[Bug 219399] System panics after several hours of 14-threads-compilation orgies using poudriere on AMD Ryzen...

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=219399

--- Comment #208 from Don Lewis <truckman at FreeBSD.org> ---
The second WHERE 0x04fffffff000 is someleftover debug stuff from an earlier
version of the program that executed some more complicated code.  I needed to
debug that code in a harmless spot in memory so that I could get that code
working right.

Even without cpuset, I think I eventually got it to crash at 0x7fffffffff40,
probably because it migrated to CPU 0 on it's own or an interrupt finally
caught it at that address, which would be less frequent on the other cores. 
There might have been other stuff running on the system at the same time.

If you pin it to some other CPU, do you see system time spike up when it gets
to 0x7fffffffff40?  I wonder if it's getting kicked into a trap handler on
every iteration when it gets to that address.  That and an interrupt happening
at the same time might sent it off into the weeds.

It would also be interesting to see the results on non-Ryzen hardware.

-- 
You are receiving this mail because:
You are the assignee for the bug.