Skip to main content
The Exaviz Cruiser exposes PCIe x1, an M.2 Key M slot for NVMe, and a SATA port via an ASM1061 bridge. All three interfaces share the CM5’s single PCIe lane and are enabled through config.txt parameters.

​
Enabling PCIe

Add the following line to /boot/efi/config.txt to activate the PCIe x1 interface:
/boot/efi/config.txt
# CM5 Exaviz carrier — PCIe/SATA will enumerate after UEFI handoff
dtparam=pciex1=on
PCIe is disabled by default on the CM5. Without dtparam=pciex1=on, neither the M.2 NVMe slot nor the SATA bridge will be visible to the kernel.

​
Enabling NVMe

With PCIe active, enable the NVMe driver overlay:
/boot/efi/config.txt
dtparam=pciex1=on
dtparam=nvme=on
After the next boot, verify enumeration: 
lsblk -d -o NAME,MODEL,SIZE,TRAN | grep nvme
Expected output: 
nvme0n1   Samsung SSD 980 500G   465.8G   nvme

​
SATA enumeration via ASM1061

The Exaviz Cruiser routes SATA through an ASM1061 PCIe-to-SATA bridge. No additional config.txt parameters are required beyond dtparam=pciex1=on. The kernel’s ahci driver handles the bridge automatically and the disk appears as /dev/sdX. 
lsblk -d -o NAME,MODEL,SIZE,TRAN | grep sata

sda   WD Red Plus 2TB   1.8T   sata
If the SATA disk is not visible after enabling PCIe, check dmesg | grep -i asm106 to confirm the bridge was detected. A missing bridge usually indicates a PCIe link training failure — check that LnkSta reports the expected speed with lspci -vvv.

​
Adding NVMe to the btrfs pool

The CM5 image uses btrfs with a @ root subvolume on eMMC. After the NVMe device is confirmed present, add it to the pool and mirror the data:
1

Add the NVMe device

btrfs device add /dev/nvme0n1 /
2

Balance to RAID 1

btrfs balance start -dconvert=raid1 -mconvert=raid1 /
This converts both data and metadata to RAID 1 across eMMC and NVMe. The balance operation runs in the background and can be monitored with:
btrfs balance status /
3

Verify the pool

btrfs filesystem show /
Both devices should appear with their sizes and used space.
Enable fstrim.timer to periodically TRIM both the eMMC and NVMe for long-term health:
systemctl enable --now fstrim.timer

​
Frigate recordings and Docker layout

Frigate NVR writes large sequential recording files that do not benefit from btrfs copy-on-write. Two dedicated subvolumes with nodatacow are used to avoid performance penalties.

​
Btrfs subvolumes

Create the subvolumes during initial image setup (or post-boot on an existing pool): 
btrfs subvolume create /mnt/@frigate
btrfs subvolume create /mnt/@docker

​
fstab entries

/etc/fstab
# Frigate recordings — nodatacow for large sequential writes
UUID=<ROOT_UUID>  /media/frigate   btrfs  compress=zstd,space_cache=v2,noatime,subvol=@frigate,nodatacow  0  0

# Docker layers — nodatacow improves layer extraction performance
UUID=<ROOT_UUID>  /var/lib/docker  btrfs  compress=zstd,space_cache=v2,noatime,subvol=@docker,nodatacow   0  0
nodatacow disables checksumming on those subvolumes. This is intentional for Frigate and Docker workloads, but means data corruption will not be detected by btrfs scrub. Back up critical data elsewhere.

​
Frigate recordings path

Frigate expects recordings at /media/frigate/recordings. With the @frigate subvolume mounted at /media/frigate, no additional configuration is needed in Frigate’s config file. The Samba share for snapshot browsing is configured at /media/frigate/snapshots. The Frigate container definition (from docker-compose.yml) mounts the path directly:
docker-compose.yml
frigate:
  image: ghcr.io/blakeblackshear/frigate:stable
  volumes:
    - /media/frigate:/media/frigate
  devices:
    - /dev/axelera0:/dev/axelera0   # Metis AIPU
    - /dev/dri:/dev/dri             # VC4 GPU for decode assist

​
SMART health monitoring

Install smartmontools for SATA disks and nvme-cli for NVMe devices. Both are listed in the stack packages: 
pacman -S smartmontools nvme-cli

​
SATA disk health

# One-time short self-test
smartctl -t short /dev/sda

# Check results
smartctl -a /dev/sda | grep -E "SMART overall|Reallocated|Pending|Uncorrectable"

​
NVMe health

# NVMe SMART log
nvme smart-log /dev/nvme0n1

# Critical warnings field — 0x00 means healthy
nvme smart-log /dev/nvme0n1 | grep critical_warning

smartctl -a /dev/sda
Enable periodic SMART polling with smartd. Add /etc/smartd.conf entries for each device and enable the daemon:
systemctl enable --now smartd

Build docs developers (and LLMs) love

Get started for freeTalk to us