown <[email protected]>
CC: Rob Herring <[email protected]>
CC: Mark Rutland <[email protected]>
CC: [email protected]
CC: [email protected]
Signed-off-by: Mark Brown <[email protected]>
|
|
Commit 412e6037324 ("spi: core: avoid waking pump thread from spi_sync
instead run teardown delayed") introduced regressions on some boards,
apparently connected to spi_mem not triggering shutdown properly any
more. Since we've thus far been unable to figure out exactly where the
breakage is revert the optimisation for now.
Reported-by: Jon Hunter <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
Cc: [email protected]
|
|
When spi_sync is running alone with no other spi devices connected
to the bus the worker thread is woken during spi_finalize_current_message
to run the teardown code every time.
This is totally unnecessary in the case that there is no message queued.
On a multi-core system this results in one wakeup of the thread for each
spi_message processed via spi_sync where in most cases the teardown does
not happen as the hw is already in use.
This patch now delays the teardown by 1 second by using a separate
kthread_delayed_work for the teardown.
This avoids waking the kthread too often.
For spi_sync transfers in a tight loop (say 40k messages/s) this
avoids the penalty of waking the worker thread 40k times/s.
On a rasperry pi 3 with 4 cores the results in 32% of a single core
only to find out that there is nothing in the queue and it can go back
to sleep.
With this patch applied the spi-worker is woken exactly once: after
the load finishes and the spi bus is idle for 1 second.
I believe I have also seen situations where during a spi_sync loop
the worker thread (triggered by the last message finished) is slightly
faster and _wins_ the race to process the message, so we are actually
running the kthread and letting it do some work...
This is also no longer observed with this patch applied as.
Tested with a new CAN controller driver for the mcp2517fd which
uses spi_sync for interrupt handling and spi_async for scheduling
of can frames for transmission (in a different thread)
Some statistics when receiving 100000 CAN frames with the mcp25xxfd driver
on a Raspberry pi 3:
without the patch:
------------------
root@raspcm3:~# for x in $(pgrep spi0) $(pgrep irq/94-mcp25xxf) ; do awk '{printf "%-20s %6i\n", $2,$15}' /proc/$x/stat; done
(spi0) 5
(irq/94-mcp25xxf) 0
root@raspcm3:~# vmstat 1
procs -----------memory---------- ---swap-- -----io---- -system-- ------cpu-----
r b swpd free buff cache si so bi bo in cs us sy id wa st
1 0 0 821960 13592 50848 0 0 80 2 1986 105 1 2 97 0 0
0 0 0 821968 13592 50876 0 0 0 0 8046 30 0 0 100 0 0
0 0 0 821936 13592 50876 0 0 0 0 8032 24 0 0 100 0 0
0 0 0 821936 13592 50876 0 0 0 0 8035 30 0 0 100 0 0
0 0 0 821936 13592 50876 0 0 0 0 8033 22 0 0 100 0 0
2 0 0 821936 13592 50876 0 0 0 0 11598 7129 0 3 97 0 0
1 0 0 821872 13592 50876 0 0 0 0 37741 59003 0 31 69 0 0
2 0 0 821840 13592 50876 0 0 0 0 37762 59078 0 29 71 0 0
2 0 0 821776 13592 50876 0 0 0 0 37593 58792 0 28 72 0 0
1 0 0 821744 13592 50876 0 0 0 0 37642 58881 0 30 70 0 0
2 0 0 821680 13592 50876 0 0 0 0 37490 58602 0 27 73 0 0
1 0 0 821648 13592 50876 0 0 0 0 37412 58418 0 29 71 0 0
1 0 0 821584 13592 50876 0 0 0 0 37337 58288 0 27 73 0 0
1 0 0 821552 13592 50876 0 0 0 0 37584 58774 0 27 73 0 0
0 0 0 821520 13592 50876 0 0 0 0 18363 20566 0 9 91 0 0
0 0 0 821520 13592 50876 0 0 0 0 8037 32 0 0 100 0 0
0 0 0 821520 13592 50876 0 0 0 0 8031 23 0 0 100 0 0
0 0 0 821520 13592 50876 0 0 0 0 8034 26 0 0 100 0 0
0 0 0 821520 13592 50876 0 0 0 0 8033 24 0 0 100 0 0
^C
root@raspcm3:~# for x in $(pgrep spi0) $(pgrep irq/94-mcp25xxf) ; do awk '{printf "%-20s %6i\n", $2,$15}' /proc/$x/stat; done
(spi0) 228
(irq/94-mcp25xxf) 794
root@raspcm3:~# cat /proc/interrupts
CPU0 CPU1 CPU2 CPU3
17: 34 0 0 0 ARMCTRL-level 1 Edge 3f00b880.mailbox
27: 1 0 0 0 ARMCTRL-level 35 Edge timer
33: 1416870 0 0 0 ARMCTRL-level 41 Edge 3f980000.usb, dwc2_hsotg:usb1
34: 1 0 0 0 ARMCTRL-level 42 Edge vc4
35: 0 0 0 0 ARMCTRL-level 43 Edge 3f004000.txp
40: 1753 0 0 0 ARMCTRL-level 48 Edge DMA IRQ
42: 11 0 0 0 ARMCTRL-level 50 Edge DMA IRQ
44: 11 0 0 0 ARMCTRL-level 52 Edge DMA IRQ
45: 0 0 0 0 ARMCTRL-level 53 Edge DMA IRQ
66: 0 0 0 0 ARMCTRL-level 74 Edge vc4 crtc
69: 0 0 0 0 ARMCTRL-level 77 Edge vc4 crtc
70: 0 0 0 0 ARMCTRL-level 78 Edge vc4 crtc
77: 20 0 0 0 ARMCTRL-level 85 Edge 3f205000.i2c, 3f804000.i2c, 3f805000.i2c
78: 6346 0 0 0 ARMCTRL-level 86 Edge 3f204000.spi
80: 205 0 0 0 ARMCTRL-level 88 Edge mmc0
81: 493 0 0 0 ARMCTRL-level 89 Edge uart-pl011
89: 0 0 0 0 bcm2836-timer 0 Edge arch_timer
90: 4291 3821 2180 1649 bcm2836-timer 1 Edge arch_timer
94: 14289 0 0 0 pinctrl-bcm2835 16 Level mcp25xxfd
IPI0: 0 0 0 0 CPU wakeup interrupts
IPI1: 0 0 0 0 Timer broadcast interrupts
IPI2: 3645 242371 7919 1328 Rescheduling interrupts
IPI3: 112 543 273 194 Function call interrupts
IPI4: 0 0 0 0 CPU stop interrupts
IPI5: 1 0 0 0 IRQ work interrupts
IPI6: 0 0 0 0 completion interrupts
Err: 0
top shows 93% for the mcp25xxfd interrupt handler, 31% for spi0.
with the patch:
---------------
root@raspcm3:~# for x in $(pgrep spi0) $(pgrep irq/94-mcp25xxf) ; do awk '{printf "%-20s %6i\n", $2,$15}' /proc/$x/stat; done
(spi0) 0
(irq/94-mcp25xxf) 0
root@raspcm3:~# vmstat 1
procs -----------memory---------- ---swap-- -----io---- -system-- ------cpu-----
0 0 0 804768 13584 62628 0 0 0 0 8038 24 0 0 100 0 0
0 0 0 804768 13584 62628 0 0 0 0 8042 25 0 0 100 0 0
1 0 0 804704 13584 62628 0 0 0 0 9603 2967 0 20 80 0 0
1 0 0 804672 13584 62628 0 0 0 0 9828 3380 0 24 76 0 0
1 0 0 804608 13584 62628 0 0 0 0 9823 3375 0 23 77 0 0
1 0 0 804608 13584 62628 0 0 0 12 9829 3394 0 23 77 0 0
1 0 0 804544 13584 62628 0 0 0 0 9816 3362 0 22 78 0 0
1 0 0 804512 13584 62628 0 0 0 0 9817 3367 0 23 77 0 0
1 0 0 804448 13584 62628 0 0 0 0 9822 3370 0 22 78 0 0
1 0 0 804416 13584 62628 0 0 0 0 9815 3367 0 23 77 0 0
0 0 0 804352 13584 62628 0 0 0 84 9222 2250 0 14 86 0 0
0 0 0 804352 13592 62620 0 0 0 24 8131 209 0 0 93 7 0
0 0 0 804320 13592 62628 0 0 0 0 8041 27 0 0 100 0 0
0 0 0 804352 13592 62628 0 0 0 0 8040 26 0 0 100 0 0
root@raspcm3:~# for x in $(pgrep spi0) $(pgrep irq/94-mcp25xxf) ; do awk '{printf "%-20s %6i\n", $2,$15}' /proc/$x/stat; done
(spi0) 0
(irq/94-mcp25xxf) 767
root@raspcm3:~# cat /proc/interrupts
CPU0 CPU1 CPU2 CPU3
17: 29 0 0 0 ARMCTRL-level 1 Edge 3f00b880.mailbox
27: 1 0 0 0 ARMCTRL-level 35 Edge timer
33: 1024412 0 0 0 ARMCTRL-level 41 Edge 3f980000.usb, dwc2_hsotg:usb1
34: 1 0 0 0 ARMCTRL-level 42 Edge vc4
35: 0 0 0 0 ARMCTRL-level 43 Edge 3f004000.txp
40: 1773 0 0 0 ARMCTRL-level 48 Edge DMA IRQ
42: 11 0 0 0 ARMCTRL-level 50 Edge DMA IRQ
44: 11 0 0 0 ARMCTRL-level 52 Edge DMA IRQ
45: 0 0 0 0 ARMCTRL-level 53 Edge DMA IRQ
66: 0 0 0 0 ARMCTRL-level 74 Edge vc4 crtc
69: 0 0 0 0 ARMCTRL-level 77 Edge vc4 crtc
70: 0 0 0 0 ARMCTRL-level 78 Edge vc4 crtc
77: 20 0 0 0 ARMCTRL-level 85 Edge 3f205000.i2c, 3f804000.i2c, 3f805000.i2c
78: 6417 0 0 0 ARMCTRL-level 86 Edge 3f204000.spi
80: 237 0 0 0 ARMCTRL-level 88 Edge mmc0
81: 489 0 0 0 ARMCTRL-level 89 Edge uart-pl011
89: 0 0 0 0 bcm2836-timer 0 Edge arch_timer
90: 4048 3704 2383 1892 bcm2836-timer 1 Edge arch_timer
94: 14287 0 0 0 pinctrl-bcm2835 16 Level mcp25xxfd
IPI0: 0 0 0 0 CPU wakeup interrupts
IPI1: 0 0 0 0 Timer broadcast interrupts
IPI2: 2361 2948 7890 1616 Rescheduling interrupts
IPI3: 65 617 301 166 Function call interrupts
IPI4: 0 0 0 0 CPU stop interrupts
IPI5: 1 0 0 0 IRQ work interrupts
IPI6: 0 0 0 0 completion interrupts
Err: 0
top shows 91% for the mcp25xxfd interrupt handler, 0% for spi0
So we see that spi0 is no longer getting scheduled wasting CPU cycles
There are a lot less context switches and corresponding Rescheduling interrupts
All of these show that this improves efficiency of the system and reduces
CPU utilization.
Signed-off-by: Martin Sperl <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
|
|
This augments the SPI core to optionally use GPIO descriptors
for chip select on a per-master-driver opt-in basis.
Drivers using this will rely on the SPI core to look up
GPIO descriptors associated with the device, such as
when using device tree or board files with GPIO descriptor
tables.
When getting descriptors from the device tree, this will in
turn activate the code in gpiolib that was
added in commit 6953c57ab172
("gpio: of: Handle SPI chipselect legacy bindings")
which means that these descriptors are aware of the active
low semantics that is the default for SPI CS GPIO lines
and we can assume that all of these are "active high" and
thus assign SPI_CS_HIGH to all CS lines on the DT path.
The previously used gpio_set_value() would call down into
gpiod_set_raw_value() and ignore the polarity inversion
semantics.
It seems like many drivers go to great lengths to set up the
CS GPIO line as non-asserted, respecting SPI_CS_HIGH. We pull
this out of the SPI drivers and into the core, and by simply
requesting the line as GPIOD_OUT_LOW when retrieveing it from
the device and relying on the gpiolib to handle any inversion
semantics. This way a lot of code can be simplified and
removed in each converted driver.
The end goal after dealing with each driver in turn, is to
delete the non-descriptor path (of_spi_register_master() for
example) and let the core deal with only descriptors.
The different SPI drivers have complex interactions with the
core so we cannot simply change them all over, we need to use
a stepwise, bisectable approach so that each driver can be
converted and fixed in isolation.
This patch has the intended side effect of adding support for
ACPI GPIOs as it starts relying on gpiod_get_*() to get
the GPIO handle associated with the device.
Cc: Linuxarm <[email protected]>
Acked-by: Jonathan Cameron <[email protected]>
Tested-by: Fangjian (Turing) <[email protected]>
Signed-off-by: Linus Walleij <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
|
|
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi into spi-4.21 mode conflict
|
|
Add flags for Octal mode I/O data transfer
Required for the SPI controller which can do the data transfer (TX/RX)
on 8 data lines e.g. NXP FlexSPI controller.
SPI_TX_OCTAL: transmit with 8 wires
SPI_RX_OCTAL: receive with 8 wires
Signed-off-by: Yogesh Gaur <[email protected]>
Reviewed-by: Boris Brezillon <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
|
|
Some devices such as the TPO TPG110 display panel require
a "high-impedance turn-around", in effect a clock cycle after
switching the line from output to input mode.
Support this in the GPIO driver to begin with. Other driver
may implement it if they can, it is unclear if this can
be achieved with anything else than GPIO bit-banging.
Cc: Andrzej Hajda <[email protected]>
Acked-by: Lorenzo Bianconi <[email protected]>
Signed-off-by: Linus Walleij <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
|
|
|
|
This attribute works the same was as the identically named attribute
for PCI, AMBA, and platform devices. For reference, see:
commit 3cf385713460 ("ARM: 8256/1: driver coamba: add device binding
path 'driver_override'")
commit 3d713e0e382e ("driver core: platform: add device binding path
'driver_override'")
commit 782a985d7af2 ("PCI: Introduce new device binding path using
pci_dev.driver_override")
If the name of a driver is written to this attribute, then the device
will bind to the named driver and only the named driver.
The device will bind to the driver even if the driver does not list the
device in its id table. This behavior is different than the driver's
bind attribute, which only allows binding to devices that are listed as
supported by the driver.
It can be used to bind a generic driver, like spidev, to a device.
Signed-off-by: Trent Piepho <[email protected]>
Reviewed-by: Jan Kundrát <[email protected]>
Tested-by: Jan Kundrát <[email protected]>
Reviewed-by: Geert Uytterhoeven <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
|
|
The of_find_spi_device_by_node() helper function is useful for other
modules too. Export the funciton as GPL like all other spi helper
functions and make it available if CONFIG_OF is enabled, because it isn't
related to the CONFIG_OF_DYNAMIC context. Finally add a stub if
CONFIG_OF isn't enabled, so others must not care about it.
Signed-off-by: Marco Felsch <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
|
|
Use the appropriate SPDX license identifier and drop the previous
license text.
Signed-off-by: Marco Felsch <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
|
|
spi: Provide SPI_CS_WORD
This provides a SPI operation mode which changes chip select after every
word, used by some devices such as ADCs and DACs.
|
|
This adds a new SPI mode flag, SPI_CS_WORD, that is used to indicate
that a SPI device requires the chip select to be toggled after each
word that is transferred.
Signed-off-by: David Lechner <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
|