Use adafruit VID & PID about adafruit_nrf52_bootloader HOT 61 CLOSED

ok by the way i made the UF2 bootloader serial .inf driver

adafruit/Adafruit_Windows_Drivers@25f6d8d

i can change it eaily and then sign it for windows 7. for windows 10/mac/linux no driver is needed :)

from adafruit_nrf52_bootloader.

OK we will do it!

from adafruit_nrf52_bootloader.

I actually already added this in the latest Windows Drivers package. Download this one:
https://github.com/adafruit/Adafruit_Windows_Drivers/releases/download/2.3.0/adafruit_drivers_2.3.0.0.exe

"%FEATHER_NRF52840% UF2 Bootloader (0029:00)"=DriverInstall, USB\VID_239A&PID_0029&MI_00
"%FEATHER_NRF52840% UF2 WebUSB dummy (0029:04)"=NullInstall, USB\VID_239A&PID_0029&MI_04
"%FEATHER_NRF52840% Arduino (8029:00)"=DriverInstall, USB\VID_239A&PID_8029&MI_00
"%FEATHER_NRF52840% CircuitPython (802A:00)"=DriverInstall, USB\VID_239A&PID_802A&MI_00

from adafruit_nrf52_bootloader.

ok i understand. thank you for explaining it to me, thach!

i think we should eventually try to figure this out to reduce but for now we can have the CDC only option with 1200 baud touch. hopefully we'll be able to fix it later :)

from adafruit_nrf52_bootloader.

hiya, im back - yep 0x002A is avilabale. lets try this! :)

Feather nRF52840  VID 0x239A    PID 0x0029   # UF2 bootloader (CDC+MSC)
                                PID 0x002A   # Arduino bootloader (CDC)
                                PID 0x8029   # Arduino application mode (CDC+MSC)
                                PID 0x802A   # CircuitPython application mode (CDC+MSC)

from adafruit_nrf52_bootloader.

Thanks, I will update it asap

from adafruit_nrf52_bootloader.

awesome, you're doing great :)

from adafruit_nrf52_bootloader.

Yeah, @microbuilder already told me the VID/PID, though, since the bootloader may still change
the usb interface. It is better to do this last, otherwise it could mess up with windows driver causing instant reset if the incorrect interface with a known VID/PID is plugged.

from adafruit_nrf52_bootloader.

Meanwhile we could use this temporary driver inf for windows here for testing. Though I will try to finialize the interface asap.
WinCDCdriver.inf.zip

PS: @ladyada I couldn't view the master list, it is possibly permission issue

from adafruit_nrf52_bootloader.

ooh ok, yep we can keep it for now - when its ready we can do the switch!
i added you as a collatorator on MBAdafruit repo, please try again :) https://github.com/adafruit/MBAdafruitBoards

from adafruit_nrf52_bootloader.

Ah thanks, I could access it now. I am finalizing the usb interface. I think we should have an extra custom interface although we didn't do much with it now, but it will allow us to have option to send command or communicate with feather without changing pid later. (Dual com port is too confusing as wiced has proved). Will update asap

from adafruit_nrf52_bootloader.

Yes I agree! i think for circuitpython single CDC UART for REPL has worked very well on M0 and M4 boards. It's true it is not great for raw data transfer but not many have complained :)
for M0 we have a lot of endpoints, so we have CDC UART + Mass Storage for disk access + HID for keyboard/mouse emulation. That seems to work for everyone
how many endpoints are there on nrf52840?

from adafruit_nrf52_bootloader.

Nrf52804 has max 8 endpoints in & out.

• Cdc use 2 in, 1 out.
• Msc use 1 in, 1 out
• one of custom will takr 1 in, 1 out

There is still 4 more pairs left, 1 of those can be used for HID in user Arduino sketch. There is plenty left :)

from adafruit_nrf52_bootloader.

ok! i think lets start with just CDC uart, then look to add MSC. Finally we can try HID or other UART later in the future. please let me know if you need help testing or debugging!
we have a few DevKits and got the bootloader loaded easily and can communicate over JLink serial for debugging work :)

from adafruit_nrf52_bootloader.

Usb and msc already work rather reliably, i will push out update that finalize the pid. You could update the driver, signed etc.. So that we could test it easier on windows.

Left work for usb would be uf2 integration and make sure it work reliably across OS.

from adafruit_nrf52_bootloader.

@hathach Just a heads up that Limor has some MSC issues on Windows 7 that are worth looking into. Transfers would fail once or twice before working the third time, etc. Probably just something really minor, but worth trying to reproduce or getting a USB capture from Limor.

from adafruit_nrf52_bootloader.

yep! will do once i get a more comfy setup - ill open up issues with debugging details!

from adafruit_nrf52_bootloader.

That is strange, I tested with windows 10 and it seems to be fine, I will pull out an windows 7 machine to test.

from adafruit_nrf52_bootloader.

I have updated the VID/PID to match the list, the usb is composed of

• cdc
• msc
• custom class ( for whatever purpose we decide to use later, maybe a command channel to factory reset/clear filesystem etc ..)

I think it is decent enough for bootloader

@ladyada please make a windows driver file, that will help other to easily test out the bootloader. Bootloader is update in Arduino repo already, I will update the circuitpython in my folk and submit an PR later along with UF2 implemenation.

from adafruit_nrf52_bootloader.

sure, can you commit your latest version to the adafruit/circuitpython usbboot branch, or give me a hex to load onto this DK, so i can test the windows driver?

from adafruit_nrf52_bootloader.

I just pushed the update bootloader binary to nrf52840 usbboot branch. It should fix your previous issue with failed uploading as well. To flash bootloader please type

make V=1 BOARD=feather52840 bootloader

to upload with usb cdc

make V=1 BOARD=feather52840 SD=s140 all
make V=1 BOARD=feather52840 SD=s140 SERIAL=/dev/ttyACM1 dfu-gen dfu-flash

The cp port for feather52840 has some issue with uart driver ( i guess), you should see the REPL prompt but couldn't type anything yet. We will fix that later after the uf2 implementation :D

from adafruit_nrf52_bootloader.

just update the uf2 bootloader support on both Arduino and CircuitPython repo. Arduino will auto generate uf2 when compilation, for circuitpython we could generate it with uf2 target

make V=1 BOARD=feather52840 SD=s140 all uf2

Note: you need to update uf2conv.py tool to latest to support hex file input, current version in master only support bin file. Simply update with git submodule update --init. I just update readme.md file here

https://github.com/adafruit/circuitpython/tree/nrf52840_usbboot/ports/nrf/boards/feather52840

Note: The REPL prompt is printed but It won't interact, I will fix that later, possibly uart driver issue.

from adafruit_nrf52_bootloader.

That will make thing easier for everyone :) btw, could we map the custom interface ( reserve now ) for libusbk with our signed driver. Currently, I use Zadig for windows binding, Mac/Linux doesn't need that.

from adafruit_nrf52_bootloader.

good question, i will research it - what do you want the custom interface for, is it debug output like we had for WICED?
FYI - for samd21/51 we use 2 endpoinds for HID keyboard/mouse.

from adafruit_nrf52_bootloader.

I am not sure for now, it depends on later need. It can be used as extra comm channel. Maybe issue an factory reset, read memory contents etc... But that custom interface does not need to be part of our signed driver at all. It presents only so that you don't have to make another VID/PID later on.

nrf52840 has 7 pairs of bulk/int endpoints. So far only 4 In, 3 out are used.It is still Btw, all HID can be merged to use a single In & out endpoint.

PS: If you don't really feel the need to include the custom interface, we could remove it. We could always add it later if needed, though you need to make another pair of VID/PID since the usb interfaces are changed.

from adafruit_nrf52_bootloader.

the custom interface is good for now - especially if it helps you with debugging :)
long term we can think about how we want to do debugging on complicated devices like nrf52 with a radio stack - there is no obvious way to do it so i will think about it :)

from adafruit_nrf52_bootloader.

@hathach:

Note: The REPL prompt is printed but It won't interact, I will fix that later, possibly uart driver issue.

I am seeing something like this problem on Linux but not on Windows or Mac, when I use the JLInk serial support. If I connect via a terminal program, I can't type at the board. If I press reset, I see part of the startup prompts, but still can't use it. I think there might be some kind of bug in the jlink support for Linux (or a Linux USB CDC bug of some kind). But the JLinkGDBServer seems to work.

from adafruit_nrf52_bootloader.

That will make thing easier for everyone :) btw, could we map the custom interface ( reserve now ) for libusbk with our signed driver. Currently, I use Zadig for windows binding, Mac/Linux doesn't need that.

Do we need to use libusbk or can we use usbser, so it will just work without installation on Windows 10?

from adafruit_nrf52_bootloader.

@dhalbert the CDC half uses usbser, so that part works on win10 right now. the special interface needs libusb on win7 and maybe win10 because its not a standard windows driver. whee :)

from adafruit_nrf52_bootloader.

@ladyada for debugging we should bring out the jlink and use the segger RT for tracing event, I haven't used much but it looks useful. Maybe when I got more familliar with cp build management, I will try to make a segger studio project for nrf52 to help with debugging :) .

I am thinking the custom interface is for Adafruit-specific command such as an AT variant, to retrieve board information, command nrf e.g send a specific payload to a specific address etc...but we may not need it at all :(.

from adafruit_nrf52_bootloader.

@dhalbert Ah thank you clarification, I am indeed linux user, I did wonder why I haven't got any complains on this from ktwon lol :)) . I did try to troubleshoot it but couldn't find a clue. But we will switch to the usb cdc on cp anyway, maybe it is not worth the effort.

Like Limor says, libusbk is only for the vendor specific (custom) interface. Registering it with libusbk with windows allow us to write a simple c/python code to communicate with it directly as separate channel (from cdc). Dual cdc port is very frustrating as wiced shows :( since the port number is quite random.

from adafruit_nrf52_bootloader.

another thing we can discuss much later is webusb - it is HID-like (and needs libusbk or similar) but then you can connect over a webbrowswer. just thinking outloud, no need to implement right now :)

from adafruit_nrf52_bootloader.

@hathach:

@dhalbert Ah thank you clarification, I am indeed linux user, I did wonder why I haven't got any complains on this from ktwon lol :)) . I did try to troubleshoot it but couldn't find a clue. But we will switch to the usb cdc on cp anyway, maybe it is not worth the effort.

This support case at Nordic seems to describe the same CDC issue on Linux (only): https://devzone.nordicsemi.com/f/nordic-q-a/35376/unable-to-send-data-via-uart-using-nrf52-dk-on-ubuntu/136786. I found this report a little difficult to follow. The original poster was doing simple serial I/O. I'm not sure I have anything to add to this, but could you take a look? It's not clear to me if updating the SDK would help or not.

from adafruit_nrf52_bootloader.

@dhalbert thanks a lot for the link, though since we are moving to the usb cdc of the chip without using the jlink cdc interface, I think I will just skip that for now. We can come back to this later on.

@ladyada oh, webusb is new to me, is it documented here https://wicg.github.io/webusb/ . I will check its requirements to see if we need to prepare anything for it.

from adafruit_nrf52_bootloader.

ok sounds good! don't worry about implementing webusb yet - its just normal 2-endpoint USB but with a special descriptor that Chrome browser can find and communicate with.

from adafruit_nrf52_bootloader.

@ladyada please also add the entry for usber in circuitpython vid/pid as well

PID 0x802A # CircuitPython application mode (CDC+MSC)

from adafruit_nrf52_bootloader.

hooray! @dhalbert can you add and re-spin/re-release?

from adafruit_nrf52_bootloader.

Let's do it for the Arduino application as well :D

summary from the list
Feather nRF52840 VID 0x239A
PID 0x0029 # UF2 bootloader (CDC+MSC) (already added)
PID 0x8029 # Arduino application mode (CDC+MSC)
PID 0x802A # CircuitPython application mode (CDC+MSC)

from adafruit_nrf52_bootloader.

OK, I'll have a PR on circuitpython master shortly.

from adafruit_nrf52_bootloader.

I'm sorry - I was confused. You want this on the Windows Drivers package, right?

from adafruit_nrf52_bootloader.

Fixed by adafruit/Adafruit_Windows_Drivers@4719610

from adafruit_nrf52_bootloader.

@dhalbert is there any problem if the webUSB dummy is not implemented !! Maybe we should add it later, anyway, I will check out to see what is webusb tomorrow.

from adafruit_nrf52_bootloader.

No problem! The "null" dummy device driver for WebUSB is only used if it shows up -- it's not required. Don't spend a lot of time on webusb right now: we experimented with it but there are various problems to solve first.

from adafruit_nrf52_bootloader.

Ok, I got it

from adafruit_nrf52_bootloader.

@ladyada Do you think we should have an additional CDC only for bootloader (without MSC UF2) when using with touch 1200 to enter DFU. Since MSC takes longer to enumerate (~3-4 seconds) and adafruit-util must wait around 4-5 seconds to be safe. and it is kind of annoying when File Explorer pop up for a few seconds everytime Arduino upload a sketch. Some OS may complain with Safe Remove message as well.

from adafruit_nrf52_bootloader.

no, i think we should stick to plain UF2 with MSC+CDC. we use it a lot with all our SAMD21 boards and we've gotten no complaints!

from adafruit_nrf52_bootloader.

@ladyada sorry, I didn't express it well enough. UF2 still show up when we are holding the DFU button or in any other ways we are using. I am talking about an addition mode where the bootloader will show up only as Serial for Arduino usage. Basically, it is done by writing to NRF_POWER->GPREGRET and soft-reset when touch 1200 happens.
https://github.com/adafruit/Adafruit_nRF52_Arduino/blob/master/cores/nRF5/wiring.c#L65

To sum up

• CDC + MSC when holding DFU button or softreset with UF2_MAGIC
• CDC only when softreset with SERIAL_MAGIC and touch 1200
• OTA: when softreset with OTA_MAGIC and or triggered by BLE event

from adafruit_nrf52_bootloader.

lets stick to only two modes:

CDC + MSC when holding DFU button or softreset with UF2_MAGIC
OTA: when softreset with OTA_MAGIC and or triggered by BLE event

we can adjust later if we get complaints but so far nobody complained about M0 and we use the same bootloader with CDC+MSC on both

from adafruit_nrf52_bootloader.

the CDC-only is only entered when we dictate it. Default mode is CDC + UF2, I really don't like to wait an extra 5 seconds with touch 1200 whenever uploading a sketch 😢

from adafruit_nrf52_bootloader.

i dont see why it needs 5 seconds tho? it doesnt with m0 :/ only the first time takes a little time. afterwards its instant

from adafruit_nrf52_bootloader.

I am not enitrely sure, I wait for 1.5 seconds now but it fails occasionally especially when uf2 drive doesn't show up yet while starting uploading. I guess while host is reading the msc uf2 contents (the firmware fiile) that could cause some sort of delay in response when we do dfu via serial. It could be usb stack's bug though. Since we uploading with serial it could be beneficial to just show up as cdc only. This is done by simply supply the host with cdc-only descriptor instead of composite descriptor. All the code remains the same.

from adafruit_nrf52_bootloader.

This delay does not show up on the Atmel USB stacks, so it's probably worth looking at some USB traces, including timing, to see what the delay is. I thought also some people are seeing a significant delay in CPy on nRF52 before the REPL connection is available, and that would be TinyUSB also, I think, since we don't see that delay in Atmel CPy either.

I have been very happy using Wireshark with USBpcap both on Linux and Windows to debug this kind of thing. Beagle is OK but does not parse the non-descriptor packets without paying more money.

from adafruit_nrf52_bootloader.

@dhalbert the first connection with serial is delayed only for CP, I run cdc with several other projects including the bootloader but doesn't encounter that issue. I probably didn't get it integrated well with CP yet. It is a different issue though.But since it is purely application code, we could always fix it later.

For the tracer, I got a really good hardware analyzer given by KTown. Under the analyzer I could only see the device NAK the IN token from host, the usb stack is not perfect I admit. The nrf5x port is only run as proof of concept just a few weeks ago, the pre-production IC is full of bug that making the sw work before that is ~ impossible.

from adafruit_nrf52_bootloader.

I just pushed the touch option for adafruit-nrfutil
adafruit/Adafruit_nRF52_nrfutil@18fa21a#diff-eeb4dae138b55a43848ee839b4e10a96R107

Our scenario is adafruit-util wiill open and disconnect circuitpython on nrf52840 at 1200 bps, cp will reset to DFU mode. TOUCH_RESET_WAIT_TIME is time for util to wait for nrf52 is enumerated as bootloader. Currently it is 1.5 second, but not enough. Often than not I see the dfu process start before the file explorer could show up the thumbdrive. Sometimes the dfu process would failed.What is my mind is

1. increase time delay
2. reduce delay time by report to host as cdc only, msc take lots more time to enumerate
3. try to hunt down the possible cause, which could take a bit of time :(

from adafruit_nrf52_bootloader.

@ladyada Yeah it is confusing, and I didn't express it very well. I admit I have issue with expression due to the remote work and lack of communication with human :( .

I will also need your help to have an extra VID/PID for the bootloader with CDC only. Like you said, later on, if we could hunt down the issue and the delay time is under 2 seconds, we could remove that mode. User won't ever notice it.

PS: Thinking about it, it is more and more the usb stack issue since we having lot of data transfer in both interfaces in a really short time. Some conflict must occur. I will try to trace it, but we really need a working thing with Arduino. As mentioned before, Arduino repo use freeRTOS, I haven't tested usb stack with freeeRTOS for ages, and the last time I test is with a very old version :(.

from adafruit_nrf52_bootloader.

PS: we have another pending issue with bootloader in OTA as well 😢 . The same code doesn't run with SD 6.1.0 while it runs with 6.0.0 despite release note says they are binary compatible
https://devzone.nordicsemi.com/f/nordic-q-a/37473/advertising-does-not-start-with-s140-6-1-0-same-hex-run-without-issue-with-s140-6-0-0

Have been tracing it in the last couple of days, at worst we will release it with 6.0.0 first.

from adafruit_nrf52_bootloader.

@ladyada is it OK that I use the PID 0x002A for the bootloader with CDC only. Any number will do though.

from adafruit_nrf52_bootloader.

Thanks, I am hoping to nail it down for other to test. Bootloader is the only thing we need to get it right in the first time :)

from adafruit_nrf52_bootloader.

changing to CDC only mode, I could get 10 out of 10 consecutive uploading. 1.5 second is probably long enough for linux. I will test with mac and windows as well. Please update the windows driver exe whenever you feel convenient :) .I am cleaning up and pushing it out :D

from adafruit_nrf52_bootloader.

ok sounds good!

from adafruit_nrf52_bootloader.

@ladyada can you update the driver exe with the new PID for Arduino bootloader, we need it to test with windows uploading

from adafruit_nrf52_bootloader.