Ancient legacy of me doing it wrong which it is nice to clear
up whilst we are here.
Signed-off-by: NJonathan Cameron <jic23@kernel.org>
Reviewed-by: NLars-Peter Clausen <lars@metafoo.de>

Nothing uses it outside of core code.
Signed-off-by: NJonathan Cameron <jic23@kernel.org>
Reviewed-by: NLars-Peter Clausen <lars@metafoo.de>

This is a necessary step in taking the buffer implementation
opaque.
Signed-off-by: NJonathan Cameron <jic23@kernel.org>
Reviewed-by: NLars-Peter Clausen <lars@metafoo.de>

Nothing outside of indiustrialio-buffer.c should be using this.
Requires a large amount of juggling of functions to avoid a
forward definition.
Signed-off-by: NJonathan Cameron <jic23@kernel.org>
Reviewed-by: NLars-Peter Clausen <lars@metafoo.de>

A multiplication for the size determination of a memory allocation
indicated that an array data structure should be processed.
Thus use the corresponding function "kmalloc_array".

This issue was detected by using the Coccinelle software.
Signed-off-by: NMarkus Elfring <elfring@users.sourceforge.net>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

A recent fix to iio_buffer_read_first_n_outer removed ret from being set by
a return from wait_event_interruptible and also added a continue in a loop
which causes the variable ret to not be set when it reaches the end of the
loop.  Fix this by initializing ret to zero.

Also remove extraneous white space at the end of the loop.

Fixes: fcf68f3c ("fix sched WARNING "do not call blocking ops when !TASK_RUNNING")
Signed-off-by: NColin Ian King <colin.king@canonical.com>
Cc: <Stable@vger.kernel.org>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

When using CONFIG_DEBUG_ATOMIC_SLEEP, the scheduler nicely points out
that we're calling sleeping primitives within the wait_event loop, which
means we might clobber the task state:

[   10.831289] do not call blocking ops when !TASK_RUNNING; state=1 set at [<ffffffc00026b610>]
[   10.845531] ------------[ cut here ]------------
[   10.850161] WARNING: at kernel/sched/core.c:7630
...
[   12.164333] ---[ end trace 45409966a9a76438 ]---
[   12.168942] Call trace:
[   12.171391] [<ffffffc00024ed44>] __might_sleep+0x64/0x90
[   12.176699] [<ffffffc000954774>] mutex_lock_nested+0x50/0x3fc
[   12.182440] [<ffffffc0007b9424>] iio_kfifo_buf_data_available+0x28/0x4c
[   12.189043] [<ffffffc0007b76ac>] iio_buffer_ready+0x60/0xe0
[   12.194608] [<ffffffc0007b7834>] iio_buffer_read_first_n_outer+0x108/0x1a8
[   12.201474] [<ffffffc000370d48>] __vfs_read+0x58/0x114
[   12.206606] [<ffffffc000371740>] vfs_read+0x94/0x118
[   12.211564] [<ffffffc0003720f8>] SyS_read+0x64/0xb4
[   12.216436] [<ffffffc000203cb4>] el0_svc_naked+0x24/0x28

To avoid this, we should (a la https://lwn.net/Articles/628628/) use the
wait_woken() function, which avoids the nested sleeping while still
handling races between waiting / wake-events.
Signed-off-by: NBrian Norris <briannorris@chromium.org>
Reviewed-by: NLars-Peter Clausen <lars@metafoo.de>
Cc: <Stable@vger.kernel.org> # 3.19+ for introduction of wake_woken
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

config structure is set to 0 when updating the buffers, so by
default config->watermark will be 0. When computing the minimum
between config->watermark and the buffer->watermark or
insert_buffer-watermark, this will always be 0 regardless of the
value set by the user for the buffer.

Set as initial value for config->watermark the maximum allowed
value so that the minimum value will always be set from one of the
buffers.
Signed-off-by: NIrina Tirdea <irina.tirdea@intel.com>
Fixes: f0566c0c ("iio: Set device watermark based on watermark of all
attached buffers")
Cc: <Stable@vger.kernel.org>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

We have the same code for computing the scan index storage size in bytes
all over the place. Factor this out into helper functions.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Acked-by: NDaniel Baluta <daniel.baluta@intel.com>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

WARN_ON() only takes a condition argument.  I have changed these to
WARN() instead.
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Cc: <Stable@vger.kernel.org>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

This patch adds a enable and disable callback that is called when the
buffer is enabled/disabled. This can be used by buffer implementations that
need to do some setup or teardown work. E.g. a DMA based buffer can use
this to start/stop the DMA transfer.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

For buffers which have a fixed wake-up watermark the watermark attribute
should be read-only. Add a new FIXED_WATERMARK flag to the
struct iio_buffer_access_funcs, which can be set by a buffer
implementation.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

Only initialize the watermark field if it is still 0. This allows drivers
to provide a custom default watermark value. E.g. some driver might have a
fixed watermark or can only support watermarks within a certain range and
the initial value for the watermark should be within this range.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

Currently the watermark of the device is only set based on the watermark
that is set for the user space buffer. This doesn't consider the watermarks
set on any attached in-kernel buffers.

Change this so that the watermark of the device should be the minimum of
the watermarks over all attached buffers.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

Change return value to 0 if no device is bound since
unsigned int cannot support negative error codes.

Fixes: f18e7a06 ("iio: Return -ENODEV for file operations if the
device has been unregistered")
Signed-off-by: NCristina Opriceana <cristina.opriceana@gmail.com>
Cc: <Stable@vger.kernel.org>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

Fix kernel docs for structures and functions in order to
remove some warnings when the documentation gets generated.
Signed-off-by: NCristina Opriceana <cristina.opriceana@gmail.com>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

This patch changes the semantics of non-blocking reads so that a
hardware fifo flush is triggered if the available data in the device
buffer is less then the requested size.

This allows userspace to accurately generate hardware fifo flushes, by
doing a non-blocking read with a size greater then the sum of the
device buffer and hardware fifo size.
Signed-off-by: NOctavian Purdila <octavian.purdila@intel.com>
Reviewed-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

In hardware mode we can not use the software demuxer, this means that the
selected scan mask needs to match one of the available scan masks exactly.

It also means that all attached buffers need to use the same scan mask.
Given that when operating in hardware mode there is typically only a single
buffer attached to the device this not an issue. Add a sanity check to make
sure that only a single buffer is attached in hardware mode nevertheless.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

For each buffer type specify the supported device modes for this buffer.
This allows us for devices which support multiple different operating modes
to pick the correct operating mode based on the modes supported by the
attached buffers.

It also prevents that buffers with conflicting modes are attached
to a device at the same time or that a buffer with a non-supported mode is
attached to a device (e.g. in-kernel callback buffer to a device only
supporting hardware mode).
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

Even if no userspace consumer buffer is attached to the IIO device at
registration we still need to compute the masklength, since it is possible
that a in-kernel consumer buffer is going to get attached to the device at
a later point.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

The same code is executed regardless ret value, so this test
can be removed.
Also fix coverity scan CID 1268786.
Signed-off-by: NLaurent Navet <laurent.navet@gmail.com>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

Currently when something goes wrong at some step when disabling the buffers
we immediately abort. This has the effect that the enable/disable calls are
no longer balanced. So make sure that even if one step in the disable
sequence fails the other steps are still executed.

The other issue is that when either enable or disable fails buffers that
were active at that time stay active while the device itself is disabled.
This leaves things in a inconsistent state and can cause unbalanced
enable/disable calls. Furthermore when enable fails we restore the old scan
mask, but still keeps things disabled.

Given that verification of the configuration was performed earlier and it
is valid at the point where we try to enable/disable the most likely reason
of failure is a communication failure with the device or maybe a
out-of-memory situation. There is not really a good recovery strategy in
such a case, so it makes sense to leave the device disabled, but we should
still leave it in a consistent state.

What the patch does if disable/enable fails is to deactivate all buffers
and make sure that the device will be in the same state as if all buffers
had been manually disabled.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

__iio_update_buffers is already a rather large function with many different
error paths and it is going to get even larger. This patch factors out the
device enable and device disable paths into separate helper functions.

The patch also re-implements iio_disable_all_buffers() using the new
iio_disable_buffers() function removing a fair bit of redundant code.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

Currently __iio_update_buffers() verifies whether the new configuration
will work in the middle of the update sequence. This means if the new
configuration is invalid we need to rollback the changes already made. This
patch moves the validation of the new configuration at the beginning of
__iio_update_buffers() and will not start to make any changes if the new
configuration is invalid.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

We only have to call the request_update() callback for a newly inserted
buffer. The configuration of the already previously active buffers will not
have changed.

This also allows us to move the request_update() call to the beginning of
__iio_update_buffers(), before any currently active buffers are stopped.
This makes the error handling a lot easier since no changes were made to
the buffer list and no rollback needs to be performed.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

Add a small helper function iio_free_scan_mask() that takes a mask and
frees its memory if the scan masks for the device are dynamically
allocated, otherwise does nothing. This means we don't have to open-code
the same check over and over again in __iio_update_buffers.

Also free compound_mask as soon a we are done using it. This constrains its
usage to a specific region of the function will make further refactoring
and splitting the function into smaller sub-parts more easier.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

While more verbose error messages are useful for debugging we should really
not put those error messages into the kernel log for normal errors that are
already reported to the application via the error code, when running in
non-debug mode.

Otherwise application authors might expect that this is part of the ABI and
to get the error they should scan the kernel log. Which would be rather
error prone itself since there is no direct mapping between a operation and
the error message so it is impossible to find out which error message
belongs to which error.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

Some devices have hardware buffers that can store a number of samples
for later consumption. Hardware usually provides interrupts to notify
the processor when the FIFO is full or when it has reached a certain
watermark level. This helps with reducing the number of interrupts to
the host processor and thus it helps decreasing the power consumption.

This patch enables usage of hardware FIFOs for IIO devices in
conjunction with software device buffers. When the hardware FIFO is
enabled the samples are stored in the hardware FIFO. The samples are
later flushed to the device software buffer when the number of entries
in the hardware FIFO reaches the hardware watermark or when a flush
operation is triggered by the user when doing a non-blocking read
on an empty software device buffer.

In order to implement hardware FIFO support the device drivers must
implement the following new operations: setting and getting the
hardware FIFO watermark level, flushing the hardware FIFO to the
software device buffer. The device must also expose information about
the hardware FIFO such it's minimum and maximum watermark and if
necessary a list of supported watermark values. Finally, the device
driver must activate the hardware FIFO when the device buffer is
enabled, if the current device settings allows it.

The software device buffer watermark is passed by the IIO core to the
device driver as a hint for the hardware FIFO watermark. The device
driver can adjust this value to allow for hardware limitations (such
as capping it to the maximum hardware watermark or adjust it to a
value that is supported by the hardware). It can also disable the
hardware watermark (and implicitly the hardware FIFO) it this value is
below the minimum hardware watermark.

Since a driver may support hardware FIFO only when not in triggered
buffer mode (due to different semantics of hardware FIFO sampling and
triggered sampling) this patch changes the IIO core code to allow
falling back to non-triggered buffered mode if no trigger is enabled.
Signed-off-by: NOctavian Purdila <octavian.purdila@intel.com>
Reviewed-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

Currently the IIO buffer blocking read only wait until at least one
data element is available.
This patch makes the reader sleep until enough data is collected before
returning to userspace. This should limit the read() calls count when
trying to get data in batches.

Co-author: Yannick Bedhomme <yannick.bedhomme@mobile-devices.fr>
Signed-off-by: NJosselin Costanzi <josselin.costanzi@mobile-devices.fr>
[rebased and remove buffer timeout]
Signed-off-by: NOctavian Purdila <octavian.purdila@intel.com>
Reviewed-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

Move all core (non-custom) buffer attributes to a vector to make it
easier to add more of them in the future.
Signed-off-by: NOctavian Purdila <octavian.purdila@intel.com>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

There was a need for non triggered software buffer type.  It can be used when
triggered model does not fit and INDIO_BUFFER_HARDWARE causes confusion because
the data stream can be obtained not directly form hardware backend.
Suggested-by: NJonathan Cameron <jic23@kernel.org>
Signed-off-by: NKarol Wrona <k.wrona@samsung.com>
Reviewed-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

Signed-off-by: NOctavian Purdila <octavian.purdila@intel.com>
Reviewed-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

We already do have the length field in the struct iio_buffer which is
expected to be in sync with the current size of the buffer. And currently
all implementations of the get_length callback either return this field or a
constant number.

This patch removes the get_length callback and replaces all occurrences in
the IIO core with directly accessing the length field of the buffer.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

If a buffer implementation does not implement the set_length() callback the
length will be static and can not be changed by userspace. Mark the length
attribute as a read only property in this case so userspace is aware of this
rather than just silently accepting any length value.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

All buffers want at least the length and the enable attribute. Move the
creation of those attributes to the core instead of having to do this in
each individual buffer implementation. This allows us to get rid of some
boiler-plate code.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

The next patch will introduce new dependencies in iio_buffer_alloc_sysfs()
to functions which are currently defined after iio_buffer_alloc_sysfs(). To
avoid forward declarations move both iio_buffer_alloc_sysfs() and
iio_buffer_free_sysfs() after those function.

This is split into two patches one moving the functions and one adding the
dependencies to make review of the actual changes easier.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

Originally device and buffer registration were kept as separate operations
in IIO to allow to register two distinct sets of channels for buffered and
non-buffered operations. This has since already been further restricted and
the channel set registered for the buffer needs to be a subset of the
channel set registered for the device. Additionally the possibility to not
have a raw (or processed) attribute for a channel which was registered for
the device was added a while ago. This means it is possible to not register
any device level attributes for a channel even if it is registered for the
device. Also if a channel's scan_index is set to -1 and the channel is
registered for the buffer it is ignored.

So in summary it means it is possible to register the same channel array for
both the device and the buffer yet still end up with distinctive sets of
channels for both of them. This makes the argument for having to have to
manually register the channels for both the device and the buffer invalid.
Considering that the vast majority of all drivers want to register the same
set of channels for both the buffer and the device it makes sense to move
the buffer registration into the core to avoid some boiler-plate code in the
device driver setup path.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

Individual drivers should not be messing with the scan mask that contains
the list of enabled channels. This is something that is supposed to be
managed by the core.

Now that the last few drivers that used it to configure a default scan mask
have been updated to not do this anymore we can unexport the function.

Note, this patch also requires moving a few functions around so they are all
declared before the first internal user.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Reviewed-by: NDaniel Baluta <daniel.baluta@intel.com>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>

The size of the allocation is currently set to the size of the pointer
rather than the structure we should actually be allocating.
Signed-off-by: NJonathan Cameron <jic23@kernel.org>
Reported-by: kbuild@01.org
Reported-by: NDan Carpenter <dan.carpenter@oracle.com>
Acked-by: NLars-Peter Clausen <lars@metafoo.de>

When copying multiple multiple samples that are adjacent in both the source as
well as the destination buffer, instead of creating a new demux table entry for
each sample just increase the length of the previous entry by the size of the
new sample. This makes the demuxing process slightly more efficient.
Signed-off-by: NLars-Peter Clausen <lars@metafoo.de>
Signed-off-by: NJonathan Cameron <jic23@kernel.org>