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

bugzilla-noreply at freebsd.org bugzilla-noreply at freebsd.org
Fri Aug 4 16:19:33 UTC 2017


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

--- Comment #204 from Don Lewis <truckman at FreeBSD.org> ---
Created attachment 185022
  --> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=185022&action=edit
program to cause Ryzen hang/reboot on tweaked FreeBSD by executing code in high
memory

If you modify the FreeBSD kernel to lower the shared page, but leave sv_maxuser
at its original value so that a user program can mmap the page at
0x7ffffffff000, the  attached program will fill that page with RET instructions
and perform calls to those in a loop.  I have not observed any issues with
calls to the RET instructions at 0x7fffffffff3f or below.  When the RET
instruction at 0x7fffffffff40 is executed, my machine will typically silently
reboot without a panic message, or it will sometimes hang with the screen
blanked.  This test is the most sensitive on core 0, which handles most
interrupts, so run under "cpuset -l 0" for best results.  The problem appears
to be triggered when the RET instruction is interrupted.

It is possible to load and execute arbitrary code for other experiments.

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


More information about the freebsd-bugs mailing list