Categories

  • Note sure where to post? Ask questions here for direct access to the ModalAI engineering team

    Support Request Format for Best Results Feature Requests 3D Models
    2306
    Topics
    11499
    Posts

    Steven GonciarS

    Hi all, we’re seeing a reproducible BLE scan crash on VOXL2 / Starling 2 with SDK v1.8.06 when BLE 5 extended advertising is present.

    QRB5165 code path: https://gitlab.com/voxl-public/system-image-build/qrb5165-kernel/-/blob/v1.8.06/net/bluetooth/hci_event.c#L5385

    From device captures (see dmesg segments describe below, I don't forum privileges to post the logs), we’re seeing LE Extended Advertising Report (0x0d) events with Num reports up to 6, followed by a kernel NULL deref in the Bluetooth RX path (hci0 hci_rx_work -> process_adv_report -> hci_bdaddr_list_lookup).

    This looks consistent with the extended-adv bounds-check issue in hci_le_ext_adv_report_evt() that upstream now guards here:
      https://github.com/torvalds/linux/blob/master/net/bluetooth/hci_event.c

    Key lines from dmesg:

    Unable to handle kernel NULL pointer dereference at virtual address 0000000000000010 Workqueue: hci0 hci_rx_work pc : hci_bdaddr_list_lookup+0x24/0x60 lr : process_adv_report+0x154/0x408 call trace includes hci_le_meta_evt and hci_event_packet Bluetooth: Unknown advertising packet type: 0x100 Bluetooth: Unknown advertising packet type: 0x4cff

    Have you seen this on QRB5165 / VOXL2 before, and is there already a downstream fix or recommended workaround / SDK version for BLE 5 extended advertising? I pulled in the guards from the latest kernel, testing today.

    Related post:
    https://forum.modalai.com/topic/5181/bluetooth-integration-on-the-voxl2

  • Commonly asked questions we get from the community

    68
    Topics
    297
    Posts

    Eric KatzfeyE

    @JB1 Our voxl-remote-id software with WiFi dongle is intended only for reference and is not meant to be a commercially ready system for our customers who need a remote id solution for their VOXL2 based drone products. Integration, testing, validation, and FAA registration with desired hardware is up to the end user. At a minimum the manufacturer id would need to be changed from ModalAI's id to your companies issued manufacturer id.

  • Questions by Product

    2198
    Topics
    11532
    Posts

    Z

    @Engineer123

    You can download all the necessary flat build files from our page: [modalai.filecloudonline.com](link url). Go into flat builds folder and download the m0054(154-1) flat build file

    With regards to actually flashing the flat build, and using modal-qdl repo:

    Make sure to put the flat build file properly in a correct folder as mentioned by the doc (./qdl/qfil_app/modalai). Make sure the file is unzipped (it should be m0054(154-1)/ all the files, NOT m0054(154-1)/m0054(154-1)/....) Some users have seen a higher qdl success rate when they ran install_dependencies.sh and run.sh with sudo commands. So sudo might work better Confirm that the voxl is in edl mode

  • 499
    Topics
    2456
    Posts

    R

    @Eric-Katzfey thanks. before I did sdk reflash I found out the px4 params were somehow corrupted. I have tried to reproduce these values without success.
    what made it work is taking a similar drawn read the params usign px4-param show -a and writing the to the corrupted one.

  • Please request review from ModalAI before building your drone or custom add-on PCB

    PCB Design Reviews System Architecture Design Reviews
    15
    Topics
    43
    Posts

    Alex KushleyevA

    Hi @Eyal ,

    I just did a quick test and I am getting 2.0ms ping time when voxl2 cpu is in perf mode, and about 2.5ms when in auto mode. VOXL2 is idle.

    the NCM networking feature works concurrently with ADB, which is great and we also have some scripts to set up NAT if you need your voxl2 to actually use the Linux host as a "router" to access outside world. that is easy to set up as well (requires some routing changes on the linux host).

    In terms of real-time streaming, it is definitely adequate for image / video streaming to a Linux host. In fact, this option is my personal favorite way of connecting to Voxl2 while developing / testing cameras and viewing multiple streams in voxl-portal. Voxl-portal uses jpeg compression, which is a lot less efficient than h264/h265 and with the NCM connection, i don't have any issues (compared to adb port forwarding, which also allows voxl-portal to work, but is about 10x slower and uses a lot more CPU, you get frame drops in voxl-portal if required bandwidth is higher than 30-50Mbps or so). I am mainly using the USB-NCM feature for development purposes (as opposed to on a flying drone), but i have not found any limitations, at least for the bench testing use case.

    In terms of CPU load affecting the speed, i just ran a test stress -c 8 which loads up all 8 cores to 100% and the ping times did not change in perf mode (2.0ms) and in auto mode, the ping times reduced from 2.5ms to 2.0ms because the cpu cores are actually running at faster speeds. I am guessing that all the USB stuff is happening in the kernel at a higher priority, so it's not affected by the user applications (unless the cpu overheats, i supposed). GPU / DSP loads should not have any effect on this, since they are separate hardware.

    I also did another test - stream 4 video streams to voxl-portal (live view), which added up to about 100Mbps of network traffic and ping time stayed between 1.5-3.0ms (sometimes lower than 2.0ms).

    I have not observed any stability issues related to NCM being active for long periods of time (days). Make sure you are using a high quality USB-C cable (nothing fancy).

    Alex