Orin Nano Devkit: JetPack 7.2 / 7.2.1 boot loops with rst_source: 0x1c in MB2

Hi,

My Orin Nano Developer Kit runs JetPack 6.2.1 (R36.4.4) fine, but JetPack 7.2 and 7.2.1 both boot-loop before UEFI. UART shows the SoC is reset by the voltage monitor (rst_source: 0x1c = VMON) at the same point in MB2 every cycle. This is a genuine Orin Nano devkit using official 19 V adapter.

What I tried

  1. Jetson ISO USB (7.2 and 7.2.1): capsule update accepted → reboot → loop.
  2. Recovery flash from Ubuntu host with R39.2.x BSP. Flash completes successfully, then cold boot loops the same way.
  3. Same recovery-flash procedure with R36.4.4 → board boots 6.2.1 normally. Repeated 6→7→6, fully reproducible.
  4. Retried 7.x with nothing attached but power, UART and boot storage. Same loop.

Failure point

MB1 reports the previous reset was VMON, boots chain B (never fails over to A). MB2 finds all 60 QSPI partitions, loads MCE/BPMP/SPE/RCE/DCE/APE/XUSB/PVA/cpu-bootloader with all integrity checks passing, then resets while loading B_secure-os as BPMP-FW is enabling clocks. Never reaches BL31/OP-TEE/UEFI.

[0000.065] I> MB1 (version: 1.0.1.17-t234-54845784-9b0d5809)
[0000.075] I> Boot-mode : Coldboot
[0000.089] C> rst_source: 0x1c, rst_level: 0x0
[0000.093] I> Total boot-chains: 2   Active-chain: 1|B  -  BR-BCT based selection
...
I> MB2 (version: 0.0.0.0-t234-54845784-9446b2bf-06Aug26-214857-0700)
I> Found 60 partitions in QSPI_FLASH (instance 0)
...
I> cpubl: plain bin integ OK
I> Partition name: B_secure-os
I> Bin@ dev:3/0 blk-94976 sz:0x400000 B_secure-os
I> Ratchet[B_secure-os] sw=0 hw=0
I> tos ra
initialized console_late
handling unreferenced clks
...
enable xusb_ss
enable xusb_fs
enable xusb_falcon
enable xusb_core_mux
enable ds
[0000.065] I> MB1 (version: 1.0.1.17-t234-54845784-9b0d5809)
[0000.071] I> t234-A01-1-Silicon (0x12347) Prod
[0000.075] I> Boot-mode : Coldboot
[0000.078] I> Entry timestamp: 0x00000000
[0000.082] I> last_boot_error: 0x0
[0000.085] I> BR-BCT: preprod_dev_sign: 0
[0000.089] C> rst_source: 0x1c, rst_level: 0x0
[0000.093] I> Total boot-chains: 2   Active-chain: 1|B  -  BR-BCT based selection
[0000.100] I> Default chain: 1|B

Questions

  1. Is a VMON reset in MB2 on R39.2.x known for any Orin Nano devkit revision? Did VMON thresholds or rail/clock bring-up change from R36 to R39?
  2. Since the R39 recovery flash completes (R39 MB1/MB2/BPMP-FW run via RCM) but cold boot from QSPI trips VMON, is this a firmware issue or a marginal module I should RMA?
  3. Why does MB1 never fail over to chain A?

Happy to run test builds or collect more logs. Thanks!

user@jetson:~$ cat /etc/nv_boot_control.conf
TNSPEC 3767-302-0005-J.1-1-0-jetson-orin-nano-devkit-super-
COMPATIBLE_SPEC 3767–0005–1–jetson-orin-nano-devkit-super-
TEGRA_BOOT_STORAGE mmcblk0
TEGRA_CHIPID 0x23
TEGRA_OTA_BOOT_DEVICE /dev/mtdblock0
TEGRA_OTA_GPT_DEVICE /dev/mtdblock0

screenlog.txt (235.1 KB)

Hi,

What is your flash method?
Please provide the steps by steps in detail.

Thanks

Hello,

Thanks for the reply. Host side: Ubuntu 24.04 x86_64 for most attempts, a few on 22.04 to rule out the host OS; firewall disabled on both. Same hosts and cable flash JetPack 6.2.1 (R36.4.3/R36.4.4) successfully with every method below, so the host setup is a known good control.

  1. Flashing qspi only with sudo ./flash.sh --no-systemimg -c bootloader/generic/cfg/flash_t234_qspi.xml jetson-orin-nano-devkit-super internal Completes and reports all partitions written. On cold boot, the boot loop. (At the time of bootloop, the storage are either empty, or containing bootable 6.2.1 os. From what I can tell, the failure is in MB2 while loading secure-os, before UEFI, so the contents of the boot storage aren’t reached; the loop is identical whether the storage holds any rootfs)
  2. Booting 7.2.1 ISO using USB, choose Y when asking for update QSPI, update process goes to 100% then reboot, the bootloop on next. (normally there should be 2 update processes, for me, the bootloop started after the 1st update, 2nd never happened.)
  3. using SDKmanager 2.4.1 direct flash method, as well as calling l4t_initrd_flash.sh , all same error as above (I don’t have the exact command now, but I was able to use the same command on 6.2.1 to bring the system back).

Hello @DavidDDD Any updates? If there is anything you want me to try, or provide any log, please let me know

Diffing 7.2.1 vs 6.2.3 boot chains, the 7.2.1 BPMP-FW DTB (bootloader/generic/tegra234-bpmp-3767-0003-3768-super.dtb) carries system-cfg { sec-faults-rst-en = <0x06>; }, which no r36 BPMP DTB has. It appears to make BPMP-FW reset the chip when a security/safety fault such as VMON is flagged.

I set it to 0 with fdtput -t x [dtb] /system-cfg sec-faults-rst-en 0 and flashed with flash.sh -g [dtb] jetson-orin-nano-devkit-super-qspi internal. The VMON reset is gone: MB1 reports rst_source: 0xb, BPMP-FW reaches bpmp: init complete, MB2 loads cpubl/tos/eks. But MB2 then dies at Trigger load TSEC leyblob with !0x380 Exception! [elr:0x50020844, spsr:0x200002cd, esr:0xbe000011, far:0x0] (SError at EL3) and the screen stays black.

Since I don’t see wide spread reports on similiar issue on the forum, I think this points to a VMON fault condition on this specific unit that 6.2.3 never acted on and that 7.2.1’s reset-on-fault handling now exposes. Therefore more likely a defective unit.

But this is based on guesswork of what sec-faults-rst-en = <0x06> means, could Nvidia confirm or correct my analysis above?