haveged fails to start (systemd)

[dutchfish]

Armbianmonitor:

http://ix.io/2SNR

root@odroidn2:~# journalctl -x | grep haveged
Mar 12 19:43:51 odroidn2 kernel: audit: type=1400 audit(1615574627.248:2): apparmor="STATUS" operation="profile_load" profile="unconfined" name="/usr/sbin/haveged" pid=1635 comm="apparmor_parser"
Mar 12 19:43:47 odroidn2 audit[1635]: AVC apparmor="STATUS" operation="profile_load" profile="unconfined" name="/usr/sbin/haveged" pid=1635 comm="apparmor_parser"
-- Subject: A start job for unit haveged.service has finished successfully
-- A start job for unit haveged.service has finished successfully.
Mar 12 19:43:52 odroidn2 haveged[1836]: haveged: Couldn't initialize HAVEGE rng 9
Mar 12 19:43:52 odroidn2 haveged[1836]: haveged starting up
Mar 12 19:43:52 odroidn2 systemd[1]: haveged.service: Main process exited, code=exited, status=1/FAILURE
-- An ExecStart= process belonging to unit haveged.service has exited.
Mar 12 19:43:52 odroidn2 systemd[1]: haveged.service: Failed with result 'exit-code'.
-- The unit haveged.service has entered the 'failed' state with result 'exit-code'.
 

[lanefu]

Yeah saw that on HC4 as well.   Not sure if it worked in earlier kernels.  Have you tried using an older kernel via armbian-config? 

[Igor]

11 hours ago, lanefu said:

Yeah saw that on HC4 as well.   Not sure if it worked in earlier kernels.  Have you tried using an older kernel via armbian-config? 

It seems related to meson64 only. On RK3399 this feature works normally.

[tparys]

I don't have a N2 or an HC4 handy, but these two links appear interesting ...

 

https://isolated.site/2019/07/08/debian-10-hangs-during-booting-on-phicomm-n1/

https://lists.opensuse.org/opensuse-bugs/2019-03/msg00250.html

 

Both have the same error code, and apparently the same solution. Notably, the phicomm N1 is also an Amlogic S905 ...

 

Any chance setting this in your /etc/default/haveged fixes it?

 

DAEMON_ARGS="-w 1024 -d16"

 

Apparently haveged is supposed to auto-detect CPU settings, but apparently doesn't always work, and the fallback doesn't either.

[iav]

I try that on my Helios4 with Armbian 21.08.0-trunk Hirsute.

Service can't start

Jun 15 20:58:54 helios4 kernel: audit: type=1326 audit(1623790734.437:21): auid=4294967295 uid=0 gid=0 ses=4294967295 subj==unconfined pid=2134 comm="haveged" exe="/usr/sbin/haveged" sig=31 arch=40000028 syscall=122 compat=0 ip=0xb6f9cb68 code=0x80000000

but I can be started from command line, with default options:

# haveged --Foreground --verbose=1
haveged: command socket is listening at fd 3
haveged starting up
haveged: ver: 1.9.8; arch: generic; vend: ; build: (gcc 10.2.1 CTV); collect: 128K
haveged: cpu: (VC); data: 16K (D); inst: 16K (D); idx: 12/40; sz: 15006/57790
haveged: tot tests(BA8): A:1/1 B:1/1 continuous tests(B):  last entropy estimate 7.99541
haveged: fills: 0, generated: 0

 

Edited June 15, 2021 by iav
fix board name

[syn0nym]

On 3/17/2021 at 12:46 PM, tparys said:

DAEMON_ARGS="-w 1024 -d16"

 

Works for me (Odroid N2+). Thanks!

[MichaIng]

This is actually is/was a bug in haveged, or the Debian build itself: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=985196

Strangely we only observed this on armv7l/armhf systems, not aarch64/arm64. However a cleaner solution is to apply the following to the systemd unit:

mkdir -p /etc/systemd/system/haveged.service.d
echo -e '[Service]\nSystemCallFilter=uname' > /etc/systemd/system/haveged.service.d/syscall.conf

But actually, the hardware random generator should work well, making haveged obsolete?
 

apt install rng-tools5 # modern
apt purge rng-tools # legacy
apt purge rng-tools-debian # legacy Debian port
apt purge haveged