New Raspberry Pi CM4 WiFi testing

Explore WiFi on recently acquired Raspberry Pi Compute Module Lite with WiFi and Bluetooth.

Motivation

  1. While provisioning the new CM4 it was noted that many IoT devices were droppinig from the LAN including TP-Link Kasa lights and outlets and Raspberry Pi Zeroes. This is not unusual, particularly the TP-Link devices, but it seemed that nearly half of the devices were disconnected and that is unusual.
  2. At that time I thought to turn off the radio on the new CM4 and discovered that wlan0 did not appear in the output of ip addr.

These may or may not be related. The milestones I can recall before taking a step back to investigate WiFi were:

  1. Install a NVME SSD with a copy of R-Pi OS (which has not been used for a while) on the Ether Board https://hankb.github.io/MkDocs-blog/tech/dphacks-CM4_Ether_Board/, attach the new CM4 to the board and boot. The board was connected to my LAN via Ethernet.
  2. Find the device on my LAN and SSH in. I discovered that this install was R-Pi OS and wanted to use Debian for this system so I recorded the output of ip addr and shut the system down. (I needed the MAC addresses for Ethernet and WiFi to remap to avoid later confusion.)
  3. Install the most recent "tested Debian" build on the card including some additional setup. The image installed was 20230612_raspi_4_bookworm.img.xz downloaded from https://raspi.debian.net/tested-images/
  4. Completed initial configuration using the Ansible Playbooks at https://github.com/HankB/polana-ansible.
  5. Continued configuration and noticed that many devices on my IoT LAN were disconnected from my Access Point (AP.) I power cycled the AP and that did not seem to help. I was concerned about possible RF interference from the new CM4 since it was located about 1 foot from the AP. I was not planning on using WiFi so I thought to disable that radio in case it was the cause of the WiFi issues. That was when I noticed that ip addr no longer displayed a device wlan0.

At this point I began a less than systematic exploration of the "missing wlan0" situation. Before I describe that, I'll list the CM4 related equipment I'm using.

  • CM4 Lite with WiFi/Ethernet and 8GB RAM. I have been using this for quite a while now. I will call it the "CM4/8GB" in the following discussion.
  • Official Raspberry Pi IO Board (on which the CM4/8GB is presently mounted.)

  • PCIe/NVME adapter on which I can mount an NVME SSD and boot either R-Pi OS or Debian.

  • CM4 Lite with WiFi/Ethernet and 2GB RAM. This is recently acquired.

  • Dphacks Ether Board. I used the CM4/8GB on this for initial testing and confirmed that everythging I tested (including WiFi) worked as expected. When I received the CM4/2GB, I installed it on this board and proceeded with configuration.

  • 256GB KIOXIA NVME SSD which (along with an NVME/USB adapter) I used for initial configuration and plan to use for further testing.

  • 128GB SK Hynix NVME SSD on which is installed a fully updated and overclocked 64 bit R-Pi OS. This was used during the initial exploration but will not be used for the more systematic testing.

initial exploration

I performed the following steps, not necessarily in the listed order, to tryt to isolate the issue. I did not record each step as I tested, hence the need to retest. Also I used existing installations which had been customized includinig overclocking.

  1. Swap the 128GB SSD to the Ether Board/CM4/2GB and boot. Confirm that wlan0 does not appear in ip addr. Confirm that there are no messages that include wlan0 or wifi in dmesg output. Confirm that there are Bluetooth (BT) related messages in dmesg output. The common factor hewre is the CM4/2GB and Ether Board.
  2. I believe I checked for (and found) wlan0 with the CM4/8GB in the IO Board and running R-Pi OS from the 128GB SSD.
  3. Install the 256GB SSD (Debian) on the IO Board with the CM4/8GB. The wlan0 device was not present. The common factor here is the S/W.

At this point I saw the need to perform a more systematic investigation to try to narrow down the reason for the missing wlan0 (and possible related WiFi issues.)

  1. Test with pristine R-Pi OS image (2023-05-03-raspios-bullseye-arm64-lite.img.xz.) No joy here. On first boot it loops in keytboard configuration and I cannot find a way to bypass that. I booted R-Pi OS (old installation) from SD card and wlan0 is not present, but wifi is disabled in /boot/config.txt. Commented that out, rebooted and still no wlan0 but I could have done something erlse like blacklisting the modules. Undid that and wlan0 is present and appears to be working (obtained IP addr using DHCP.) I wrote the same image to an SD card and was able to boot w/out any issue. And the wlan0 device was present. The next test is to copy the SD card to the NVME SSD and boot from that.
  2. Image initialized on an SD card and copied to NVME SSD booted CM4/8GB/IO Board w/out difficulty and wlan0 was present.
  3. Transferred the NVME SSD to the CM4/2GB/Ether Board. System booted, logged in and found that wlan0 was present.
  4. Performed an upgrade (still on CM4/2GB/Ether Board) and rebooted. wlan0 still present.

In other words, on both H/W setups, wlan0 was present. Next series involves testing with Debian 20230612_raspi_4_bookworm.img.xz. No joy. Does not boot from NVME. Debugging that is an issue outside the scope of this investigation. 20230101_raspi_4_bookworm.img.xz seems to be hanging at the same place.

  1. Copy 20230612_raspi_4_bookworm.img.xz to the NVME SSD and try to boot the CM4/8GB/IO Board. Hung <ctrl><alt><del> led to a halted system. Repeated power cycles resulted in the same result.
  2. Copy 20230101_raspi_4_bookworm.img.xz to the NVME SSD and repeat trying to boot the same H/W. It came up to the point where it reported mmc0: Timeout waiting for cmd interrupt (whichj has been fixed in subsequent kernels.) Working with this variant was abandoned (and later resumed.)
  3. Copy 20230612_raspi_4_bookworm.img.xz to an SD card and attempt boot on CM4/8GB/IO Board. Boot completes and results in a login prompt on the console. ip addr lists only lo and eth0. Installed pciutils and usbutils. lspci lists nothing. lsusb lists 2 hubs and a keyboard.
  4. Copy 20230101_raspi_4_bookworm.img.xz to another SD card and boot. Repeatedly comes up with the mmc0: Timeout waiting for cmd interrupt but recollection from previous testing is that this does not happen 100% of the time. After 4 unsuccessful boots (requiring power cycling) the OS came up normally and presented a login prompt. The wlan0 device was not present.

Conclusion

This appears to be a S/W issue. R-Pi OS in the most recent version (based on Bullseye) brings up wlan0 on the CM4/8GB/IO Board reliably. Recent versions of Debian (Bookworm) do not.

The problems with other devices on my local LAN seem to be unrelated.

Future

  1. Determine why wlan0 is not coming up on Debian Bookworm. This is not a high priority as I do not plan to use WiFi at present. I would nevertheless be willing to test if someone can point me in the right direction.
  2. Determine why Debian Bookworm does not boot from NVME. The interesting thing is that using my playbook to copy the image to the boot media and tweaking a few other things results in a system that does boot. One thing the playbook does is to resize the root partition (and creates an extra partition for ZFS) and that may bypass the boot problem.
  3. Test with Buster and Testing to see if results are different.
  4. Determine why R-Pi OS loops on keyboard configuration when booting from NVME SSD. Also not a high priority but if someone in a position to fix this is interested, I'd be happy to test.