Regression in 6.17: ASMedia ASM1164 SATA controller forced to max_performance, preventing disk spin-down

[pbexiga]

Hello, first time poster - and I must admit, I used AI (Claude) to help me piece together all the information for this post. Although I did run through a bunch of troubleshooting to understand where the issue was before going to AI for help. I've done everything to try to identify what could potentially be waking the disks, until I realized that the behavior changed after upgrading from 6.14 to 6.17.

Hardware:

• UGREEN DXP4800 Plus NAS
• OpenMediaVault 7.x running on Proxmox kernel packages
• ASMedia ASM1164 SATA controller (PCI ID: 1b21:1164, rev 02)
• 4x Seagate ST2000DM008 spinning disks in ZFS RAIDZ1

On kernel 6.14, the ASM1164 SATA link power management policy is correctly set to keep_firmware_settings, allowing disks to spin down and stay down. On kernel 6.17, the policy is forced to max_performance for all 4 hosts, which prevents disks from entering or staying in standby — they spin back up within minutes regardless of userspace configuration.

Kernel 6.14.11-5-pve (working):

$ cat /sys/class/scsi_host/host*/link_power_management_policy
keep_firmware_settings
keep_firmware_settings
keep_firmware_settings
keep_firmware_settings

Kernel 6.17.13-1-pve (broken):

$ cat /sys/class/scsi_host/host*/link_power_management_policy
max_performance
max_performance
max_performance
max_performance

Attempting to set med_power_with_dipm or min_power on 6.17 returns Operation not supported. The policy appears to be hard-locked to max_performance for this controller on 6.17.

dmesg output is identical on both kernels:

ahci 0000:05:00.0: SSS flag set, parallel bus scan disabled
ahci 0000:05:00.0: AHCI vers 0001.0301, 32 command slots, 6 Gbps, SATA mode
ahci 0000:05:00.0: 4/4 ports implemented (port mask 0xf)
ahci 0000:05:00.0: flags: 64bit ncq sntf stag pm led only pio sxs deso sadm sds apst

lspci -vvv key details:

05:00.0 SATA controller: ASMedia Technology Inc. ASM1164 Serial ATA AHCI Controller (rev 02)
  LnkCap: ASPM L0s L1, ClockPM+
  LnkCtl: ASPM Disabled; ClockPM+
  Kernel driver in use: ahci

Workaround: Rolling back to kernel 6.14 restores correct behavior.

Has anyone else hit this with ASMedia controllers on 6.17? Is there a known kernel parameter or AHCI quirk that could restore keep_firmware_settings behavior without rolling back?

 

[ste-bm]

Hi, I just came across the same issue with kernel 6.17 where disks consume more power in idle.
With 6.14 I don't have this problem and disk can be set to med_power_with_dipm.

6.17

for f in /sys/class/scsi_host/host*/link_power_management_policy; do echo "$f: $(cat "$f")"; done
/sys/class/scsi_host/host0/link_power_management_policy: max_performance
/sys/class/scsi_host/host1/link_power_management_policy: max_performance
/sys/class/scsi_host/host2/link_power_management_policy: max_performance
/sys/class/scsi_host/host3/link_power_management_policy: max_performance

and

homelab:~# echo 'med_power_with_dipm' > '/sys/class/scsi_host/host1/link_power_management_policy';
-bash: echo: write error: Operation not supported

For now I moved back to 6.14; It'd be good to know once it's fixed or someone has a workaround.

It took me hours to get to the bottom of it, I thought Eth drivers were the cause of increased power consumption as it often happened in the past to me.