Bug 29289

Summary: Constant I/O from process kworker events_freezable_power
Product: Mageia Reporter: Ken Arromdee <arromdee2>
Component: RPM PackagesAssignee: Kernel and Drivers maintainers <kernel>
Status: NEW --- QA Contact:
Severity: normal    
Priority: Normal CC: davidwhodgins, rolfpedersen
Version: 8   
Target Milestone: ---   
Hardware: All   
OS: Linux   
Whiteboard:
Source RPM: kernel 5.10.52-desktop-1.mga8 CVE:
Status comment:

Description Ken Arromdee 2021-07-25 19:03:47 CEST
Description of problem:
The LED on my computer for drive activity is active every few seconds no matter what I do (as long as Linux is running).  I do not need to be running X or a desktop for this problem to happen.

Looking at iotop shows me that there is a kernel process doing I/O which is usually events_freezable_power_, sometimes events_freezable or mm_percpu_wq (pid doesn't change).  The process doesn't show disk reads/writes, just I/O.  No other process shows either I/O or disk reads/writes.  Presumably my drive light shows all IDE activity, not just drive access.

I just upgraded my computer to an MSI A520M Pro motherboard, Ryzen 5 5650x CPU, and more memory.  This problem did not happen on my previous setup which was an AMD FX-4350 on a different motherboard.

The only reference online I could find to a similar problem is this: https://bbs.archlinux.org/viewtopic.php?id=241971 Looking at the recommendations in that thread, my /sys/module/workqueue/parameters/power_efficient is already N and although my drive is ext4 it doesn't seem to allow journal_async_commit.

I have no idea if this problem is dangerous or cosmetic, or if the process is normal and the new motherboard is just stupid about when it turns on the drive light.  The system doesn't have other problems like high CPU usage or disk errors.

Version-Release number of selected component (if applicable):


How reproducible:
Probably just reproducible by me.  Although whether other people have this process may help show whether it's actually a problem.

Steps to Reproduce:
1. Look at disk drive light
2. Run iotop
Comment 1 Lewis Smith 2021-07-25 22:07:06 CEST
Thank you for your report.

To tell us something about your system, please post the O/P of:
 $ inxi -MSDp

CC: (none) => lewyssmith

Comment 2 Ken Arromdee 2021-07-26 03:55:34 CEST
/home/arromdee% inxi -MSDp
System:
  Host: localhost Kernel: 5.10.52-desktop-1.mga8 x86_64 bits: 64 
  Desktop: fVWM2 2.6.9 Distro: Mageia 8 mga8 
Machine:
  Type: Desktop Mobo: Micro-Star model: A520M PRO (MS-7D14) v: 1.0 
  serial: <superuser required> BIOS: American Megatrends v: 1.00 
  date: 08/21/2020 
Drives:
  Local Storage: total: 1.82 TiB used: 635.32 GiB (34.1%) 
  ID-1: /dev/sda vendor: Samsung model: SSD 850 EVO 2TB size: 1.82 TiB 
Partition:
  ID-1: / size: 57.42 GiB used: 11.08 GiB (19.3%) fs: ext4 dev: /dev/sda5 
  ID-2: /home size: 1.14 TiB used: 527.16 GiB (45.0%) fs: ext4 
  dev: /dev/sda7 
  ID-3: /media/windows size: 585.94 GiB used: 97.05 GiB (16.6%) fs: ntfs 
  dev: /dev/sda2 
  ID-4: /media/windows-boot size: 100 MiB used: 28.6 MiB (28.6%) fs: ntfs 
  dev: /dev/sda1 
  ID-5: swap-1 size: 29.3 GiB used: N/A (0.0%) fs: swap dev: /dev/sda6 
/home/arromdee%
Comment 3 Lewis Smith 2021-07-26 21:30:12 CEST
Thank you for that.

My own box is a humdrum uncomplicated desktop system with SATA disc. I have been running for a bit with nothing user active. The machine disc light flashes occasionally and variably: sometimes idle for longish intervals, sometimes a few flashes in a short time, or single ones at shorter intervals. This seems normal: there are things like cron, mgaapplet, msec.
Running just iotop, same behaviour; and like the report, mostly at the top a couple of kworker things.

> I just upgraded my computer
Is the installed Mageia system the same as before?
Is it possible that the two systems indicate(d) disc activity differently?

In any event, I feel that what you are seeing is 'normal' - except for your "every few seconds".
--------------------
The Arch thread you found in comment 0 is very similar, and worth a read. At one point it says:
"This constant idle disk writing (a SSD killer btw) is caused by the "workqueue power-efficient mode" power management option. The kernel config flag is WQ_POWER_EFFICIENT_DEFAULT and should be off by default. There is a good reason this is off in the kernel defconfig."
Which of course may not be relevant.

In the light of that discussion, assigning this to the kernel team.

Assignee: bugsquad => kernel
CC: lewyssmith => (none)

Comment 4 Dave Hodgins 2021-07-26 22:44:17 CEST
Do you have smartd.service or hddtemp.service running?

CC: (none) => davidwhodgins

Comment 5 Ken Arromdee 2021-07-27 02:41:59 CEST
>Is the installed Mageia system the same as before?

Yes.

>Is it possible that the two systems indicate(d) disc activity differently?

It's possible but I have no way to determine this.

iotop always shows it as I/O, not as disk reads/disk writes.  I occasionally get a [jdb2/sda7-8] which is writing a journal, and that does show disk reads/writes, but that goes away if I leave the machine running without doing anything, so it's probably unrelated.

>The kernel config flag is WQ_POWER_EFFICIENT_DEFAULT and should be off by default.

As I said, I checked this and it's already set to 'N'.

>Do you have smartd.service or hddtemp.service running?

No.  These aren't in mcc and don't seem to be running.

[root@localhost arromdee]# service --status-all|grep smart
[root@localhost arromdee]# service --status-all|grep hdd
[root@localhost arromdee]#
Comment 6 Ken Arromdee 2021-07-27 03:23:03 CEST
Further information: This problem doesn't appear while I am on the grub2 menu.  It does appear if I reboot into single user mode or into the Mageia 8.0 disk using rscue.
Comment 7 Ken Arromdee 2021-07-27 03:45:09 CEST
And now that I look more closely, this isn't that similar to the archlinux thread.  That one has writes every minute, and they actually show up as writes, not only as I/O.  This one is every two seconds or so and they are I/O only.  I'm only assuming that the I/O is turning on the drive light, since there's no real disk activity.

Even if it is, it may be that my BIOS is just buggy and shouldn't be turning on the light.  It's the kind of bug that would go unnoticed, because Windows accesses the disk a lot and that'd mnask it.
Comment 8 Dave Hodgins 2021-07-27 04:19:00 CEST
Try booting with the kernel options pcie_port_pm=off pcie_aspm=off
Comment 9 Ken Arromdee 2021-07-27 07:38:05 CEST
Does not help.

/home/arromdee% cat /proc/cmdline
BOOT_IMAGE=/boot/vmlinuz-5.10.52-desktop-1.mga8 root=UUID=312bf85e-3d2e-4ba7-ac79-411bd00afc2a ro pcie_port_pm=off pcie_aspm=off nosplash quiet noiswmd resume=UUID=ab7e7f00-83e0-4d28-9671-745379099456 audit=0 rd.systemd.unit=multi-user.target vga=791
/home/arromdee%
Comment 10 Rolf Pedersen 2021-07-27 14:57:15 CEST
(In reply to Ken Arromdee from comment #5)
...
> 
> >Do you have smartd.service or hddtemp.service running?
> 
> No.  These aren't in mcc and don't seem to be running.
> 
> [root@localhost arromdee]# service --status-all|grep smart
> [root@localhost arromdee]# service --status-all|grep hdd
> [root@localhost arromdee]#

[rolf@x570i ~]$ sudo service --status-all|grep smart
[rolf@x570i ~]$ rpm -qf `which service`
initscripts-9.78-27.mga8

I see the cmdline in Comment #9 references systemd, an "upgrade" to initscripts, aiui.

Whereas `service` on my machine returns the same as you, I'm not so ambitious as to try to hang on to initscripts.  

[rolf@x570i ~]$ sudo systemctl -a | grep smart
  smartd.service                                                                                                                                                          loaded    active   running   Self Monitoring and Reporting Technology (SMART) Daemon
[rolf@x570i ~]$ rpm -qf `which systemctl`
systemd-246.15-1.mga8

FWIW.

CC: (none) => rolfpedersen

Comment 11 Ken Arromdee 2021-07-27 15:48:47 CEST
[root@localhost arromdee]# systemctl -a | grep smart
[root@localhost arromdee]#