Skycontroller `nap` image: hardware and identity (static analysis)
Static read of the Skycontroller
nap-linux-2.0.0-rc3image by one agent, 2026-10-04. Status: Agent-cited, with these claims re-derived by the commissioning session: the whole device-type name table inusr/lib/libarsdk.so(functionarsdk_device_type_str, all 23 ids decoded from its jump table: 0x901 BEBOP, 0x903 SKYCTRL, 0x90c BEBOP_2, 0x90f SKYCTRL_2, 0x913 SKYCTRL_NG, 0x915 SKYCTRL_2P and the rest), and the update-file table in the same library (id paired with filename pointer: 0x903nap_update.plf, 0x90fmpp_update.plf, 0x915mpp2_update.tar, plus 0x901, 0x90c and 0x914). The 0x901 and 0x90c rows agree with the Bebop 1 and Bebop 2 images’ own model ids, which is the control. Not re-derived here: the kernel, pin map, peripheral and xmega sections.What it means, with the live unit: see
skycontroller-live-vs-image.md. Short form: the image announces 0x913 SKYCTRL_NG and ships asnap_update.plf; the live Skycontroller 2 reports hostnamempp, modelSkycontroller 2and themppgamepad id, and the same library pairsmpp_update.plfwith 0x90f SKYCTRL_2. The nap image is therefore not the Skycontroller 2’s firmware. The “recollection” labelled R below is the agent’s memory, not evidence.
Image: firmware/skycontroller_nap_update.plf, build nap-linux-2.0.0-rc3, extracted at firmware/extracted/skycontroller-nap/. All paths below are relative to rootfs/ unless they start with partitions/ or metadata.json. Nothing was connected to; no firmware secrets are quoted. Tags: V = Verified (bytes or listing read), A = Agent-cited (earlier note), I = Inference, O = Open, R = Recollection.
Scratch artifacts (not in the repo): scratchpad/nap-kernel.bin (decompressed kernel, 6,401,540 bytes), nap-kernel.elf (same bytes wrapped for llvm-objdump, VMA 0 so offsets equal file offsets), nk.txt (strings), xmega.bin (flat image of lib/firmware/xmega.hex).
Summary
Section titled “Summary”- SoC is a Parrot7 (“P7”) part on Linux 3.4.11+ (SMP PREEMPT, built Wed Aug 9 11:18:30 CEST 2017, gcc 4.6.3 Sourcery 2012.03-57, jenkins job
nap-release). The board is a machine calledNAP(mach nr 0x2703), added as a variant inside Parrot’s FC7100 workbench board code. The kernel is a gzip payload inside a nested PLF in outer section 2, loaded at 0x82000000. (V) - This kernel has no
/sys/kernel/hsistable, no Milos or Mykonos3 strings. Its pin map is a plain user-GPIO export list in the board init code, which I decoded: LED lines on gpio 55 to 58, bottom-board update line gpio 150, HDMI reset 154, Wi-Fi reset 148, USB reset 179, 11 board-revision strap pins. (V) - Peripherals: TI WL18xx Wi-Fi on SDIO (used as the AP,
wlan1), Atheros AR6004 hw3.0 Wi-Fi on USB (the STA link to the drone, which wifid callsqca9375), MPU6050 IMU at i2c 0x68, AK8963 magnetometer at i2c 0x0c, Silicon Image SiI HDMI transmitter at i2c 0x3b, Mali-400 GPU and a hardware video decoder, TI shared-transport Bluetooth/GPS path onttyPA1, GPS protocolai2.bcm43526appears only in the build manifest andusb.ids, not as hardware. (V) - Bottom board is an Atmel ATxmega32C4 flashed over Atmel DFU (03eb:2fd9) from
lib/firmware/xmega.hex: 13,122 bytes of program starting at 0x0, no version string. It enumerates as USB 03eb:2407 “NAP GAMEPAD”, HID report 1 = 24 buttons plus 8 axes of signed 8 bit (-100 to 100), report 2 = 16 byte vendor channel. (V) - Identity: the image calls itself
Skycontroller, model id0x0913, USB pid 6860, updater filenap_update.plf, default SSID prefixnap-ng-. Its ownlibarsdk.sodecodes 0x0913 asSKYCTRL_NG, which is a different enum entry fromSKYCTRL_2(0x090f). The decode is validated against the repo’s Bebop images (0x0901 and 0x090c). (V) - The same library’s update table maps
nap_update.plfto device type 0x0903SKYCTRL,mpp_update.plfto 0x090fSKYCTRL_2,mpp2_update.tarto 0x0915SKYCTRL_2P. So in ARSDK’s naming this image is not the update file of the entry called SKYCTRL_2. (V for the table, I for what it means) - Hardware features plus a script line about wiping Android files when updating from 1.x.x suggest the
naphardware line had an earlier Android-based 1.x software. That points at the original Skycontroller hardware family running a new “NG” firmware, but the image never says “Skycontroller 2” and nothing in it proves which retail product Ryan’s unit is. (I, O) - Things in the image that do separate controller variants:
ro.factory.typeisbebopormilos, chosen from/customer_prod/type.txtor, when that file is missing, from whether the serial starts with a fixed 10 character prefix (milos if it does). The mapping folders differ on the emergency button. The image ships mpp-family and mpp2-family udev rules and source-tree entries for hardware it does not carry. (V)
Identity evidence table
Section titled “Identity evidence table”| Evidence | Where | Says | Tag |
|---|---|---|---|
ro.product.model "Skycontroller", ro.product.model.id "0x0913", ro.updater.filename "nap_update.plf", ro.product.usb.pid 6860 with comment “USB pid from SkyController” |
etc/boxinit.rc lines 20 to 32 |
Product string has no generation. Pid 6860 decimal is 0x1ACC (arithmetic). | V |
ro.parrot.build.project=nap, uid=nap-linux-2.0.0-rc3, ro.build.date Wed Aug 9 11:09:49 CEST 2017, ro.debuggable=1 |
etc/build.prop, etc/default.nap.prop |
Release candidate, debug flavour. Same builder hostname as the kernel banner (jenkins@977af7249957). |
V |
Device-type name table in libarsdk.so, function arsdk_device_type_str at 0x1fc08 (Thumb, tbb switch on id minus 0x900) |
usr/lib/libarsdk.so |
0x901 BEBOP, 0x903 SKYCTRL, 0x90c BEBOP_2, 0x90f SKYCTRL_2, 0x911 PAROS, 0x913 SKYCTRL_NG, 0x914 ANAFI4K, 0x915 SKYCTRL_2P, 0x916 CHIMERA (also RS, JS, EVO, RS3, POWERUP, EVINRUDE, WINGX; 0x904, 0x908, 0x912 give UNKNOWN) | V |
| Validation of that decode | bebop-2-4.7.1/rootfs/etc/default.mykonos3board.prop has ro.product.model.id=0901, default.milosboard.prop has 090c |
Matches BEBOP and BEBOP_2 above, so the table reading is right. | V |
Update-file table at literal base 0x553ec in libarsdk.so (18 entries of id and filename pointer) |
usr/lib/libarsdk.so |
0x901 bebopdrone_update.plf, 0x90c bebop2_update.plf, 0x911 paros, 0x914 anafi4k_update.tar, 0x916 chimera, 0x903 nap_update.plf, 0x90f mpp_update.plf, 0x915 mpp2_update.tar. 0x913 has no entry. | V |
Media folder names SkyController/ and SkyController_2/ |
usr/lib/libarsdk.so strings at 0x3a3e5 and 0x3a3f4 |
The SDK treats the two as separate products. | V |
Default SSID built as nap-ng- plus a suffix |
usr/bin/nap_factory_reader (rodata 0x11177, format %s%s) |
“NG” is in the image’s own naming for this firmware. | V |
ro.factory.type logic: read /customer_prod/type.txt; if absent (errno 2), read serial; if serial is longer than 9 chars and its first 10 chars equal a fixed product-prefix string, type is milos; otherwise bebop |
usr/bin/nap_factory_reader read_type at Thumb 0x10b3e to 0x10c36 |
Selects etc/mppd/{bebop,milos}. Both names also appear in usr/lib/libpinst.so machine tables (“bebop”, “Milos board”), and milosboard is the Bebop 2 board in the Bebop 2 image. I read the strncmp operand as 10 bytes at rodata 0x11143. |
V (logic), I (meaning of type) |
bebop vs milos mapping folders |
etc/mppd/ |
Emergency bound to button Z only in bebop (A, from SKYCONTROLLER_NOTES.md; not re-read) |
A |
udev: input name mpp2_gamepad, input name qpnp_pon (typed mpp2_power), USB 03eb:2406 (mpp_gamepad, mpp_bb), USB 03eb:2407 (nap_gamepad, nap_bb) |
lib/udev/rules.d/55-juba-gamepad.rules, 56-juba-bb.rules |
The software supports three hardware families, only nap hardware is described elsewhere in the image. libmpphw.so knows the types mpp_bb, nap_bb, mpp2_gamepad. Only libmppd_hw_nap.so ships. |
V |
Build manifest lists products mpp, mpp2, nap, packages atmel-mpp, atmel-mpp2, atmel-nap (path aftermarket/NAP/Atmel), linux-msm8909, lk-msm8909, qcom-charger, qcacld-2.0, qca6174-firmware, bcm43526, network-monitor-bcm43526b, and also linux-p7 |
etc/manifest.xml (100 projects) |
It is one manifest for a source tree serving several products, so a name there is not evidence of hardware in this image. The kernel actually built is linux-p7 (path strings in the kernel). |
V |
boxinit.rc comment: “When updating from 1.x.x, wipe the data partition to remove android-related files”, with factory_reset android |
etc/boxinit.rc lines 74 to 76 |
Earlier 1.x software of this hardware was Android-based. Kernel also carries android_usb adb, mtp and accessory gadget strings. | V |
libpinst.so machine-name table (Mykonos, Mykonos3 board, Milos board, Mpp board, FC7100 Workbench, NAP, Chimera board, plus mpp2, anafi, anafi4k) |
usr/lib/libpinst.so strings 0x3cce to 0x4042 |
Generic Parrot platform list, NAP and Mpp board are separate entries. |
V |
hardware_compatibility 183847359 (0x0AF549BF); inner kernel PLF has 0x31F8B344 |
metadata.json; header of partitions/section-00002-type-3.bin |
Opaque in the image. Nothing in the rootfs decodes it (libpinst only has the string “hardware”). | V (values), O (meaning) |
ro.product.model.id is also read by mppd and avabye |
strings in usr/bin/mppd, usr/bin/avabye |
Likely how the controller announces itself to the app. Not traced. | I |
The question: does anything separate a Skycontroller 2 from an original Skycontroller?
Section titled “The question: does anything separate a Skycontroller 2 from an original Skycontroller?”What the image gives, in order of weight:
- The SDK table inside the image has separate entries SKYCTRL (0x903), SKYCTRL_2 (0x90f), SKYCTRL_NG (0x913), SKYCTRL_2P (0x915). This image announces 0x913 and ships as
nap_update.plf, which the same library files under 0x903. So by the image’s own tables it is neither the “mpp_update.plf” product (0x90f SKYCTRL_2) nor the “mpp2_update.tar” product (0x915). (V) - The hardware in this image is Parrot7 with HDMI out, GPS, an IMU and compass, a TI Wi-Fi AP chip and an Atheros USB Wi-Fi client. The sibling families named in the same tree (mpp2) bring a Qualcomm power-key input (
qpnp_pon) and Qualcomm kernel and Wi-Fi source projects. That Qualcomm side is absent from this rootfs. (V that it is absent, I that it describes another product) - The Android-wipe line shows the nap family shipped Android software before 2.0.0. (V)
- R (Recollection, not evidence): I believe the original Skycontroller ran Android and had an HDMI output, and that Parrot’s SDK called the Bebop 2 pack controller “Skycontroller 2”. If true, the nap hardware would be the original, and “NG” would be the 2.0.0 firmware generation. What would confirm it: any string mapping 0x0913,
SKYCTRL_NGornapto a marketing name (the image has none), or the model plate and the firmware version read from Ryan’s unit (ro.product.model.idin its properties).
Net: the image does not say “Skycontroller 2” anywhere, and it contradicts the assumption that nap equals SKYCTRL_2 in ARSDK naming. The earlier note label “Skycontroller 2” for this image rests on inference from services and should be treated as unproven. (I)
One more split that does exist in the image: the bebop versus milos factory type and its mapping folders. That is a factory pairing or SKU switch inside this one firmware, not a generation switch. Which one Ryan’s unit uses is not in the image (/customer_prod is empty). (V, O)
SoC and kernel
Section titled “SoC and kernel”| Item | Finding | Tag |
|---|---|---|
| Packaging | Outer PLF section 2 (type 3, 3,237,088 bytes) is itself a PLF: magic PLF!, header version 13, file_type 1 (outer is 2), entry point 0x82000000, platform 6, application 103, one section (type 0, 3,236,704 bytes, load 0x82000000, stored uncompressed). That section is an ARM self-extracting kernel: gzip stream at inner file offset 0x5b38 (1f 8b 08), decompressing to 6,401,540 bytes with a 361 byte unused tail. Same layout the Bebop 2 notes (11h) describe, gzip offset differs (0x5ae8 there). |
V |
| Version | Linux version 3.4.11+ (jenkins@977af7249957) (gcc version 4.6.3 (Sourcery CodeBench Lite 2012.03-57)) #1 SMP PREEMPT Wed Aug 9 11:18:30 CEST 2017. Module vermagic 3.4.11+ SMP preempt mod_unload modversions ARMv7. Rootfs modules at lib/modules/3.4.11+/. |
V |
| Tree | Source paths /data/jenkins/jobs/nap-release/workspace/packages/linux-p7/arch/arm/mach-parrot7/{core,devices,pinctrl,uart,p7_pwm,p7mu,sdhci,usb,vdec,wl18xx,board-common,board-sysfs,board-fc7100-workbench,board-fc7100-workbench-variant,fc7100-module,fc7100-mezzs,...}.c. No board-nap.c name is present. |
V |
| SoC | Messages p7: not a Parrot7 chip, p7: found Parrot7 chip v%d [0x%08x], p7: found invalid Parrot7 chip revision; Parrot7 GPU clock messages; p7mu power unit; p7-temperature iio device (read by libmpphw.so for P7 and P7MU temperatures); console UART is the Parrot serial driver, port ttyPA3 (ro.boot.console=ttyPA3). Exact P7 revision and core model are runtime-detected: O. |
V, O |
| Machine descriptors | Two descriptors found by pointer: FC7100 Workbench (nr 0x270d) at VMA 0xc05c4ec8, and NAP (nr 0x2703) at 0xc05c4f0c. Kernel VMA = file offset + 0xc0008000 (checked by 1,432 string-pointer matches). _PARROT_FC7100 and _PARROT_NAP appear as symbol-like strings. No device tree: this is a 3.4 board-file kernel. |
V |
| Board identification | NAP hw 0x%08x and nap mb rev: the init code reads 11 strap GPIOs (table at VMA 0xc0410380: 183, 184, 185, 186, 187, 189, 191, 203, 204, 205, 206) and logs the result as the board hardware revision. |
V (table read), I (that they form the revision) |
| Not present | hsis, Mykonos3, Milos, Evinrude, mk3_hsis: zero matches in the decompressed kernel. Control: with the same /usr/bin/grep -a call, Linux version, user_gpio, ATMEG_UPDATE, p7mu all match. Rootfs etc/boxinit.rc line 33 still reads /sys/kernel/hsis/pcbrev with default 00.00 for the USB bcdDevice, so on this kernel that falls back to 00.00. |
V |
| Boot console | ro.boot.console=ttyPA3, ro.debuggable=1, adb over TCP port 9050 property set; adbd also runs a USB gadget (etc/boxinit.d/50-adb.rc, VID 19cf, pid from ro.product.usb.pid, function adb). |
V |
| Other kernel facts | Built-in: Android USB gadget (adb, mtp, accessory), TI shared transport (stc, stll, hcill) with a GPS channel driver gpsdrv, sdhci, ramoops, ION. FC7100 workbench strings (maXTouch, LVDS serializer, wau8822 audio) belong to the workbench code in the same kernel, not to the controller. |
V |
Pin map (this kernel, decoded)
Section titled “Pin map (this kernel, decoded)”There is no /sys/kernel/hsis here. The numbers come from the user-GPIO export function in the NAP board init, file offset 0x5a9890 to 0x5a9ad8 (VMA 0xc05b1890 to 0xc05b1ad8). Each call is bl 0xc001df84(gpio, mode, name), where the name pointer is a string at VMA 0xc04ffe88 and up. The log strings Export user in GPIO %d with alias '%s' and the sysfs path /sys/devices/platform/user_gpio/<NAME>/value used by bottom_board_update show these become user_gpio nodes. The second argument is 0 or 2 in the calls; I did not decode its meaning (O).
| user_gpio name | GPIO | arg | Evidence it matches use | Tag |
|---|---|---|---|---|
| LED_CLK | 55 | 2 | libnapbb.so drives /sys/class/gpio/gpio55..58 (A) |
V, A |
| LED_LATCH | 56 | 2 | same | V, A |
| LED_BLANK | 57 | 2 | same | V, A |
| LED_SIN | 58 | 2 | same. Four lines named clock, latch, blank, serial in: a bit-banged serial LED driver. | V, I |
| ATMEG_UPDATE | 150 | 2 | bottom_board_update writes 0 here after flashing |
V |
| hdmi-nrst | 154 | 0 | kernel logs Resetting HDMI chip |
V |
| wifi-rst | 148 | 0 | name only | V |
| wifi-wake-up | 138 | 0 | name only | V |
| namp-pwdn | 151 | 2 | name only. The “namp” prefix is unexplained (audio amplifier power-down is my guess). | V, O |
| ATMEG_GPIO_A, B, C, D | 134, 135, 136, 137 | 0 | four lines to the xmega, role unknown | V, O |
| usb-rst | 179 | 0 | name only | V |
| (unnamed in this list) | 140 | 32 | configured by a separate helper (bl 0xc0023bbc, args 140 and 32); 140 is also the i2c irq field of the HDMI transmitter |
V, I |
I2C board table (i2c_board_info, 40 byte entries at VMA 0xc05c8564): mpu6050 addr 0x68 irq 146, siihdmi addr 0x3b irq 140, ak8963 addr 0x0c irq 142. Whether the irq values are GPIO numbers or Linux irq numbers is not decided (O). A mounting matrix string 0, -1, 0; -1, 0, 0; 0, 0, -1; sits right after the machine name at 0xc04fff40 (file 0x4f7f40), referenced by one pointer in data; which sensor it belongs to I did not trace (O).
The Bebop 2 hsis numbers are not valid here. For example, in the Bebop 2 table GPIO 199 is reset_psoc and 9 is reset_wifi; in this kernel 199 is not exported and wifi reset is 148.
Peripherals
Section titled “Peripherals”| Peripheral | What the image shows | Tag |
|---|---|---|
| Wi-Fi, drone link (STA) | etc/boxinit.d/50-wifi.rc: wifid --mode STA --ip 192.168.42.3 --dhclient, WIFID_CFG80211_CHIP=qca9375. usr/lib/wifid/cfg80211driver.so strings for qca9375 load module ath6kl_usb from /lib/modules/ath6k/ and use firmware dir /lib/firmware/ath6k/AR6004/hw3.0 (fw.ram.bin link to fw.ram.080717.bin, per-region bdata*.bin, board data and MAC softmac*.bin are links into /tmp generated at runtime). ath6kl_usb.ko describes “Atheros AR600x USB devices”. So: Atheros AR6004 over USB, labelled qca9375 in wifid. No Broadcom driver or firmware in the rootfs (control: the same recursive /usr/bin/grep -rIl -a -i finds ath6k in lib/modules/ath6k/ath6kl_usb.ko and usr/lib/wifid/cfg80211driver.so, and finds bcm43 only in etc/manifest.xml and usr/share/usb.ids). |
V |
| Wi-Fi, AP | wifid --mode AP --ip 192.168.44.1 --suffix ap, chip wl18xx, interface renamed to wlan1 by 60-wl18xx.rules for the SDIO device under acs3-sdhci.0/mmc0:0001:2. Firmware lib/firmware/ti-connectivity/wl18xx-fw-4.bin, wl18xx-conf.bin, wl1271-nvs.bin. modules.alias: sdio:c*v0097d4076* maps to wlcore_sdio (TI, 0x4076). Started only when gps.socket.ready=1. |
V |
| Bluetooth and GPS path | btgpsdemux opens /dev/ttyPA1, exposes abstract sockets @/tmp/bt.socket and @/tmp/gps.socket, sets property gps.socket.ready, uses a user_gpio named bt-rst. Kernel has TI shared transport with a GPS channel driver. That suggests a TI combo chip on the WL18xx side sharing one UART between Bluetooth and GPS. |
V, I |
| GPS | etc/gps_config.txt is two lines: $protocol=ai2 and $tty=@/tmp/gps.socket. usr/lib/libgps.so (multi-vendor: ai2, sirf, ubx, furuno parsers, NMEA) has send_set_ai2_protocol and parse_ai2_msg. etc/gps_firmware.bin is 44,341 bytes: magic bytes 07 ab cd 00, a short header, about 0x48 bytes of repeating ASCII filler 45NMGNSS, then Thumb-2 looking code with 0x001xxxxx and 0x002e4000 addresses; entropy about 7.0 bits/byte after the header. The format does not name a vendor and I cannot identify the chip family from the file alone. gpsd looks for /etc/gps/<ro.hardware>/gps_config.txt and gps_firmware.bin, but the image has the files directly in /etc and no etc/gps/ directory, and no file sets ro.hardware. |
V, O |
| IMU | ro.sensors.imu.name "mpu6050", trigger mpu6050-dev0, kernel i2c entry mpu6050 at 0x68, modules mpu6050.ko, mpu6050_i2c.ko, mpu6050_accel_gyro.ko. |
V |
| Magnetometer | ro.sensors.magneto.name "ak8963", kernel i2c entry ak8963 at 0x0c, driver akm8975.ko (“AKM8975 compass driver”, rejects non-AK8975 ids in its message; the board table says ak8963, so that driver is used for the AK8963 entry). |
V |
| Sensor daemon | sensorsd-mpp (libiio, Eigen) as in the earlier notes. |
A |
| HDMI | Kernel has sii HDMI 9024A tx, hdmi-nrst, switch video-hotplug; i2c siihdmi at 0x3b. usr/bin/hdmi_hotplug (udev rule 99-hdmi.rules on switch/hdmi) loads ump, p7ump, hx_uio, hx270_vdec, mali, sets hdmi_plugged, starts hudd (Qt5 eglfs); on unplug stops hudd and unloads the drivers. Modules: mali.ko (Mali-400 path parrot/gpu/mali400), hx270-vdec.ko. |
V |
| USB device gadget | etc/boxinit.d/50-adb.rc: iManufacturer “Parrot”, idVendor 19cf, idProduct 6860, iProduct = model plus build version, bcdDevice from /sys/kernel/hsis/pcbrev (falls back to 00.00 on this kernel), function adb only. |
V |
| USB host and phone link | aoad (libusb) probes attached devices for Android Open Accessory protocol and talks to the phone; usbnet udev rule handles asix, cdc_ether, stmmaceth adapters; rndis_host.ko present; usb-rst gpio 179. |
V |
| Audio, battery, charger, display panel, buzzer | Not found as components. A namp-pwdn line exists. |
O |
Bottom board
Section titled “Bottom board”| Item | Finding | Tag |
|---|---|---|
| MCU | ATxmega32C4. usr/bin/bottom_board_update runs dfu-programmer atxmega32c4 flash /lib/firmware/xmega.hex --force --validate-first --ignore-outside, writes 0 to ATMEG_UPDATE, then dfu-programmer atxmega32c4 launch. The bundled dfu-programmer is 0.7.2; its target table row for atxmega32c4 carries vendor/product word 0x03eb2fd9 and flash size 0x8000 (32 KB), which matches the udev trigger 52-bottomboard.rules (Atmel DFU 03eb:2fd9). |
V |
| Trigger and gating | udev runs the updater when 03eb:2fd9 enumerates. The script’s first action is sprop mpp.services.start 1, and etc/boxinit.rc starts class mpp (mppd, aoad, hudd) only on that property. So the main app waits for the bottom board script. Whether the xmega is held in DFU mode on every boot, so --validate-first skips the write when the image matches, is my reading (I). |
V, I |
| Firmware file | lib/firmware/xmega.hex: Intel HEX, 36,930 bytes of text, 13,122 bytes of data in one contiguous run 0x0000 to 0x3341, no gaps, no extended address records. Starts with an interrupt vector table of RJMP words. No readable version string, no build date; so I cannot give a firmware version (O). The package is aftermarket/NAP/Atmel at revision d7aadb84... (manifest, so source tree state, not a version). |
V |
| USB identity it presents | Device descriptor at flash 0x32da: bcdUSB 2.00, class 0, ep0 packet 8, VID 0x03eb, PID 0x2407, bcdDevice 0x0100, 1 configuration. Strings stored in flash: NAP GAMEPAD and www.parrot.com. The PID matches the udev nap_gamepad and nap_bb rules; PID 0x2406 is the mpp variant the same rules also handle. |
V |
| HID report descriptor | At flash 0x324c. Report ID 1: Generic Desktop Joystick collection, 24 buttons (usage page 9, 1 to 24, 1 bit each), 8 axes with usages 0x30 to 0x37, 8 bits each, logical -100 to +100. Report ID 2: vendor page 0xFF00, 16 byte input and 16 byte output (this is presumably the hidraw channel libnapbb uses for LEDs, calibration, thresholds; I). This settles one open stick-path question: the stick range on the wire is signed 8 bit in -100..100, so the signed-char scaling in mppd is not a rescale (V for the descriptor, I for the mppd consequence; the evdev range seen by libevdev was not read). |
V, I |
| Attached lines | GPIO 134 to 137 ATMEG_GPIO_A..D, GPIO 55 to 58 LED serial lines, GPIO 150 ATMEG_UPDATE, all from the pin map above. |
V |
| Calibration | Not in the firmware file as far as I looked (no attempt to decode the code). Stick centres and ranges are properties ro.factory.js.centers and ro.factory.js.min_max per the earlier notes; still no setter found. |
A |
What was not examined
Section titled “What was not examined”- Disassembly of the xmega code itself (AVR), so no statement about what it does with the
ATMEG_GPIO_*lines or its protocol beyond the descriptors. libnapbb.so,libmppd_hw_nap.so,mppdandsensorsd-mppinternals;hudd,libhud, Qt, fonts.- The meaning of the second argument to the user-GPIO export call, the helper at 0xc0023bbc, and the other board init pieces outside the user-GPIO block (sdhci and pinctrl tables, UART and USB setup, PWM). I only decoded the one function whose pointers I could resolve.
- The p7 pinctrl table (pad muxing) and the board revision decode from the 11 strap pins.
- The bootloader and what the flashed partitions are; this image has only the rootfs and the kernel section. The
/factoryand/customer_prodUBI volumes are empty in the extraction. - Decoding
hardware_compatibilityvalues (no table in the image). gps_firmware.binbeyond its header and entropy, and howgpsdactually finds it.- The Wi-Fi firmware and board-data files in
lib/firmware/ath6kandti-connectivitybeyond names and sizes. I read no Wi-Fi key, password or serial from any file and none appear in this report. - The Bebop 2
hsisnumbers were taken from the earlier notes, not re-decoded.
Open questions
Section titled “Open questions”- Which retail product does
ro.product.model.id 0x0913(SKYCTRL_NG) andnapcorrespond to? The image has no marketing name; the unit’s own properties or label would settle it. - Is the original Skycontroller really the nap hardware (R, section above)? A string in the Android-era 1.x firmware or the app would confirm; not in this repo.
- What do
ro.factory.typevaluesbebopandmilosmean physically: the drone the pack was sold with, or a hardware revision? The image only shows the selector and that the serial prefix rule exists. The prefix string innap_factory_reader(10 characters, at rodata 0x11143) is a serial-number prefix; I did not compare it with any real unit serial. - Is
gpsd’s lookup path/etc/gps/<ro.hardware>/satisfied at boot by something not in the image, and what doesro.hardwarehold? - Which GPS chip is it? Protocol
ai2is selected, the library supports several vendors, and the firmware format names none. R: the AI2 name is, I believe, from the SiRF family; confirming would need strings inlibgps.so’s ai2 parser or a part number on the board. - Exact Parrot7 revision and CPU core: runtime-detected, absent from the static kernel.
- Meaning of GPIO 134 to 137 (
ATMEG_GPIO_A..D) and 138 (wifi-wake-up), and the export mode argument 0 or 2. - The inner and outer
hardware_compatibilityvalues (0x31F8B344 and 0x0AF549BF): what do they gate in the installer? - xmega firmware version: the image has no way to read it; the unit reports it (if at all) over the 16 byte vendor report.
- Why the udev rules and manifest carry
mpp2and Qualcomm pieces that this image lacks: shared tree (most likely, I) or an intended later hardware swap.
Comments
Spotted a mistake, or tried this on your own Bebop? Leave a note. Comments are kept on this site's own server.