Move some logic from Subscriber into uORBManager. This reduces calls from the
modules to the uORB manager, improving performance in protected build.
Signed-off-by: Jukka Laitinen <jukkax@ssrc.tii.ae>
It was a mistake to mix these two together, it is simpler to implement the boardctl interface
for the protected build, if the boardctl ioctls are different
Signed-off-by: Jukka Laitinen <jukkax@ssrc.tii.ae>
When building uORB for NuttX flat build, or for some other target, everything
works as before.
When building uORB for NuttX protected or kernel build, this does the following:
- The kernel side uORB library reigsters a boardctl handler for calls from userspace
and services the boardctl_ioctls by calling the actual uORB functions
- For user mode binaries, the uORBManager acts as a proxy, making boardctl_ioctl calls to the
kernel side
Signed-off-by: Jukka Laitinen <jukkax@ssrc.tii.ae>
This is the case for boards with digital readout, like v5x, but still
enable the battery_status module for external analog driver options.
An alternative would be to not run battery_status depending on config.
Since uORB is split into kernel and userspace parts, it is no longer possible to just
link uORB into px4_platform, which is used in both kernel and user side.
Signed-off-by: Jukka Laitinen <jukkax@ssrc.tii.ae>
- Allocate and free the node name in uORBDeviceNode.
- Add protected build support by de-allocating the name with kmm_free, when
running in kernel side. strdup allocates from the kernel heap in NuttX kernel
space.
- Remove the CDev::unregister_driver_and_memory(), it is no longer used
Signed-off-by: Jukka Laitinen <jukkax@ssrc.tii.ae>
In NuttX protected build there are separate work queues in kernel and user sides.
pthreads are only available in user side, so use tasks and kthreads for
memory protected builds.
Signed-off-by: Jukka Laitinen <jukkax@ssrc.tii.ae>
This is more logical place for the "backend" implementation than
directly under platform.
This also allows making other implementations as "real" drivers, as well as proper configuration via Kconfigs
Signed-off-by: Jukka Laitinen <jukkax@ssrc.tii.ae>
Don't compare pointers to metadata, but the metadata contents.
In protected/kernel build there are two sets of metadata, on on kernel
side and another in user side. Thus the comparison of pointers would just
always fail. Compare orb_id instead
Signed-off-by: Jukka Laitinen <jukkax@ssrc.tii.ae>
Proxy all calls to the DeviceNode through Manager;
- This hides the DeviceNode from publishers and subscribers
- Manager can be made an interface class between user and kernel spaces in protected build
- This doesn't increase code size or harm the performance on flat build
Signed-off-by: Jukka Laitinen <jukkax@ssrc.tii.ae>
Make possible to define chip-select pin for internal SPI.
By defining chip-select pin, it's possible to start specific SPI device only.
This allows to have several same type of sensors on the same bus with different orientation.
Add common functions, implemented for nuttx, and link to architecture specific libraries
Make a separate library to wrap nuttx random number generator as "os_random".
Signed-off-by: Jukka Laitinen <jukkax@ssrc.tii.ae>
Using mixers on the IO side had a remote benefit of being able to
override all control surfaces with a radio remote on a fixed wing.
This ended up not being used that much and since the original design
10 years ago (2011) we have been able to convince ourselves that the
overall system stability is at a level where this marginal benefit,
which is not present on multicopters, is not worth the hazzle.
Co-authored-by: Beat Küng <beat-kueng@gmx.net>
Co-authored-by: Daniel Agar <daniel@agar.ca>
When the last WorkItem is deleted, it is removed from a work queue and the
queue is being stopped. But, the queue itself might get deleted in the middle,
in a higher priority thread than where the WorkItem deletion was performed from
If the WorkQueue::Detach accesses the member variables after this, there is memory
corruption
This happens in particular when launching i2c or spi devices in
I2CSPIDriverBase::module_start:
- The "initializer" is deleted when the instance is not found and the iterator
while loop continues.
- The workqueue is deleted in the middle of "initializer" deletion when the
WorkQueueRunner returns.
This prevents deletion of the WorkQueue before the Detach has been finished,
in the specific case that the ::Detach triggers the deletion
Signed-off-by: Jukka Laitinen <jukkax@ssrc.tii.ae>
There's an increasing amount of slow logged topics at 1-2Hz, which were all
updated in the same logger iteration, leading to data bursts. For log
streaming this started to exceed uart buffer sizes. By distributing updates
more equal over time those bursts are removed, reducing buffer size
requirements.
Tests showed during steady state a reduction of maximum topic updates per
iteration from 40 down to 17.
Also the SD log buffer fill level is more constant.