datum
April 9, 2026, 10:11am
1
I have a Radxa Dragon Q6A with the Radxa 256GB UFS module (Kioxia THGJFGT1E45BAILB).
It is powered by a Radxa DC12 36W power adapter via the 3-pin connector, but I’ve tried a USB-C PD adapter too. It has a 13mm all-copper heatsink with fan that keeps it cool (less than 60 °C in stress test)
I flashed the radxa-dragon-q6a_noble_gnome_r2.output_512.img.xz image on a USB, booted and then downloaded and flashed radxa-dragon-q6a_noble_gnome_r2.output_4096.img.xz on UFS.
The system boots up fine from UFS but after a while it gets stuck or reboots randomly!
I stress tested CPU and memory with stress-ng and they are fine.
Running fio to test UFS (when booted from USB), and using dmesg -w , I see lots of ufshcd-qcom errors, mostly PM and hibern8 issues. Also, instead of 1811/557 MB/s (specs for Radxa OS using UFS module Kioxia THGJFGT1E45BAILB), fio reports ~60 MB/s R/W !!
It doesn’t look like a hardware issue, more like a UFS driver issue, but like many linux drivers there isn’t a separate loadable module, ufshcd‑qcom driver is compiled into the kernel.
uname -r shows that I’m running 6.18.2-3-qcom kernel.
I opted for UFS for better power efficiency and comparable speed to NVMe, and I’m very disappointed.
Has anyone faced similar issues?
Yes. R2 here too. UFS3 is original RADXA btw.
Powered by POE (X4 model). Had some issues getting the fan on the POE HAT working.
Rebooting using reboot command not always works; sometimes it time’s out.
I will post the kernel messages in a later post.
[ 326.034889] audit: type=1131 audit(1775806393.264:2203): pid=1 uid=0 auid=4294967295 ses=4294967295 msg=‘unit=packagekit comm=“systemd” exe=“/usr/lib/systemd/systemd” hostname=? addr=? terminal=? res=success’
datum
April 10, 2026, 3:01pm
4
This is very similar if not identical to what I get.
It “stutters” (freezes for 2-3 secs) every couple of minutes and after about an hour it freezes permanently or reboots. I have 12GB RAM and a huge heatsink with fan on it and the temperature is usually below 40°C so it’s definitely not throttling or anything like that.
dmesg.dump.3.zip (28.6 KB)
Thanks for your message. We’re actively investigating this issue and will update you shortly.
You can try update boot firmware to 260120Resource Downloads | Radxa Docs
Thank you for your reply.
Regarding: 2601020
[ 0.029057] DMI: Radxa Computer Co., Ltd. Radxa Dragon Q6A/Radxa Dragon Q6A, BIOS 6.0.260120.BOOT.MXF.1.0.1-00549-KODIAKWP-1 01/20/2026
I think I am already running this version, am I correct?
Thanks in advance!
I have currently 3 dragon-q6a’s waiting for the UFS to work so I can deploy them and try to get SUSE MicroOS ported to them.
Kind regards,
Pascal
Pascal
datum
April 14, 2026, 8:02pm
8
I just updated firmware to 6.0.260120.BOOT.MXF.1.0.1-00549-KODIAKWP-1
No change.
I attach a fresh dmesg dump.
6.0.260120.dump.4.zip (26.4 KB)
Ok, I will review the dmesg dump
I tried using KIOXIA UFS and could see error logs, but there was no reboot issue. Additionally, I tested a Samsung UFS and haven’t encountered any problems so far. I suspect there might be an issue with the KIOXIA-related driver.
Prolongued it hangs or restarts.
Will there be a fix for the KIOXA related driver?
The UFS is sold under the RADXA brand as is the dragon-q6a, so I am assuming they ought to be compatible.
Thanks in advance.
Kind regards,
Pascal
datum
April 16, 2026, 10:32am
12
Maybe DTB uses Samsung settings for all UFS modules, and they are incompatible with Kioxia?
So far, I have had a total of 3 permanent freezes and 2 reboots, so these may be hard to detect, but stuttering is easily detectable (short freezes of 1-5 seconds every few minutes). There’s no stutter when booting from USB, the problem exists only when booting from UFS.
The log contains HW_H8_ENTER_CNT=102, HW_H8_EXIT_CNT=0, SW_H8_ENTER_CNT=76, SW_H8_EXIT_CNT=178, seeing this I though that UFS hibernation was the culprit.
auto_hibern8 was set to 150ms, I disabled it until next reboot:
echo 0 | sudo tee /sys/bus/platform/devices/1d84000.ufshc/auto_hibern8
the log showed:HW_H8_ENTER_CNT=0, HW_H8_EXIT_CNT=0, SW_H8_ENTER_CNT=26, SW_H8_EXIT_CNT=26
Unfortunately stuttering continued, even with UFS hibernation completely off, so I guess it’s not a UFS hibernation issue (see attached auto_hibern8=0.log).
Maybe lowering speed by running it at a lower gear? or adjusting regulator voltages?
There seems to be a DTB overlay mechanism in place, but I don’t want to mess with DTB on my own. If Radxa created an overlay, I can test it.
Demetrios
auto_hibern8=0.zip|attachment (8.9 KB)
1 Like
I’m currently testing a workaround by lowering the UFS communication speed to see if it resolves the issue and allows the system to function properly.
I’ve applied the setting limit-hs-gear = <2>; in the device tree, and so far the UFS subsystem is running without any error logs. You can go ahead and try it on your end.https://github.com/chenchongbiao/linux-qcom/releases/download/test/q6a-kernel.zip
Command
unzip q6a-kernel.zip
cd q6a-kernel
sudo apt reinstall ./*deb
Unfortunately; that didn’t help..
Still hangs and a lot of angry kernel messages:
pascal@dragon-q6a-n05:~$ uname -a#4 SMP PREEMPT_DYNAMIC Fri Apr 17 11:17:34 CST 2026 aarch64 aarch64 aarch64 GNU/Linux
[ 101.766973] audit: type=1105 audit(1776444363.505:161): pid=1520 uid=1002 auid=1002 ses=2 msg=‘op=PAM:session_open grantors=pam_limits,pam_env,pam_env,pam_permit,pam_umask,pam_unix acct=“root” exe=“/usr/bin/sudo” hostname=? addr=? terminal=/dev/pts/0 res=success’
1 Like
/dev/sda:
datum
April 18, 2026, 10:14pm
18
I agree with Pascal, this patch did not resolve the issue.
Identical pa_err=0x80000003 and dl_err=0x80004000
HW_H8_EXIT_CNT still 0 (PHY cannot exit hardware hibernate)
ufshcd_eh_device_reset_handler still failing with -110
As for the symptoms:
I see no change in stuttering.
Did not run long enough for reboot or complete freeze to occur.
UFS speed is crippled (not somewhat slower):
Lowering UFS link from HS-G3 to HS-G2 did not fix the problem nor did disabling hibernation.sudo cat /sys/kernel/debug/regulator/regulator_summaryvreg_l10c_0p88 be 880mV instead of 720mV?
Demetrios
dmesg_before&after.zip (52.1 KB)
info_before&after.zip (3.5 KB)
If I have any new findings, I’ll be sure to share them here promptly.
min-microvolt 720mV
