-
Notifications
You must be signed in to change notification settings - Fork 89
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
lora: Initial commit of LoRaWAN example #456
Conversation
Any ideas?
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is cool! TTN is great for having an 'out-of-box' working example. If I can find a minute, I'll try to do a similar for Helium at some point (which has much better penetration in North America than TTN).
5e43d89
to
1a6ee32
Compare
CI fixed! |
1a6ee32
to
f118410
Compare
Marking as blocked until we can resolve the GPL question. @tock/core-wg This is something we need to reason through, either on a core team call, or out of band. tl;dr:
|
I don't think anyone is reasonably getting in any trouble here as long as they either don't depend on this library for a distributed program, or license any program that does depend on it under GPL, but the question is reasonable and understanding the answer is probably important. My interpretation is that including it in this manner is fine. Anyone who doesn't use it in their program won't actually link against it, "components" in the libtock-c repo are ultimately licensed individually anyway (primarily under MIT/Apache), and the affects for an actual program compiled/distributed from this repo depend on what is actually distributed and what it depends on. I'd be more concerned if we made any core parts of userland depend on GPL code. If C had a good package management story, we would anyway just be saying that programs should depend on this other library, distributed elsewhere to use it, which would have the same implications and I don't think the question would come up. In this case, we'd just be hosting a copy of that library in the same repo. I don't see a significant difference. Basically my vote is that this is fine. It's only actually a dependency if it's used for a particular application (in which case that application as a whole I believe would need to be distributed under GPL).
I think it's cleaner if the GPL code lives exclusively in a single example, less clean but perfectly good (and more reasonable) if it is only ever used/linked from individual example applications under the examples folder, and a red flag if we start linking against it from dependencies of applications (don't see that happening here).
I think this would be overkill. |
b6c78a8
to
0dea5c4
Compare
OK, looking into this a bit more deeply, two things:
RadioLib is implemented in a way that allows for pluggable hardware implementations via the A trivially portable version of this would implement an instance of Now, I realize that implementing a driver for the SX1262 is non-trivial. It's ~1000 LoC in RadioLib. Now, it's clearly possible to use Tock + libtock-c to build these kinds of applications (as the PR clearly demonstrates). It's perfectly fine to have applications out in the world that don't adhere to the particular layering choice that we advocate upstream, and generally supporting exporting a SPI device (in this case) directly to userspace is good to support for that reason. However, I think the question is: do we want to have these kinds of examples upstream? No doubt we already have a few, but I'd argue those are basically legacy or tests and should be removed (or retained merely as tests). I suggest there are at least two options:
|
I agree, but I'm also torn. It took a year implementing the lr1110 stack using the same SPI/GPIO HAL interface (#458). The examples in that would too only work for one radio chip. The radio lib driver for that is massive (and is also very new and didn't exist when we started). Like OpenThread, I kind of feel like these are necessary evils to make progress on other networking goals. |
Have you read the entire PR? This is specifically not board specific (obviously you need a LoRa radio). The GPIO pin mapping and SPI allocation is handled in the kernel. So this application can run on any board with a SX1261. I agree that this example only supports the SX1261 module. It is easy to change the module in RadioLib to support any module that RadioLib supports, such as the LR1110 for example. It obviously would be great to have the module support in the kernel. But that's yet another thing that needs to be re-written in a Tock way. So until someone goes ahead and does that I think this is a good start and works well. This implementation doesn't stop anyone from writing an implementation in the kernel and migrating to that in the future. It's probably also worth noting that the #458 implementation only works on a single board. So maybe that needs more scrutiny? |
@alistair23 from the code: // now we can create the radio module
// pinout corresponds to the SparkFun LoRa Thing Plus - expLoRaBLE
// NSS pin: 0
// DIO1 pin: 2
// NRST pin: 4
// BUSY pin: 1
I have read the entire PR. Though I'm realizing now, that I had not read the RadioLib Tock adaptation library, which depends on libtock-c, which is a convoluted enough dependency that I admit I don't really understand now what happens with those pins that are passed in.
I agree, and I have not yet really looked at that PR, though one important difference is that #458 adds tests only, while this is moving examples from |
Fair. The comment is trying to describe that this is what I tested with. I did also test with a SX1261 externally attached via SPI, which also works. That does require changes to The code in Tock where the GPIO lines are mapped looks like this let sx1262_gpio = components::gpio::GpioComponent::new(
board_kernel,
LORA_GPIO_DRIVER_NUM,
components::gpio_component_helper!(
apollo3::gpio::GpioPin,
0 => &peripherals.gpio_port[36], // H6 - SX1262 Slave Select
1 => &peripherals.gpio_port[39], // J8 - SX1262 Radio Busy Indicator
2 => &peripherals.gpio_port[40], // J9 - SX1262 Multipurpose digital I/O (DIO1)
3 => &peripherals.gpio_port[47], // H9 - SX1262 Multipurpose digital I/O (DIO3)
4 => &peripherals.gpio_port[44], // J7 - SX1262 Reset
),
)
.finalize(components::gpio_component_static!(apollo3::gpio::GpioPin));
#432 is actually the PR that moves LoRa to examples. The PR backlog is so large that PRs are starting to build on each other :) For clarity, this was originally in examples and moved out because it didn't support RISC-V (which it does now) |
@alistair23 Once #432 is in, can you rebase this? I think my disposition is similar to Brad, i.e., networking stuff is hard and a working example has real value, even if it's a bit less hardware agnostic / purely layered than some idea thing might be, but I'll look at that more in depth early next week. |
0dea5c4
to
ae70ee6
Compare
Rebased, this is just a single commit and can be merged once #432 is |
ae70ee6
to
55bb231
Compare
Okay! #432 is in, but this needs one more quick rebase now it looks like. |
55bb231
to
8aedbf6
Compare
Done |
PR opened: jgromes/RadioLib#1313 It breaks their CI though, as there is a backwards incompatible name change. I have updated this PR to use the new names and not break the CI. So once this is merged I can get jgromes/RadioLib#1313 merged and fix everything up. Let's try not to block this for too long |
2586d93
to
e54c7f0
Compare
Ping! It'd be good to get this in so then I can get jgromes/RadioLib#1313 merged |
@ppannuto hey there! Just my two cents on one of the points raised:
It probably makes more sense to move the Tock HAL file to the tock project itself. The HAL files present in RadioLib started out as examples that turned out to be more easy to maintain as part of the project source. However, unlike the other HALs which are pretty generic (e.g. for Raspberry Pi), the Tock HAL is very specific. So I'm not opposed to moving it, in fact I would support it ;) |
@ppannuto can we maybe move forward with this and kind of iterate our way past the circular dependency? |
I pushed an update that copies over my updated RadlioLib Tock HAL to the libtock-c repo, and has the application use our HAL to build. @alistair23 This is compile-tested only; sorry, I don't have this hardware on-hand. |
OK, I think we just want @alistair23 to confirm this still works. |
Signed-off-by: Alistair Francis <[email protected]>
Bump to the 7.1 release plus a few extra commits to fix a pin disable bug in the Tock HAL. Signed-off-by: Alistair Francis <[email protected]>
Signed-off-by: Alistair Francis <[email protected]>
Still works! |
46438e4
to
2b8e837
Compare
This builds on top of #432 and adds a LoRaWAN example where a Tock system can send data to TTN.