Hi Hailo Community,
I am experiencing a persistent firmware loading failure with my Hailo-10H module on a Raspberry Pi 5 (using the official Raspberry Pi AI HAT+ and the official 27W Power Supply). I have been troubleshooting this for 3 weeks and suspect a hardware issue with the DDR subsystem on the module.
Environment:
Host: Raspberry Pi 5 (16GB)
OS: Raspberry Pi OS (Debian Trixie / Kernel 6.12)
Kernel Page Size: Switched to 4KB (kernel8.img)
Driver Version: 5.2.0 (compiled via DKMS for the current kernel)
Boot Parameters: iommu=off, pcie_aspm=off, dtparam=pciex1_gen=2 (also tested on Gen 1)
The Issue:
The driver successfully probes the device, reads the SKU-ID (6), and writes the customer_certificate.bin, scu_fw.bin, and the signed DTB. However, it consistently fails at Stage 2 (SOC firmware / image-fs) with a timeout error -110.
Critical SCU Log Output:
The SCU log reveals a fatal hardware interrupt during DDR initialization:
text
[ 10.206903] hailo1x 0001:01:00.0: SCU log:
[ 10.206905] hailo1x 0001:01:00.0: [ 0.000000] SCU-FW v1.9.0 [b9f32b9a]
[ 0.000000] [FWK] Module initialization complete!
[ 0.048765] [BOOT] SCU-FW notified PCIe host, ready to receive boot DTB image
[ 0.063746] [BOOT] PCIe host completed sending boot DTB image
[ 0.084734] [SAFETY] SCU failure: safety fatal interrupt
[ 0.084910] [SAFETY] safety_agg_status = 0x00000080
[ 0.085078] [DDR] DDR failed: unexpected safety fatal interrupt
[ 0.085276] [DDR] subsys: 0x7C068008=0x00000020 (mask=0x00000037)
KĂłd pouĹľĂvajte opatrne.
Steps already taken:
Verified all firmware files are present in /lib/firmware/hailo/hailo10h/.
Ensured 4KB PAGE_SIZE is active.
Disabled IOMMU and ASPM.
Tested with PCIe Gen 1 for signal integrity.
Performed multiple cold boots (30s power-off).
Checked the FPC cable connection (shielded 35mm cable).
Since the SCU explicitly reports a Safety Fatal Interrupt and DDR failure immediately after the DTB transfer, is this a confirmed hardware defect? Could it be a power delivery issue on the AI HAT+ or a faulty DDR controller on the Hailo-10H module itself?
Any guidance on whether this requires an RMA would be greatly appreciated.
Thanks!
Of course, I followed the official procedure first. Followed the official Raspberry Pi installation procedure for the Hailo-10H (AI HAT+2), which failed at the firmware loading stage with a timeout error. Attempts using alternative methods, including one from RonsTechHub, resulted in the same error message: “hailo1x 0001:01:00.0: Timeout waiting for firmware file”. The user emphasizes that a previous SCU log indicated a Safety Fatal Interrupt related to DDR failure, preventing the hailo1x driver from activating the board and running hailo-apps examples.
I can also do a clean install with an RPI Connect account, which I’ll let you know and you can connect directly. I’ve tried that so many times that I know what the result will be. I should get a new PCIex cable today. Maybe that will make a difference.
log:
Setting up h10-hailort-pcie-driver (5.1.1) …
WARNING: apt does not have a stable CLI interface. Use with caution in scripts.
build-essential/stable,now 12.12 arm64 [installed]
Failed to install PCIe driver to the DKMS tree. Trying to install PCIe driver without DKMS
Please reboot your computer for the installation to take effect.
dpkg -l | grep hailo
ii h10-hailort 5.1.1 arm64 AI acceleration runtime library for Hailo hardware
ii h10-hailort-pcie-driver 5.1.1 all Hailo PCIe driver and firmware for AI hardware acceleration
ii hailo-h10-all 5.1.1 all Hailo-10 support (metapackage)
ii hailo-models 1.0.0-2 all AI models for Hailo modules
ii hailo-tappas-core 5.1.0 arm64 Core components of the Hailo TAPPAS platform
ii python3-h10-hailort 5.1.1-1 arm64 HailoRT Python API, which wraps the runtime library
ii python3-hailo-tappas 5.1.0 arm64 Python binding for tappas
ii rpicam-apps-hailo-postprocess 1.11.1-1 arm64 Raspberry Pi AI HAT+ post-processing plugin for rpicam-apps
Hello, did you fixed this issue? I see (AI HAT+) in the topic subject but do you realy have (AI HAT+2), I mean are you sure about Hailo10 you need to install but not Hailo8? Here is my “lspci -v“ command output that shows the device is realy connected and identified. If you have no devices here with “Co-processor” it mean’s you have no connection. But if you have Hailo-8 there you just need to chose another firmware
bus master and few other options are missing (maybe it’s the PCIex cable issue), you can try to reconect or change PCIex gen3 to gen1. Also check for the driver files with “find / -name “hailo_pci.ko” 2>/dev/null”. I have next output:
/usr/lib/modules/6.12.47+rpt-rpi-v8/kernel/drivers/media/pci/hailo/hailo_pci.ko.xz
/usr/lib/modules/6.12.47+rpt-rpi-2712/kernel/drivers/misc/hailo_pci.ko.xz
You can try to remove all of them and re-install the driver only again
