[Discuss] Issues with new Linux computer mostly addressed

[Alan W. Irwin]

> [A kind responder to one of my bug reports concerning the lockups said the following]

> To prevent random kernel lock ups with Ryzen, enable RCU_NOCB_CPU in
> the kernel configuration and boot the kernel with the rcu_nocbs=0-X
> command line parameter. X is the cpu thread count -1 [which would be
> 15 for my Ryzen 7 1700]. To fix this with bios, set to Typical Current
> Idle in the bios Advanced/AMD CBS menu.
>
> Using this comment to refine my google search terms I quickly found
> <https://community.amd.com/thread/225795> and
> <https://bugzilla.kernel.org/show_bug.cgi?id=196683> which confirmed
> that my Ryzen linux lockup problems are common, but could be
> addressed by the two methods (kernel parameter or BIOS option) he discussed
> with some users finding that only the latter option worked, but they 
> typically
> needed the absolutely latest BIOS update to get access to that option.
>
> So as a result I plan to rebuild my Debian kernel with CONFIG_RCU_NOCB_CPU=y 
> and set
> the kernel parameter rcu_nocbs=0-15 (15 is appropriate for Ryzen 7 1700)
> for that custom kernel.  And I hope that will finally solve these lockups[...]

Status report:

We have been running this system with the above combination of custom kernel and
special kernel parameter that is recommended to help stabilize all
Linux Ryzen systems.  This uptime test has been run with our standard
two-user mode where Barbara accesses her desktop on this box with an
X-terminal (X server run and displayed on separate low-end hardware)
while I simultaneously access my desktop on this box with a local X
server (which uses the AMD RX 550 graphics card attached to this box
to display results).

The uptime is now 5+ days without a lockup with us both being quite
active users.  This result exceeds our previous uptime record for this
hardware in our standard two-user mode by roughly a factor of two.  So
this is a promising first result which makes me believe I am finally
(after 6 months of uncertainty) on the right track for stabilizing
this box.  However, we plan to continue this two-user uptime
experiment until the system locks up or ~40 days uptime (whichever
comes first) to see whether more steps are required to completely
stabilize this box or whether running this custom kernel with special
kernel parameter is all that is necessary to stabilize this particular
Ryzen system.

Alan
__________________________
Alan W. Irwin

Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________

Linux-powered Science
__________________________