Hi Andrew,
I have tried it with your suggested command line, and it "feels"
like it makes the issue occur less often (but one never knows with
sparse intermittent issues).
The first time I ran my test, it took 23 minutes before the error
occurred.
However, I picked that when running two processes simultaneously,
the issue occurs reasonably fast.
I have, however, not been able to verify during runtime that the
parameter really arrived in Xen because /proc/cmdline didn't show
the Xen parameters. But I have taken a screen shot of what I have
done to submit the parameter.
If you tell me how I can view the boot options in a running system
to be certain the option has is effective, I'm happy to run the test
again.
Side remark: To make reasonably sure the issue really only happens
with Xen, I have re-verified that without Xen the issue does not
exist, even with two instances for more than 1.5 hours.
Regards
~Joe
On 12/18/23 09:40, Andrew Cooper wrote:
On 18/12/2023 3:34 pm, Joe Tretter wrote:
Hello,
I discussed the below problem with the QubesOS team on Github
(https://github.com/QubesOS/qubes-issues/issues/4493) and they suggest
that this seems to be a problem with Xen, and suggested that I post it
to this e-mail address.
I have problems restoring backups in QubesOS release 4.1.2 on one of
my machines.
Other users reported the issue too, but no QubesOS developer seemed to
be able to reproduce it, therefore nothing happened for a while and
the assumption has been that it's some sort of hardware problem.
I analyzed the problem down to the "scrypt" tool
(https://www.tarsnap.com/scrypt.html) falsely complaining that the
password would be wrong.
I proceeded and re-compiled the "scrypt" tool in it's latest version
and found that the self-tests show intermittent erratic behavior.
This issue only happens with QubesOS/Xen Kernel, if I boot into a
kernel without Xen, the problem can't be observed.
The problem can only be observed on one of my machines which is a Dell
Inspiron 5675 with an AMD Ryzen 7 1700 Eight-Core Processor.
I hope you can help with this issue.
So the issue is specific to your AMD Zen1 system, and only when Xen is
in the mix.
Can you try booting Xen with spec-ctrl=eager-fpu and see if that changes
the behaviour ?
~Andrew
|
