Hi Michael,
I built a 4.9.137 kernel and added some trace code and found that the
issue occurs during device_shutdown() in the linux kernel. Specifically
when it tries to shutdown the following device:
00:1d.0 USB controller: Intel Corporation Atom Processor Z36xxx/Z37xxx
Series USB EHCI (rev 11)
As an experiment we removed the ehci-pci.ko and ehci-hcd.ko kernel
modules so that ehci would not be used. By doing this we found that
reboots and powerdowns worked 100% of the time.
This issue does not occur when I use antiX-17.1_386-core.iso (does not
used systemd) or a very old 32 bit PCLOS distro (does not used systemd).
I performed more testing with antiX-17.1_386-core.iso (does not used
systemd) and found that this issue DOES occur.
I was able to confirm that the very old 32 bit PCLOS distro (does not
used systemd) does not have this issue.
This is logged right before systemd-shutdown calls the final
reboot() system call. If it get's stuck after that, it looks like it is
a kernel issue.
I think you are right, this looks like a kernel issue, not a systemd
issue. Currently I have logged this as a systemd issue, how can I change
this to a kernel issue?
Other investigations showed (ehci was not forcibly removed):
* PCLOS *with linux 3.2 is OK*
* PCLOS with linux 4.8.7 is not OK
* Debian Wheezy *with linux 3.2 is OK*
* Debian Jessie with linux 3.18 is not OK
* Debian Stretch with linux 4.9.137 is not OK
Regards, Andrew
On 08/11/18 02:11, Michael Biebl wrote:
Am 07.11.18 um 04:24 schrieb Andrew Pavlomanolakos:
Hi,
I tried with another kernel from backports:
root@entry115:~# uname -a
Linux entry115 4.18.0-0.bpo.1-686-pae #1 SMP Debian 4.18.6-1~bpo9+1
(2018-09-13) i686 GNU/Linux
The same issue was reproduced with kernel 4.18.0-0.bpo.1-686-pae. Here
is the trace:
Ok, thanks for testing.
I'm out of ideas at this point, so we should take this upstream.
For that it would be best if you can reproduce that with the current
upstream version, i.e. v239, which is also available from stretch-backports.
If the issue is still reproducible then, it would be great if you can
file a report athttps://github.com/systemd/systemd/issues
Regards,
Michael
